diff --git a/.gitattributes b/.gitattributes index ded000fc84d107e25ba7136cff61a5339a729743..1717bfdb3e8a1e7d014aea90382c32003ac53235 100644 --- a/.gitattributes +++ b/.gitattributes @@ -113,3 +113,10 @@ assets/lightx2v-8/17.mp4 filter=lfs diff=lfs merge=lfs -text assets/lightx2v-8/18.mp4 filter=lfs diff=lfs merge=lfs -text assets/lightx2v-8/19.mp4 filter=lfs diff=lfs merge=lfs -text assets/lightx2v-8/20.mp4 filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_1_1.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_1_2.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_1_3.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_2_1.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_2_2.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_2_3.jpg filter=lfs diff=lfs merge=lfs -text +assets/ref2va-inputs/ref2va_test_4_1.jpg filter=lfs diff=lfs merge=lfs -text diff --git a/REPORT.md b/REPORT.md index f731e473f19175cb8e8a6d9857ce90c44fef8ab6..d4f9ce8592c3ba445e21fdf32da71d175e8fe950 100644 --- a/REPORT.md +++ b/REPORT.md @@ -1,101 +1,96 @@ # Sol-H3 / HyperFlow 测试结果 -HSG 与 Poly 原始测速均已完成，共 46 组实测配置。效果区目前为旧版 5 秒原始社区提示词结果，未按 H3 结构重写；15 秒修正版正在准备。 +HSG、Poly 共 57 条独立实测记录。全部来自本次 Slurm 推理输出；GitHub B300 历史数字单独标注。 -[HF 对比 Space（公开）](https://huggingface.co/spaces/Lawrence-cj/sol-h3-hyperflow-comparison) 包含 20 个相同场景、4 组 checkpoint，共 80 段新生成的带音频视频。 +[公开对比 Space](https://huggingface.co/spaces/Lawrence-cj/sol-h3-hyperflow-comparison) · [逐次实测计时 CSV](benchmark-evidence.csv) · [完整公开结果 JSON](data.json) -## HSG 速度 +## 15 秒效果样片 -硬件实测为 NVIDIA GB200，单卡 189471 MiB，4 卡/节点；8 卡使用两节点。1 卡测试实际使用 1 个 GPU 进程，按集群规则分配 4 卡。预热一次后测试三次，取中位数；计入文本编码、DiT 和音视频 VAE，不含权重加载、预热编译、MP4 编码。 +20 个场景 × 4 个 H3 checkpoint，共 80 段重新生成的音视频。全部使用 1344×768、362 帧、24 FPS，目标时长 15 秒；同一场景的 prompt 和 seed 完全相同。效果区使用 HSG 4 卡、BF16、SOL/BSA。 -| Checkpoint / 计算 | 使用卡数 | 5 秒 / 124f | 10 秒 / 243f | 15 秒 / 362f | -|---|---:|---:|---:|---:| -| hyperflow-8 / BF16 | 1 | 25.560 s | 71.960 s | 137.426 s | -| hyperflow-8 / BF16 | 4 | 5.519 s | 13.347 s | 24.273 s | -| hyperflow-8 / BF16 | 8 | 3.224 s | 7.476 s | 13.338 s | -| hyperflow-8 / MXFP8 | 8 | 2.759 s | 6.552 s | 12.065 s | -| fasth3-4 / BF16 | 1 | 14.144 s | 37.886 s | 71.524 s | -| fasth3-4 / BF16 | 4 | 2.996 s | 7.041 s | 12.621 s | -| fasth3-4 / BF16 | 8 | 1.666 s | 3.754 s | 6.676 s | -| lightx2v-8 / BF16 | 4 | 5.536 s | — | — | -| lightx2v-4 / BF16 | 4 | 2.964 s | — | — | +旧版 5 秒样片使用了普通社区文本，未整理为 H3 格式。本版已按 H3 T2VA 的 `integrated_multimodal_description`、`overall_soundscape`、`non_diegetic_music` 三字段整理；首镜头为 `[Shot 1]`，切镜时间在 15 秒内，原文中的明确台词保留并加上说话人和 `` 标签。社区原文与本次实际输入分别展示，新增镜头节奏、环境音等明确属于本次整理，并不声称是作者的原版结构化提示词。 -1 卡采用 dense，4/8 卡采用 SOL/BSA、INT8 QKV 通信、FP8 输出通信和并行 VAE。主效果对比使用 4 卡 BF16；MXFP8 仅列为独立速度配置。这些近似运算可能改变画面和音频。 +20 条实际输入通过字段顺序、镜头编号、时间范围、引用完整性、语音标签及哈希检查；80 段输出逐一核对输入哈希、seed、步数、362 帧、几何尺寸、FPS、双声道音频，并完整解码音视频。此检查证明数据对应关系和媒体完整性，不等同于模型逐条服从了全部镜头和台词。 -## 与 GitHub 原速度比较 +## 计时与环境 -原表测于 B300。本次使用默认 quiet-room prompt、seed 20260903；原表未提供完整 prompt。因此这是跨硬件参照比较，不是完全同条件复现。 +速度表使用独立的 [quiet-room 测速 prompt](benchmark_prompt.txt)，seed 20260903。每种配置预热一次、测三次取中位数，包含文本编码、DiT、视频 VAE 和音频 VAE；排除权重加载、预热编译和 MP4 编码。效果卡片则显示该复杂 prompt 的单次耗时，可能含新形状首次编译，因此不与预热后的测速表混用。 -| FastH3 使用卡数 | 时长 | 本次 GB200 | GitHub B300 | 耗时差异 | -|---:|---:|---:|---:|---:| -| 1 | 5s | 14.144s | 13.745s | +2.90% | -| 1 | 10s | 37.886s | 37.813s | +0.19% | -| 1 | 15s | 71.524s | 52.260s | +36.86% | -| 4 | 5s | 2.996s | 2.918s | +2.67% | -| 4 | 10s | 7.041s | 6.993s | +0.69% | -| 4 | 15s | 12.621s | 12.542s | +0.63% | -| 8 | 5s | 1.666s | 1.653s | +0.79% | -| 8 | 10s | 3.754s | 3.732s | +0.59% | -| 8 | 15s | 6.676s | 6.612s | +0.97% | +HSG 和 Poly 都重新建立了独立推理环境，PyTorch 2.10.0+cu130、CUDA toolkit 13.0.2、Triton 3.6.0、cuDNN 9.15.1.9、Transformers 5.8.1；Diffusers 与 cuDNN frontend 均固定到相同提交。HSG Python 3.12.13，Poly Python 3.12.3。硬件均报告 NVIDIA GB200，4 卡/节点，8 卡跨两节点；单卡测试只启用一个 GPU 进程。 -9 组 FastH3 配置中，8 组比原参照慢 0.19–2.90%；单卡 15 秒慢 36.86%，不能视为全部速度一致。本次没有对这一长视频性能差异作进一步瓶颈归因。 +## HSG 本次实测 -## Poly 速度 +| Checkpoint / 计算 | 使用卡数 | 5 秒 / 124f | 10 秒 / 243f | 15 秒 / 362f | +|---|---:|---:|---:|---:| +| fasth3-4 / BF16 | 1 | 14.144 s | 37.886 s | 71.524 s | +| fasth3-4 / BF16 | 4 | 2.996 s | 7.041 s | 12.621 s | +| fasth3-4 / BF16 | 8 | 1.666 s | 3.754 s | 6.676 s | +| hyperflow-8 / MXFP8 | 1 | 22.428 s | 65.486 s | 129.516 s | +| hyperflow-8 / MXFP8 | 4 | 4.607 s | 11.552 s | 21.511 s | +| hyperflow-8 / MXFP8 | 8 | 2.759 s | 6.552 s | 12.065 s | +| hyperflow-8 / BF16 | 1 | 25.560 s | 71.960 s | 137.426 s | +| hyperflow-8 / BF16 | 4 | 5.519 s | 13.347 s | 24.273 s | +| hyperflow-8 / BF16 | 8 | 3.224 s | 7.476 s | 13.338 s | +| lightx2v-4 / BF16 | 4 | 2.964 s | — | — | +| lightx2v-8 / BF16 | 4 | 5.536 s | — | — | -Poly 同为每节点 4 张 NVIDIA GB200，8 卡使用两个节点；Slurm 无 GPU GRES，按整节点分配。实际推理使用 1、4 或 8 个 GPU 进程。计时、prompt、seed、权重哈希和推理依赖版本与 HSG 一致；Poly Python 为 3.12.3，HSG 为 3.12.13，驱动及集群网络环境也有差异。 +## POLY 本次实测 | Checkpoint / 计算 | 使用卡数 | 5 秒 / 124f | 10 秒 / 243f | 15 秒 / 362f | |---|---:|---:|---:|---:| -| hyperflow-8 / BF16 | 1 | 24.940 s | 70.312 s | 135.545 s | -| hyperflow-8 / BF16 | 4 | 5.514 s | 13.581 s | 24.797 s | -| hyperflow-8 / BF16 | 8 | 3.122 s | 7.369 s | 13.218 s | -| hyperflow-8 / MXFP8 | 8 | 2.714 s | 6.550 s | 12.104 s | | fasth3-4 / BF16 | 1 | 13.841 s | 37.857 s | 71.750 s | | fasth3-4 / BF16 | 4 | 3.077 s | 7.225 s | 12.986 s | | fasth3-4 / BF16 | 8 | 1.631 s | 3.787 s | 6.707 s | -| lightx2v-8 / BF16 | 4 | 5.112 s | — | — | +| hyperflow-8 / MXFP8 | 1 | 21.857 s | 63.848 s | — | +| hyperflow-8 / MXFP8 | 4 | 4.649 s | 11.837 s | 22.035 s | +| hyperflow-8 / MXFP8 | 8 | 2.714 s | 6.550 s | 12.104 s | +| hyperflow-8 / BF16 | 1 | 24.940 s | 70.312 s | 135.545 s | +| hyperflow-8 / BF16 | 4 | 5.514 s | 13.581 s | 24.797 s | +| hyperflow-8 / BF16 | 8 | 3.122 s | 7.369 s | 13.218 s | | lightx2v-4 / BF16 | 4 | 2.928 s | — | — | +| lightx2v-8 / BF16 | 4 | 5.112 s | — | — | + +## GitHub 历史参照（B300） + +以下数字从已公布的 Sol-H3 源码快照引用，未当作本次实测。原表没有公开完整测速 prompt，因此属于跨硬件参照，不是严格同条件复现。 + +| FastH3 4-step BF16 卡数 | 5 秒 | 10 秒 | 15 秒 | +|---:|---:|---:|---:| +| 1 | 13.745 s | 37.813 s | 52.260 s | +| 4 | 2.918 s | 6.993 s | 12.542 s | +| 8 | 1.653 s | 3.732 s | 6.612 s | + +HSG 单卡 15 秒为 71.524 秒，相对 GitHub 的 52.260 秒慢 36.86%。该差距在两套新环境中都出现，不能归因为本次忘了重新配环境；也不能在没有 B300 分阶段记录时，把原因确定为某一项依赖或硬件。 + +POLY 单卡 15 秒为 71.750 秒，相对 GitHub 的 52.260 秒慢 37.29%。该差距在两套新环境中都出现，不能归因为本次忘了重新配环境；也不能在没有 B300 分阶段记录时，把原因确定为某一项依赖或硬件。 + +## 实际启用的加速 + +- BF16 主测试开启融合 RMSNorm/调制/残差、QKNorm/RoPE、SwiGLU，以及 AdaLN 预计算。HyperFlow 缓存完整 `(t,r)` 条件。 +- 多卡启用 SOL/BSA 与 Ulysses；实际记录每次请求的稀疏注意力调用和 INT8 QKV / FP8 输出通信。部分层与前几个步骤按既定策略保留 dense，并非漏开稀疏。 +- 单卡使用发布实现的 dense 路径。视频 VAE decoder 编译；多卡额外启用视频 VAE 分片与批处理。音频 VAE 未做同样的并行改造。 +- MXFP8 作为独立配置测试，量化第 2–46 层共 180 个线性层；与 BF16 主表分别列出。首次仅测了 8 卡，现在补测 1/4 卡；不能把最初 BF16 主表描述为所有可选量化均已开启。 +- 融合 SwiGLU 的大张量地址已改为 64 位，独立 GB200 边界数值检查与完整单卡 15 秒推理通过。 + +## HyperFlow 8-step 与 LightX2V 8-step + +两者每个视频都做 8 次联合音视频 DiT 前向，但时间条件与采样网格不同。 + +| 项目 | HyperFlow v1.0 | 本次 LightX2V 8-step v1.0 | +|---|---|---| +| 时间条件 | 当前时间 t 与区间终点 r | 单时间条件 | +| 时间嵌入 | 独立两套嵌入，gate 0.25 | 原始 H3 时间嵌入 | +| 网格 | 权重元数据内的固定非均匀网格 | 对应推理配置的均匀原始网格 | +| video / audio shift | 12 / 3 | 6 / 3 | +| LoRA rank / alpha | 256 / 256 | 128 / 8 | + +HyperFlow 保留原 Euler 状态更新；额外设计主要在两时间条件、时间嵌入和对应训练网格。接入时同时安装这些逻辑，不能仅把它作为普通 LoRA 换文件。每种 adapter 在新推理实例中加载，避免 LoRA 累加；该验证不代表支持已融合模型的原地热切换。 + +GPU 效果验证覆盖 T2VA。FL2VA/Ref2VA 仅完成已有结构和数值检查，没有新增 GPU 效果验证。HyperFlow 所述训练画布为 124 帧，15 秒属于更长时长的实际测试。 + +## 来源与复现 -### Poly 与 HSG / GitHub 的对比 - -| Checkpoint | 卡数 | 生成时长 | Poly 耗时 | 相对 HSG | 相对 GitHub B300 | -|---|---:|---:|---:|---:|---:| -| hyperflow-8 | 1 | 5s | 24.940s | -2.43% | — | -| hyperflow-8 | 1 | 10s | 70.312s | -2.29% | — | -| hyperflow-8 | 1 | 15s | 135.545s | -1.37% | — | -| hyperflow-8 | 4 | 5s | 5.514s | -0.08% | — | -| hyperflow-8 | 4 | 10s | 13.581s | +1.75% | — | -| hyperflow-8 | 4 | 15s | 24.797s | +2.16% | — | -| hyperflow-8 | 8 | 5s | 3.122s | -3.18% | — | -| hyperflow-8 | 8 | 10s | 7.369s | -1.43% | — | -| hyperflow-8 | 8 | 15s | 13.218s | -0.90% | — | -| fasth3-4 | 1 | 5s | 13.841s | -2.14% | +0.70% | -| fasth3-4 | 1 | 10s | 37.857s | -0.08% | +0.12% | -| fasth3-4 | 1 | 15s | 71.750s | +0.32% | +37.29% | -| fasth3-4 | 4 | 5s | 3.077s | +2.69% | +5.43% | -| fasth3-4 | 4 | 10s | 7.225s | +2.61% | +3.31% | -| fasth3-4 | 4 | 15s | 12.986s | +2.89% | +3.54% | -| fasth3-4 | 8 | 5s | 1.631s | -2.06% | -1.28% | -| fasth3-4 | 8 | 10s | 3.787s | +0.87% | +1.47% | -| fasth3-4 | 8 | 15s | 6.707s | +0.46% | +1.44% | - -Poly FastH3 除单卡 15 秒外，相对原 B300 耗时差异为 -1.28% 至 +5.43%。单卡 15 秒耗时 71.750 秒，比原参照慢 37.29%；两个 GB200 集群都复现了这一差距。 - -正值表示耗时更长。页面的 80 段效果样片保持使用 HSG 的 4 卡 BF16 结果；Poly 的 23 组测速分别生成了带音频视频并完成全量解码校验。 - -## Checkpoint 适配和验证 - -- 新文件是 HyperFlow 8-step LoRA。其 rank/alpha 为 256/256、gate 为 0.25，采用两套时间嵌入与固定 sigma 网格，video/audio shift 为 12/3；不能仅按普通 LoRA 换文件。已加入自动识别、完整目标校验、两时间 AdaLN 缓存和 Sol-H3 加速。 -- 每次切换已融合的 adapter 使用新推理实例；HSG 实测冷加载约 6–7 分钟，权重读取与模型初始化均不计入表中的预热后推理耗时。此结果不代表支持原地热切换。 -- 两时间缓存覆盖全部 50 层、8 个步骤，约占 0.433 GiB，并释放每卡约 24.234 GiB 的投影权重。 -- 修复了原有 SwiGLU 的 32 位地址溢出。独立 GB200 大张量测试复现原版边界后读错，64 位实现通过数值检查；修复后的单卡 15 秒完整推理已通过。 -- HSG 和 Poly 的新环境均通过 9 项 CPU 数值/流水线测试；所有最终作业均以 0:0 退出。80 段样片与 46 个测速视频均完成帧数、分辨率、FPS、双声道音频及全量解码检查。 -- 已查看全部 80 段视频的预览帧，并检查四组模型第一个样例的多时间点画面；浏览器验证四路完整播放、步数筛选、20 场景切换和窄屏布局。画质优劣仍需逐段结合动作和音频判断。 -- GPU 测试覆盖 T2VA。FL2VA / Ref2VA 仅完成加载结构和数值测试，尚未做 GPU 端到端验证；10/15 秒也超出了 HyperFlow 所述的 124 帧训练范围。 - -## 代码与记录 - -- 本地代码提交：`bcce779cbe6bd7aa46d92adb4182136fc90a5724`，分支 `codex/hyperflow-h3`。已部署到本次 HSG 和 Poly 目录，未推到公开 Sana 仓库。 -- 所有输出保留 checkpoint SHA256、基础模型 revision、seed、NFE、集群、卡数和推理参数，见 [data.json](data.json)。 -- Prompt 来源：[BeatAPI Prompt Gallery](https://github.com/BeatAPI/awesome-minimax-h3-prompts)，逐条保留原作者链接；统一为 1344×768、124 帧、24 FPS、seed 42–61。时长/横竖屏调整记录在每条 prompt 下。 -- 基础代码与原始参照：[NVlabs/Sana Sol-H3](https://github.com/NVlabs/Sana/tree/sol-engine/models/minimax_h3/Sol-H3)。 +- 基础代码：[NVlabs/Sana Sol-H3](https://github.com/NVlabs/Sana/tree/sol-engine/models/minimax_h3/Sol-H3)，本次适配提交 `bcce779cbe6bd7aa46d92adb4182136fc90a5724`。 +- 社区提示词：[BeatAPI awesome-minimax-h3-prompts](https://github.com/BeatAPI/awesome-minimax-h3-prompts)，逐条保留原作者链接和原始文本。 +- 比较仅包含 H3：HyperFlow 8-step、FastH3 4-step、LightX2V 8-step v1.0、LightX2V 4-step v1.2。 +- 公开数据包含 checkpoint SHA256、基础模型 revision、prompt/seed、NFE、GPU 数、精度、逐次计时和媒体哈希；内部节点与调度日志保留在本地实验记录。 diff --git a/app.js b/app.js index a3d30258d94795edffe104300b71e062f3e0efeb..fcb79a6a3448aa5350c42e2c7a99e2306fd69101 100644 --- a/app.js +++ b/app.js @@ -2,7 +2,7 @@ let data, sampleIndex = 0, stepFilter = "all"; const $ = s => document.querySelector(s); const text = (tag, value, cls) => { const e = document.createElement(tag); e.textContent = value; if (cls) e.className = cls; return e; }; -const videos = () => [...document.querySelectorAll("video")]; +const videos = () => [...document.querySelectorAll("#gallery video")]; function renderSample() { videos().forEach(v => v.pause()); const s = data.samples[sampleIndex]; @@ -12,6 +12,7 @@ function renderSample() { $("#scene-seed").textContent = `SEED ${s.seed}`; $("#position").textContent = `${sampleIndex + 1} / ${data.samples.length}`; $("#prompt").textContent = s.prompt; + if ($("#source-prompt")) $("#source-prompt").textContent = s.source_prompt; $("#adaptations").textContent = (s.adaptations || []).join(" "); const source = document.createElement("a"); source.href = s.source.url; source.target = "_blank"; source.rel = "noreferrer"; source.textContent = `${s.source.name} · 原始来源 ↗`; $("#attribution").replaceChildren(source, text("span", ` · BeatAPI Prompt Gallery · 原示例 ${s.source_duration}`)); diff --git a/assets/ref2va-inputs/ref2va_test_1_1.jpg b/assets/ref2va-inputs/ref2va_test_1_1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..418ec0b0cfb38013a1a2dd251be59eff92dd036a --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_1_1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1665c52d9e20ace593cbc30d1e17b5cf996ad25a59c7f1890d636711da93dbd3 +size 344933 diff --git a/assets/ref2va-inputs/ref2va_test_1_2.jpg b/assets/ref2va-inputs/ref2va_test_1_2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6d2b914f2b64da4d9fd9114900efde717dbd6da9 --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_1_2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df77cc201252e90cb828132863d4126c4d62d512f79299ec2651bcc9be8c5b1f +size 396459 diff --git a/assets/ref2va-inputs/ref2va_test_1_3.jpg b/assets/ref2va-inputs/ref2va_test_1_3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0f6af6bea53f9466f7c6af5dfd4b3b7c59751bd --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_1_3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82cdb7a20d6ec0d7c6c396cc8e99d3f5fe49a2f67f7c3e5e75d45102f4a91062 +size 390923 diff --git a/assets/ref2va-inputs/ref2va_test_2_1.jpg b/assets/ref2va-inputs/ref2va_test_2_1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6a8a089be477067a9098f1c1e2ca5271c28ec52d --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_2_1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3808f06a4b21910dcc04a4f62f6fe95714d0c9773cd9f81fd67695d4eb924373 +size 383601 diff --git a/assets/ref2va-inputs/ref2va_test_2_2.jpg b/assets/ref2va-inputs/ref2va_test_2_2.jpg new file mode 100644 index 0000000000000000000000000000000000000000..68ff4fbfbe1befaede73b90e7eb1b6cbedfc9585 --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_2_2.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b62cf2f93c256975c80fd58869216a819749a8f3aab4bc03b396b6147f0848eb +size 278692 diff --git a/assets/ref2va-inputs/ref2va_test_2_3.jpg b/assets/ref2va-inputs/ref2va_test_2_3.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c2c772b5e34484a5b52b967058251c301dbbe8b9 --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_2_3.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1aac91407587e4457ad191e3867e5f24dd878870ca4bbafc40ca24c605c86df +size 194821 diff --git a/assets/ref2va-inputs/ref2va_test_4_1.jpg b/assets/ref2va-inputs/ref2va_test_4_1.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04577f97d769757f614508ddfb7687aabf0a23c1 --- /dev/null +++ b/assets/ref2va-inputs/ref2va_test_4_1.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b54a682135c5e60c3584df15754c8d54177738686a0778ca89f0a326fe50bbd +size 290315 diff --git a/assets/v2-15s/fasth3-4/01.jpg b/assets/v2-15s/fasth3-4/01.jpg new file mode 100644 index 0000000000000000000000000000000000000000..642e96fbfdc7a443fd504ef28c2f27102075207f Binary files /dev/null and b/assets/v2-15s/fasth3-4/01.jpg differ diff --git a/assets/v2-15s/fasth3-4/02.jpg b/assets/v2-15s/fasth3-4/02.jpg new file mode 100644 index 0000000000000000000000000000000000000000..04452a46b911e9b90ec0420f0b9792ea7f0dee0c Binary files /dev/null and b/assets/v2-15s/fasth3-4/02.jpg differ diff --git a/assets/v2-15s/fasth3-4/03.jpg b/assets/v2-15s/fasth3-4/03.jpg new file mode 100644 index 0000000000000000000000000000000000000000..644706c80973e8ef43cb8ff1ba898e40df7c8b08 Binary files /dev/null and b/assets/v2-15s/fasth3-4/03.jpg differ diff --git a/assets/v2-15s/fasth3-4/04.jpg b/assets/v2-15s/fasth3-4/04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..035b053186812035da988532c919d6a953d4b001 Binary files /dev/null and b/assets/v2-15s/fasth3-4/04.jpg differ diff --git a/assets/v2-15s/fasth3-4/05.jpg b/assets/v2-15s/fasth3-4/05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..457d5b2afa5aedc1f0aee8508be196ea362deb18 Binary files /dev/null and b/assets/v2-15s/fasth3-4/05.jpg differ diff --git a/assets/v2-15s/fasth3-4/06.jpg b/assets/v2-15s/fasth3-4/06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bc3a68aaf39751a3d00fb160cb2f13cb1be44d75 Binary files /dev/null and b/assets/v2-15s/fasth3-4/06.jpg differ diff --git a/assets/v2-15s/fasth3-4/07.jpg b/assets/v2-15s/fasth3-4/07.jpg new file mode 100644 index 0000000000000000000000000000000000000000..18461bdbbecba6437e1bb378a9d788e7f0256c5b Binary files /dev/null and b/assets/v2-15s/fasth3-4/07.jpg differ diff --git a/assets/v2-15s/fasth3-4/08.jpg b/assets/v2-15s/fasth3-4/08.jpg new file mode 100644 index 0000000000000000000000000000000000000000..aed30767ee6ddff643e1018bf7b7feae66855992 Binary files /dev/null and b/assets/v2-15s/fasth3-4/08.jpg differ diff --git a/assets/v2-15s/fasth3-4/09.jpg b/assets/v2-15s/fasth3-4/09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..621e3ca078be0bdbaf595702ba4c0bf1a79e089f Binary files /dev/null and b/assets/v2-15s/fasth3-4/09.jpg differ diff --git a/assets/v2-15s/fasth3-4/10.jpg b/assets/v2-15s/fasth3-4/10.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1e4063550f9b3ba8af58695743589d424dce3555 Binary files /dev/null and b/assets/v2-15s/fasth3-4/10.jpg differ diff --git a/assets/v2-15s/fasth3-4/11.jpg b/assets/v2-15s/fasth3-4/11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e50988cfd744b2d9dddcb62e4f6af128e9bcce23 Binary files /dev/null and b/assets/v2-15s/fasth3-4/11.jpg differ diff --git a/assets/v2-15s/fasth3-4/12.jpg b/assets/v2-15s/fasth3-4/12.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ec6c2a8b51ad26ccc64335c355d54fd9942a6eb2 Binary files /dev/null and b/assets/v2-15s/fasth3-4/12.jpg differ diff --git a/assets/v2-15s/fasth3-4/13.jpg b/assets/v2-15s/fasth3-4/13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0ca17eb260da074b8e63f5d084e70eb2e0ba459 Binary files /dev/null and b/assets/v2-15s/fasth3-4/13.jpg differ diff --git a/assets/v2-15s/fasth3-4/14.jpg b/assets/v2-15s/fasth3-4/14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4ac2243227095edb9391efaccd5cf6807b611419 Binary files /dev/null and b/assets/v2-15s/fasth3-4/14.jpg differ diff --git a/assets/v2-15s/fasth3-4/15.jpg b/assets/v2-15s/fasth3-4/15.jpg new file mode 100644 index 0000000000000000000000000000000000000000..462ced78bbde22c2e68c4d30daf26fb22558fe81 Binary files /dev/null and b/assets/v2-15s/fasth3-4/15.jpg differ diff --git a/assets/v2-15s/fasth3-4/16.jpg b/assets/v2-15s/fasth3-4/16.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fd5b8615dc92ff6699023f6bf648773bf583fc0b Binary files /dev/null and b/assets/v2-15s/fasth3-4/16.jpg differ diff --git a/assets/v2-15s/fasth3-4/17.jpg b/assets/v2-15s/fasth3-4/17.jpg new file mode 100644 index 0000000000000000000000000000000000000000..11e543934d26936701116b033c5e67d31bfd02bc Binary files /dev/null and b/assets/v2-15s/fasth3-4/17.jpg differ diff --git a/assets/v2-15s/fasth3-4/18.jpg b/assets/v2-15s/fasth3-4/18.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fb12f368e26100949c77cd2535de96469f937b4f Binary files /dev/null and b/assets/v2-15s/fasth3-4/18.jpg differ diff --git a/assets/v2-15s/fasth3-4/19.jpg b/assets/v2-15s/fasth3-4/19.jpg new file mode 100644 index 0000000000000000000000000000000000000000..db6306379e35079b2edb2d9ec5d4410483fef30e Binary files /dev/null and b/assets/v2-15s/fasth3-4/19.jpg differ diff --git a/assets/v2-15s/fasth3-4/20.jpg b/assets/v2-15s/fasth3-4/20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ce8d4afa0ac874c4551f6fbba31c5bc88243d452 Binary files /dev/null and b/assets/v2-15s/fasth3-4/20.jpg differ diff --git a/assets/v2-15s/hyperflow-8/01.jpg b/assets/v2-15s/hyperflow-8/01.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5fe4d555002116821d98b62289c7f64070246dc8 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/01.jpg differ diff --git a/assets/v2-15s/hyperflow-8/02.jpg b/assets/v2-15s/hyperflow-8/02.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7b88f855af5078560158dc939e85cc43a3f5eb14 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/02.jpg differ diff --git a/assets/v2-15s/hyperflow-8/03.jpg b/assets/v2-15s/hyperflow-8/03.jpg new file mode 100644 index 0000000000000000000000000000000000000000..90195621cff2b38a99498231c1a08f5f29a9c8e2 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/03.jpg differ diff --git a/assets/v2-15s/hyperflow-8/04.jpg b/assets/v2-15s/hyperflow-8/04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c0f85f863abd1d21828292ad9523cf1fec1cdc85 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/04.jpg differ diff --git a/assets/v2-15s/hyperflow-8/05.jpg b/assets/v2-15s/hyperflow-8/05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e187d6a3906e99ddfdef129fe8cb0c8f6b6cd0a2 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/05.jpg differ diff --git a/assets/v2-15s/hyperflow-8/06.jpg b/assets/v2-15s/hyperflow-8/06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4d187b5e48a789c4c5b460d99e5675d81dd496ec Binary files /dev/null and b/assets/v2-15s/hyperflow-8/06.jpg differ diff --git a/assets/v2-15s/hyperflow-8/07.jpg b/assets/v2-15s/hyperflow-8/07.jpg new file mode 100644 index 0000000000000000000000000000000000000000..067ba5a79a04b05a7bbd46bcfabffbbf3d9e862e Binary files /dev/null and b/assets/v2-15s/hyperflow-8/07.jpg differ diff --git a/assets/v2-15s/hyperflow-8/08.jpg b/assets/v2-15s/hyperflow-8/08.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3b9c2e033a12b601007c56cbdbc52518d0a9f9aa Binary files /dev/null and b/assets/v2-15s/hyperflow-8/08.jpg differ diff --git a/assets/v2-15s/hyperflow-8/09.jpg b/assets/v2-15s/hyperflow-8/09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..273082f7a5fbb192875d710e4bea783e594bb16d Binary files /dev/null and b/assets/v2-15s/hyperflow-8/09.jpg differ diff --git a/assets/v2-15s/hyperflow-8/10.jpg b/assets/v2-15s/hyperflow-8/10.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2dd08969c93d8d92e0a1e3bf7c41b6363739395d Binary files /dev/null and b/assets/v2-15s/hyperflow-8/10.jpg differ diff --git a/assets/v2-15s/hyperflow-8/11.jpg b/assets/v2-15s/hyperflow-8/11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4f195a0d326e3c84c050a29fdff6b45de641aa63 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/11.jpg differ diff --git a/assets/v2-15s/hyperflow-8/12.jpg b/assets/v2-15s/hyperflow-8/12.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ca15ae369874fee59c31f589ae049feec668c0ad Binary files /dev/null and b/assets/v2-15s/hyperflow-8/12.jpg differ diff --git a/assets/v2-15s/hyperflow-8/13.jpg b/assets/v2-15s/hyperflow-8/13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..efd6d25d2a74cc7641b62ec008b7c46ecfdb274d Binary files /dev/null and b/assets/v2-15s/hyperflow-8/13.jpg differ diff --git a/assets/v2-15s/hyperflow-8/14.jpg b/assets/v2-15s/hyperflow-8/14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d87c327b7492599d886324bef176375d23c26502 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/14.jpg differ diff --git a/assets/v2-15s/hyperflow-8/15.jpg b/assets/v2-15s/hyperflow-8/15.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7bc5cc1f2ad2b684cad7a13d55a7878637b49d5d Binary files /dev/null and b/assets/v2-15s/hyperflow-8/15.jpg differ diff --git a/assets/v2-15s/hyperflow-8/16.jpg b/assets/v2-15s/hyperflow-8/16.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1a1d850d088fd94dbded2efa957afeabeff2b91a Binary files /dev/null and b/assets/v2-15s/hyperflow-8/16.jpg differ diff --git a/assets/v2-15s/hyperflow-8/17.jpg b/assets/v2-15s/hyperflow-8/17.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c60567845981a9f1d7fb82d83eb1f90df330c9fd Binary files /dev/null and b/assets/v2-15s/hyperflow-8/17.jpg differ diff --git a/assets/v2-15s/hyperflow-8/18.jpg b/assets/v2-15s/hyperflow-8/18.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e6af0bb0c165a0acc0013e9e2ba27f258b4b0817 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/18.jpg differ diff --git a/assets/v2-15s/hyperflow-8/19.jpg b/assets/v2-15s/hyperflow-8/19.jpg new file mode 100644 index 0000000000000000000000000000000000000000..2194fd147e12b1b3416a94389c76efb188e80bc1 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/19.jpg differ diff --git a/assets/v2-15s/hyperflow-8/20.jpg b/assets/v2-15s/hyperflow-8/20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5fbedb0478bfca9b1212900a13ac83cb77b85d09 Binary files /dev/null and b/assets/v2-15s/hyperflow-8/20.jpg differ diff --git a/assets/v2-15s/lightx2v-4/01.jpg b/assets/v2-15s/lightx2v-4/01.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c97564391b71b4e2191acfa6d0a642560bead935 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/01.jpg differ diff --git a/assets/v2-15s/lightx2v-4/02.jpg b/assets/v2-15s/lightx2v-4/02.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7975bccc71a01d3c4eac44e7124f72953e6d8b27 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/02.jpg differ diff --git a/assets/v2-15s/lightx2v-4/03.jpg b/assets/v2-15s/lightx2v-4/03.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cf90859b1f9d377a4ef7b0e42947e5cf064ef817 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/03.jpg differ diff --git a/assets/v2-15s/lightx2v-4/04.jpg b/assets/v2-15s/lightx2v-4/04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4a4f9b208c48aa82a5d5745b9b29147ed0ae1093 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/04.jpg differ diff --git a/assets/v2-15s/lightx2v-4/05.jpg b/assets/v2-15s/lightx2v-4/05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..48465a2d1aa9d1b7494242b86aec318093e7fa54 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/05.jpg differ diff --git a/assets/v2-15s/lightx2v-4/06.jpg b/assets/v2-15s/lightx2v-4/06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f248aa91768e396cb716c3d020f16610b1bc6a30 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/06.jpg differ diff --git a/assets/v2-15s/lightx2v-4/07.jpg b/assets/v2-15s/lightx2v-4/07.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3ddca87ce2581eb3a93ddfe9594f5e784d0a2da0 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/07.jpg differ diff --git a/assets/v2-15s/lightx2v-4/08.jpg b/assets/v2-15s/lightx2v-4/08.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6a1c21380252578ddfed4dbe316a3a70fc32ae03 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/08.jpg differ diff --git a/assets/v2-15s/lightx2v-4/09.jpg b/assets/v2-15s/lightx2v-4/09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9399d04c62ef1e96dbb72fd4df5a41db85b43454 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/09.jpg differ diff --git a/assets/v2-15s/lightx2v-4/10.jpg b/assets/v2-15s/lightx2v-4/10.jpg new file mode 100644 index 0000000000000000000000000000000000000000..781f8c469a3244dbc1565c646c8a469c004ff38a Binary files /dev/null and b/assets/v2-15s/lightx2v-4/10.jpg differ diff --git a/assets/v2-15s/lightx2v-4/11.jpg b/assets/v2-15s/lightx2v-4/11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ead21825f76cc5015145e84e74d4cffb60d30191 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/11.jpg differ diff --git a/assets/v2-15s/lightx2v-4/12.jpg b/assets/v2-15s/lightx2v-4/12.jpg new file mode 100644 index 0000000000000000000000000000000000000000..57c013a4ebf170eaa3dcc65c76fe0226aa8dac2d Binary files /dev/null and b/assets/v2-15s/lightx2v-4/12.jpg differ diff --git a/assets/v2-15s/lightx2v-4/13.jpg b/assets/v2-15s/lightx2v-4/13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a31332851a835711f7ced84d5541aff71cd29b0e Binary files /dev/null and b/assets/v2-15s/lightx2v-4/13.jpg differ diff --git a/assets/v2-15s/lightx2v-4/14.jpg b/assets/v2-15s/lightx2v-4/14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dd7368dc45d7fca44f47e7ce59d4d4fb1b4376db Binary files /dev/null and b/assets/v2-15s/lightx2v-4/14.jpg differ diff --git a/assets/v2-15s/lightx2v-4/15.jpg b/assets/v2-15s/lightx2v-4/15.jpg new file mode 100644 index 0000000000000000000000000000000000000000..880ddc58953158e903e1d27c06bbdab3fb004c45 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/15.jpg differ diff --git a/assets/v2-15s/lightx2v-4/16.jpg b/assets/v2-15s/lightx2v-4/16.jpg new file mode 100644 index 0000000000000000000000000000000000000000..d81590e941a586219c9336fcaa5a60c4adb7b8ca Binary files /dev/null and b/assets/v2-15s/lightx2v-4/16.jpg differ diff --git a/assets/v2-15s/lightx2v-4/17.jpg b/assets/v2-15s/lightx2v-4/17.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b3229a6bc5c00bbf3499107a73ecb838ae2de37f Binary files /dev/null and b/assets/v2-15s/lightx2v-4/17.jpg differ diff --git a/assets/v2-15s/lightx2v-4/18.jpg b/assets/v2-15s/lightx2v-4/18.jpg new file mode 100644 index 0000000000000000000000000000000000000000..bebc70e2315342a2ca701f03ca00ea69238dfe3b Binary files /dev/null and b/assets/v2-15s/lightx2v-4/18.jpg differ diff --git a/assets/v2-15s/lightx2v-4/19.jpg b/assets/v2-15s/lightx2v-4/19.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7a2e29631f2699e63463cd9fae4897939e879010 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/19.jpg differ diff --git a/assets/v2-15s/lightx2v-4/20.jpg b/assets/v2-15s/lightx2v-4/20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..652d98949a23d0f7eb512434a4f0f9841c0cae09 Binary files /dev/null and b/assets/v2-15s/lightx2v-4/20.jpg differ diff --git a/assets/v2-15s/lightx2v-8/01.jpg b/assets/v2-15s/lightx2v-8/01.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cfa3defdb040e10b0de31a6f7fc7ba8bd02786e3 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/01.jpg differ diff --git a/assets/v2-15s/lightx2v-8/02.jpg b/assets/v2-15s/lightx2v-8/02.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4096dbf39c3ef7269c7bf45065dcf28cdb7dcb67 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/02.jpg differ diff --git a/assets/v2-15s/lightx2v-8/03.jpg b/assets/v2-15s/lightx2v-8/03.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b1838f493c0aa5457a13a44bfb5633bb0b0d3837 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/03.jpg differ diff --git a/assets/v2-15s/lightx2v-8/04.jpg b/assets/v2-15s/lightx2v-8/04.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eca773c5449132e50e704096387dcaefe81f48bc Binary files /dev/null and b/assets/v2-15s/lightx2v-8/04.jpg differ diff --git a/assets/v2-15s/lightx2v-8/05.jpg b/assets/v2-15s/lightx2v-8/05.jpg new file mode 100644 index 0000000000000000000000000000000000000000..48a3f3c663c84ad9902cf643807b48f2b1b1a9d0 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/05.jpg differ diff --git a/assets/v2-15s/lightx2v-8/06.jpg b/assets/v2-15s/lightx2v-8/06.jpg new file mode 100644 index 0000000000000000000000000000000000000000..af1fe7aa2ab7d0ec2a32366e49fbb96ee26c8aca Binary files /dev/null and b/assets/v2-15s/lightx2v-8/06.jpg differ diff --git a/assets/v2-15s/lightx2v-8/07.jpg b/assets/v2-15s/lightx2v-8/07.jpg new file mode 100644 index 0000000000000000000000000000000000000000..1890a571fbbeaba2e0bad806c58092d988cc326c Binary files /dev/null and b/assets/v2-15s/lightx2v-8/07.jpg differ diff --git a/assets/v2-15s/lightx2v-8/08.jpg b/assets/v2-15s/lightx2v-8/08.jpg new file mode 100644 index 0000000000000000000000000000000000000000..217e91bc3fedb15bac81d0ca104d8c27a4794c0a Binary files /dev/null and b/assets/v2-15s/lightx2v-8/08.jpg differ diff --git a/assets/v2-15s/lightx2v-8/09.jpg b/assets/v2-15s/lightx2v-8/09.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6a553a90a05ba7e45d9582b7575e6b4be942cb24 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/09.jpg differ diff --git a/assets/v2-15s/lightx2v-8/10.jpg b/assets/v2-15s/lightx2v-8/10.jpg new file mode 100644 index 0000000000000000000000000000000000000000..829b9165ecda19ccddbaef6e83b8bf09c50225ff Binary files /dev/null and b/assets/v2-15s/lightx2v-8/10.jpg differ diff --git a/assets/v2-15s/lightx2v-8/11.jpg b/assets/v2-15s/lightx2v-8/11.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ea695fec71ef74559411b0002010f729fdbfc41d Binary files /dev/null and b/assets/v2-15s/lightx2v-8/11.jpg differ diff --git a/assets/v2-15s/lightx2v-8/12.jpg b/assets/v2-15s/lightx2v-8/12.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7127b2f36d2e0ef5a35946aa32cc347da81bc518 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/12.jpg differ diff --git a/assets/v2-15s/lightx2v-8/13.jpg b/assets/v2-15s/lightx2v-8/13.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cc3a64d3a18ebf64f529d2c8c4e0150c74f09a35 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/13.jpg differ diff --git a/assets/v2-15s/lightx2v-8/14.jpg b/assets/v2-15s/lightx2v-8/14.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c9557e1201137bc0145203f91324039eec3fcff1 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/14.jpg differ diff --git a/assets/v2-15s/lightx2v-8/15.jpg b/assets/v2-15s/lightx2v-8/15.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0d85f27277bcdffa6682931594f3c4f1730963a1 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/15.jpg differ diff --git a/assets/v2-15s/lightx2v-8/16.jpg b/assets/v2-15s/lightx2v-8/16.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6f1dac39712f0e9e30b4aa2233f35f4fe6b949f3 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/16.jpg differ diff --git a/assets/v2-15s/lightx2v-8/17.jpg b/assets/v2-15s/lightx2v-8/17.jpg new file mode 100644 index 0000000000000000000000000000000000000000..7403d7d79db5fb397ff22ae3825e54917c7a36fd Binary files /dev/null and b/assets/v2-15s/lightx2v-8/17.jpg differ diff --git a/assets/v2-15s/lightx2v-8/18.jpg b/assets/v2-15s/lightx2v-8/18.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0576be2b9247af60017e76ff861eaad1bb6019e5 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/18.jpg differ diff --git a/assets/v2-15s/lightx2v-8/19.jpg b/assets/v2-15s/lightx2v-8/19.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9687e77f671d7f041d1cbb5918b019f2166ad487 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/19.jpg differ diff --git a/assets/v2-15s/lightx2v-8/20.jpg b/assets/v2-15s/lightx2v-8/20.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8e65e26bbe342b94e036b884f33376efa3240160 Binary files /dev/null and b/assets/v2-15s/lightx2v-8/20.jpg differ diff --git a/benchmark-evidence.csv b/benchmark-evidence.csv index 46addacc2bddc063bb922b673d1466616c4d6444..188156a4491081f9991e6cb1714c33a12c463acb 100644 --- a/benchmark-evidence.csv +++ b/benchmark-evidence.csv @@ -8,9 +8,15 @@ hsg,fasth3-4,4,none,5,124,NVIDIA GB200,27.3790018921718,3.0292545007541776,2.996 hsg,fasth3-4,8,none,10,243,NVIDIA GB200,10.703646476846188,3.7656888221390545,3.7543425438925624,3.752968884073198,3.7543425438925624,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,fasth3-4,8,none,15,362,NVIDIA GB200,243.63984071277082,6.691152920015156,6.652609454933554,6.675593917723745,6.675593917723745,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,fasth3-4,8,none,5,124,NVIDIA GB200,26.054769210983068,1.7181520899757743,1.6623261668719351,1.6656489758752286,1.6656489758752286,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,1,mxfp8,10,243,NVIDIA GB200,65.33505730889738,65.48566196393222,65.6046371506527,65.33273641113192,65.48566196393222,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,1,mxfp8,15,362,NVIDIA GB200,129.85651914030313,129.51566505106166,129.90104290517047,129.23715650010854,129.51566505106166,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,1,mxfp8,5,124,NVIDIA GB200,42.42403999203816,22.414969386998564,22.482791357208043,22.4277695142664,22.4277695142664,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,1,none,10,243,NVIDIA GB200,71.77961420686916,71.96028976980597,72.14366453094408,71.87395159387961,71.96028976980597,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,1,none,15,362,NVIDIA GB200,137.42840397078544,137.85603942908347,137.42607980174944,136.93979512900114,137.42607980174944,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,1,none,5,124,NVIDIA GB200,40.22856629593298,25.559987880755216,25.68099537305534,25.446089582052082,25.559987880755216,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,4,mxfp8,10,243,NVIDIA GB200,18.84309404483065,11.544127263128757,11.55223888810724,11.559273229911923,11.55223888810724,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,4,mxfp8,15,362,NVIDIA GB200,28.802728682756424,21.485745965968817,21.545224134344608,21.510514890309423,21.510514890309423,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +hsg,hyperflow-8,4,mxfp8,5,124,NVIDIA GB200,33.40463601704687,4.606777192093432,4.593384526204318,4.615335638169199,4.606777192093432,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,4,none,10,243,NVIDIA GB200,113.69129496766254,13.347808117046952,13.34682912286371,13.33125697216019,13.34682912286371,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,4,none,15,362,NVIDIA GB200,235.1847071829252,24.262370044831187,24.287067759316415,24.273042456246912,24.273042456246912,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" hsg,hyperflow-8,4,none,5,124,NVIDIA GB200,30.525569978170097,5.545859571546316,5.51854232000187,5.509740656707436,5.51854232000187,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" @@ -31,9 +37,14 @@ poly,fasth3-4,4,none,5,124,NVIDIA GB200,29.798163390005357,3.0818069269880652,3. poly,fasth3-4,8,none,10,243,NVIDIA GB200,10.177076838008361,3.789100617010263,3.7869828450056957,3.783537929004524,3.7869828450056957,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,fasth3-4,8,none,15,362,NVIDIA GB200,12.688988487992901,6.722846845994354,6.70660141800181,6.693554066994693,6.70660141800181,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,fasth3-4,8,none,5,124,NVIDIA GB200,32.965851381988614,1.632835114010959,1.6313898380030878,1.6286788399884244,1.6313898380030878,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +poly,hyperflow-8,1,mxfp8,10,243,NVIDIA GB200,63.84408454998629,63.75714229498408,63.8483764519915,63.88921740002115,63.8483764519915,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +poly,hyperflow-8,1,mxfp8,5,124,NVIDIA GB200,30.864501964009833,21.856848203984555,21.847355916979723,21.877207930985605,21.856848203984555,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,1,none,10,243,NVIDIA GB200,70.50726453700918,70.32355213700794,70.31223342899466,70.11613318399759,70.31223342899466,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,1,none,15,362,NVIDIA GB200,135.63105801402708,135.65047668898478,135.54529769800138,135.5381122569961,135.54529769800138,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,1,none,5,124,NVIDIA GB200,71.83553511201171,24.93977749699843,24.9366850979859,24.955815777997486,24.93977749699843,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +poly,hyperflow-8,4,mxfp8,10,243,NVIDIA GB200,18.310874960006913,11.83686811098596,11.855843441007892,11.825516913988395,11.83686811098596,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +poly,hyperflow-8,4,mxfp8,15,362,NVIDIA GB200,28.251016705005895,22.0627290699922,22.035492566996254,22.032053962000646,22.035492566996254,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" +poly,hyperflow-8,4,mxfp8,5,124,NVIDIA GB200,21.500160736002726,4.695578521001153,4.63270866998937,4.648682238010224,4.648682238010224,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,4,none,10,243,NVIDIA GB200,47.6237547690107,13.509233238000888,13.580609458003892,13.604155460983748,13.580609458003892,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,4,none,15,362,NVIDIA GB200,59.1786147270177,24.773393165989546,24.86349293999956,24.796716023003682,24.796716023003682,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" poly,hyperflow-8,4,none,5,124,NVIDIA GB200,64.3763722970034,5.503104650997557,5.514375450991793,5.521031700016465,5.514375450991793,"warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding" diff --git a/benchmark_prompt.txt b/benchmark_prompt.txt new file mode 100644 index 0000000000000000000000000000000000000000..51bb7c2be647db2c92043a9663fe600b302b4792 --- /dev/null +++ b/benchmark_prompt.txt @@ -0,0 +1 @@ +A continuous cinematic shot of morning light moving across a quiet room. diff --git a/data.json b/data.json index b547a78e1b62554022219b2cc178421459f6ef5b..c606c8dc753d9a88c3841f8db21bf3f477219819 100644 --- a/data.json +++ b/data.json @@ -69,17 +69,22 @@ }, "source_prompt": "\"A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.\"\n\n- time: 15min ish\n- 768x768", "source_duration": "5s", - "prompt": "A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.", - "prompt_sha256": "6ede0fa934a1dc755b53a87f6747f8e9e39b87c912ac52370171dfac33815847", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cute little girl walks through a sunny forest when a small, friendly dinosaur with a light, cheerful English voice (S1) steps out, waves, and says: [English] hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style. Pace the meeting, sneeze and shared laughter over a continuous 15-second scene, allowing time to see the butterflies settle around both characters.\n\noverall_soundscape: Light forest wind, birds and small footsteps accompany the meeting. A comic sneeze releases a fluttering rush of butterfly wings, followed by the girl and dinosaur laughing together.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "954337465d95f9ab7b1fd512ba3879fa925f8e668f493e314c9d84a91e99fad3", + "duration": 15, "seed": 42, "adaptations": [ - "Removed source author runtime and resolution notes outside the quoted prompt." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 5s，本次明确扩展为 15 秒。" ], "source_title": { "en": "via MLX Serve by", "zh": "via MLX Serve by" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 2, @@ -96,15 +101,22 @@ }, "source_prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", "source_duration": "10s", - "prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", - "prompt_sha256": "713096bb780dfbaeb577baa49e60b44666cd2ce400e9ca9f97d8d00fcae796a0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] amateur handheld pov footage of a tourist with an excited, conversational English voice (S1) in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: [English] Wow look at that!, then they show back the room Use one uninterrupted 15-second handheld take. Switch off the room light around halfway through, hold the window view for the spoken reaction, then turn back to show the room.\n\noverall_soundscape: A steady enclosed-room ventilation hum and soft carpeted footsteps continue under slight handheld handling noise. A light-switch click marks the change to the darkened room.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "712f02d91f39b4d5a029cf4c3396609ac32d151511bf04949501f697139405e5", + "duration": 15, "seed": 43, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 10s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "The last thing you see in your first and last", "zh": "The last thing you see in your first and last" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 3, @@ -121,15 +133,21 @@ }, "source_prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", "source_duration": "15s", - "prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", - "prompt_sha256": "3de5bdf5968d3cb12d87fca7e975541ea97cee72e354a989fbe70ef0f19b7e10", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality. In one continuous 15-second camera move, establish the dawn lake first, raise and spiral the droplets in the middle, hold the completed readable word before its final liquid collapse.\n\noverall_soundscape: Quiet mountain wind, small lake ripples and faint distant dawn birds establish the lakeside. Delicate water movement grows into the rushing liquid-letter collapse and a crisp shower of droplets.\n\nnon_diegetic_music: Slow, sustained string notes and sparse high piano tones rise gently with the water lettering and resolve as the droplets fall.", + "prompt_sha256": "7abb51ba6a7327cb9266b13da9a8fa218928aeb0d97f444c82dc95ea21c245c2", + "duration": 15, "seed": 44, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Ultra cinematic macro shot: A calm mountain lake at dawn", "zh": "Ultra cinematic macro shot: A calm mountain lake at dawn" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 4, @@ -146,15 +164,21 @@ }, "source_prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", "source_duration": "15s", - "prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", - "prompt_sha256": "0f456ed5bd7c35beb4dfc627e3b3082bc499140191221431d6a8fd6aa413e6ed", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Photorealistic premium food commercial, extreme macro high-speed product cinematography. A single ripe strawberry falls toward a white ceramic bowl in a sunlit spring greenhouse. Its red seeds and moist surface are sharply resolved against soft green glasshouse bokeh. Keep the strawberry's screen position, size, orientation and downward motion consistent through every seasonal match cut. No hands, labels or additional fruit.\n[Shot 2] At 00:03.000, the shot transitions to a hot summer picnic in a seamless seasonal match cut around the same falling strawberry and bowl. Warm sunlight replaces the greenhouse light; distant insects accompany its uninterrupted descent.\n[Shot 3] At 00:06.000, the shot transitions to an autumn kitchen with amber window light, preserving the strawberry and bowl alignment and the same downward motion. Soft indoor room tone replaces the outdoor insects.\n[Shot 4] At 00:09.000, the shot transitions to a candlelit winter dining room. The strawberry completes its fall into fresh cream in the white bowl, sending one graceful crown-shaped splash upward. Hold the detailed splash and its settling ripples through the final seconds. One continuous musical phrase connects all four seasons.\n\noverall_soundscape: Spring birds and a gentle greenhouse breeze change to summer insects, quiet autumn kitchen ambience, then the still winter dining room. The final cream splash and falling droplets remain clearly audible.\n\nnon_diegetic_music: One uninterrupted light piano-and-plucked-string phrase runs at a moderate tempo across all four seasons, resolving with the strawberry splash.", + "prompt_sha256": "0c4614d6b1c88a86974b1393eab7a574817abc11c37bf72e3cb53e2d9bbf718f", + "duration": 15, "seed": 45, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Strawberry seasonal match-cut food commercial", "zh": "草莓四季匹配剪辑食品广告" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 5, @@ -171,15 +195,21 @@ }, "source_prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", "source_duration": "15s", - "prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", - "prompt_sha256": "745f72b9452c89f647147db06593c1202b51e0af1a2f2b769f48d1bbdc3e0dfb", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown. Maintain one continuous 15-second rear chase shot with brief arcs that reveal the pursuers when the robot looks back; preserve the robot’s identity and forward motion.\n\noverall_soundscape: Rapid metallic footfalls and small robot servo movements run through the city ambience. Heavy pursuing footsteps, drone rotors and passing machinery track the chase without intelligible dialogue.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be75f61112b0d84692fb5ce91ced32ae50c3f1a3a1ad6b115fba2089944d4073", + "duration": 15, "seed": 46, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Little Robot Cyberpunk Escape", "zh": "小机器人赛博朋克逃亡" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 6, @@ -196,15 +226,21 @@ }, "source_prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", "source_duration": "15s", - "prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", - "prompt_sha256": "99600269200a60ed5566c134a5fdb59a19a7543be9c1e1e8bb9d68661a1d8320", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field. Keep one continuous 15-second medium shot, with a very slow small push-in as she touches the flowers, smiles and finally closes her eyes to feel the breeze.\n\noverall_soundscape: Gentle shore wind moves the ribbons and reeds while small waves wash over the pebbles. Quiet bouquet and fabric rustles accompany her hand movements.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7e05bce3aa444d99e753af9bcc4865bd2934e2c9102f5a1ce63f952302516d15", + "duration": 15, "seed": 47, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3 Max", "zh": "Created with MiniMax H3 Max" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 7, @@ -221,15 +257,22 @@ }, "source_prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", "source_duration": "15s", - "prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", - "prompt_sha256": "022127583b887243e5d890afedd26196c9e3a7cf6665e714e0b8259b867c8c07", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western artist with a relaxed conversational English voice (S1) sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience. Her brief natural English remarks follow her painting actions. Keep the artist, easel and skyline consistent over the 15-second take. Begin with her positioning the easel, follow several visible finishing brush strokes, then give the final canvas-to-skyline comparison time to read.\n\noverall_soundscape: Soft city-park wind, distant traffic and lakeside birds surround the easel. Light wood movements, brush-on-canvas friction and cloth handling follow the artist’s actions.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "f39be188be70b922d6337aa1436e276c98592d5adf01475408b43bbd227e6cbc", + "duration": 15, "seed": 48, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "Created with MiniMax H3 on .", "zh": "Created with MiniMax H3 on ." - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 8, @@ -246,15 +289,21 @@ }, "source_prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", "source_duration": "15s", - "prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", - "prompt_sha256": "1c1f768a47fd7db3bf3402c3e52f828f3de297d6a4ef32f449c7f75cb0cbb651", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western female street photographer with a warm conversational English voice (S1) walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says: [English] Look at that, then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism. Over one continuous 15-second handheld shot, follow her noticing and photographing the café scene, then turn with her as she shows the camera display, speaks and continues walking.\n\noverall_soundscape: Layered downtown footsteps, distant traffic and soft café room spill continue beneath a camera shutter click. The dog shifts and lightly pants beside the seated man.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be91933cc24647e7105ff883517384c8c83c2ccef42801f72555f509c467e319", + "duration": 15, "seed": 49, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3.", "zh": "Created with MiniMax H3." - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 9, @@ -271,15 +320,22 @@ }, "source_prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", "source_duration": "15s", - "prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", - "prompt_sha256": "dd3b9a92fa180a0f865cf0f4e1869330f0aedf3b94bfa2a4df24ff87cd65ba1a", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. The woman, with a light melodic singing voice (S1), sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look. She performs a wordless vocal melody. Keep the low-angle backward tracking shot continuous for the full 15 seconds. Leave enough space for a full turn and return to eye contact without cutting.\n\noverall_soundscape: Light footsteps on stone, wind through the alley and natural fabric movement accompany the performance. Her soft laughter and breaths are synchronized to the visible movement.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7c2058cc5dc84a848b29169d61ecc607158e3ce4e68dd07d43622c10bfd34d91", + "duration": 15, "seed": 50, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词或歌词；未编造逐字台词，歌唱使用无词旋律。" + ], "source_title": { "en": "Low-Angle Fashion Tracking Film", "zh": "低机位时尚跟拍短片" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 10, @@ -296,15 +352,22 @@ }, "source_prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", "source_duration": "14s", - "prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", - "prompt_sha256": "f778e6b31c8426ea25d455ac33f5b0767d095059e1a69fbc2d2b97f0969b7eb0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose. Follow the complete obstacle sequence in one 15-second handheld take, with the bell contact visible in the final seconds.\n\noverall_soundscape: Backyard birds and distant neighborhood ambience surround quick paw taps and excited dog panting. Plastic poles rattle, the fabric tunnel scrapes, and a small bell rings once at the finish.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "b441f7d70cc0c7a54efccb096492e34839955225bc31781700cf560af438bc8d", + "duration": 15, "seed": 51, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Generated a french bulldog with MiniMax H3", "zh": "Generated a french bulldog with MiniMax H3" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 11, @@ -321,15 +384,21 @@ }, "source_prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", "source_duration": "15s", - "prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", - "prompt_sha256": "12a8a0c253466cbb68c3f4896f05c48ac47a2a4612fde6d00519fd3cf7813e0e", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX. Over 15 seconds, establish the glacier, follow the expanding fractures and airborne ice, hold the completed readable ice title, then finish on its violent shatter.\n\noverall_soundscape: Deep frozen-ocean wind surrounds spreading ice creaks and resonant fractures. Heavy ice collapse becomes a crystalline debris storm, followed by the sharp final title shatter.\n\nnon_diegetic_music: Sustained low strings swell with the ice fracture, joined by deep percussion at the shelf collapse and final title shatter.", + "prompt_sha256": "4450f406537c05f4eaf4263bdbb5e1e4d60f9b8e253dbe76aac8d0be5e2b24bd", + "duration": 15, "seed": 52, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with Hailuo H3.", "zh": "Created with Hailuo H3." - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 12, @@ -346,15 +415,21 @@ }, "source_prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", "source_duration": "15s", - "prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", - "prompt_sha256": "1622e4ef07a926746ba75db53791e1ee48aaa6d1da9cbdf512e40f27daf6cd0b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX. Keep the camera move continuous over 15 seconds: establish the floating monolith, enter the spreading fracture, orbit the suspended fragments, hold the completed title briefly, then finish on fragments expanding toward the lens.\n\noverall_soundscape: A projectile impact triggers fine traveling glass cracks and a swelling structural groan. Dense crystalline shattering and passing shards follow the orbit, with a second pressure-wave crack and outward burst at the end.\n\nnon_diegetic_music: A sparse low electronic pulse builds beneath bowed-metal tones, with two deep percussion hits at the main and final glass explosions.", + "prompt_sha256": "b198e1ddb0d1d5c79cab984ce06e7944ecd4efbca41f105960c7312c540dd572", + "duration": 15, "seed": 53, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "SHATTER.", "zh": "SHATTER." - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 13, @@ -371,15 +446,21 @@ }, "source_prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", "source_duration": "15s", - "prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", - "prompt_sha256": "543d21958bff63c8152dd8eafe797abae9972759bb8d0fe35a73fa0d456de3f4", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below. Perform the described camera path as one continuous 15-second move, ending on the high overhead view of the lone figure and exploding surf.\n\noverall_soundscape: Strong coastal wind drives the coat fabric. Deep ocean roar becomes clearer as the camera descends, with violent waves striking rock beneath the final overhead view.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "8a9fb882646cdb521b55ea777eea72c255e9b3863f257602d34d5bda4da3b879", + "duration": 15, "seed": 54, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Cliffside Descent to the Ocean", "zh": "悬崖下行至海边" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 14, @@ -396,17 +477,22 @@ }, "source_prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, vertical 9:16, social media ad quality, highly realistic natural skin texture, not overly retouched.", "source_duration": "15s", - "prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched.", - "prompt_sha256": "992264f4313ef249573cc318eb959dd1e4f84f8c9ef13d0a6dc4fda46e877d40", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] UGC-style skincare video featuring a realistic young woman with a clear conversational English voice (S1) speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched. Her short natural English presentation stays about her skincare routine. Use a continuous 15-second landscape smartphone shot framed to keep both her face and the bottle clearly visible. She presents the serum, applies it to her face, and returns her attention to the camera.\n\noverall_soundscape: Quiet bathroom room tone, small glass-bottle handling sounds and soft hand-on-skin movements accompany the presentation. Keep breaths and clothing movement natural and close.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "61f30fc7218593117f92f6c7e46d55e03de10b31642a4122a69a9eba43d40b4c", + "duration": 15, "seed": 55, "adaptations": [ - "Normalized vertical 9:16 to landscape 16:9 for the shared 1344x768 canvas." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" ], "source_title": { "en": "Bathroom Mirror Skincare UGC Ad", "zh": "浴室镜前护肤 UGC 广告" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 15, @@ -423,15 +509,22 @@ }, "source_prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", "source_duration": "14s", - "prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", - "prompt_sha256": "11220e80abd1c920147a97e2b8ccc2436a5c0b8fecedb0b63751a2297c07f3af", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A low distorted English voice from the phone speaker (S1) says: [English] Finally. I’ve been trying to switch places with you. The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\" Keep the cheap boat-mounted camera viewpoint continuous for the 15-second scene, with only the specified white-flash glitch interrupting the image. Establish the fishing action, allow the phone line to finish, show the trapped image and then the wet double before the abrupt final black frame.\n\noverall_soundscape: Low lake water lapping, boat creaks and fishing-reel strain accompany the fisherman’s heavy breathing. Wet weeds scrape the hull, the phone glitches with clipped electronic noise, then water splashes as the soaked double enters the boat.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "778e3b3e9a6ca16d59312279338db624dd6281b72e5ade48402f63ed4b25f617", + "duration": 15, "seed": 56, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Horror Film Study 335284", "zh": "恐怖短片案例 335284" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 16, @@ -448,15 +541,22 @@ }, "source_prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", "source_duration": "14s", - "prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", - "prompt_sha256": "03fa5bde85bff179c3e1b7bfcb9f378a3e9a1a931a80faff11dbf7ad57e1fc7d", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic science-fiction suspense, realistic spaceship crew in full futuristic uniforms boarding a dark abandoned spacecraft in response to a distress signal. A smooth tracking shot follows them through a narrow metallic corridor; flashlight beams cross flickering warning lights. One crew member with a restrained mid-pitched English voice (S1) says: [English] Signal's coming from deeper inside. Radio static, a low ship hum, creaking metal and hard boot echoes accompany their advance.\n[Shot 2] At 00:04.500, the camera cuts to a wide view of a large chamber with hundreds of open, empty cryogenic pods. A second crew member with an uneasy, slightly higher English voice (S2) looks from the abandoned pods toward a final closed door and says nervously: [English] Where did everyone go? The captain approaches that door as the camera pushes forward with the crew. Pod machinery resonates beneath their footsteps.\n[Shot 3] At 00:09.000, the camera cuts behind the captain as the door hisses open. Bright daylight reveals an ordinary present-day elementary school classroom, children at desks, backpacks on the floor and colorful posters on the walls. A teacher at the front with a clear, matter-of-fact English voice (S3) turns toward the astronauts and scolds them: [English] You're late again. Take your seats. The astronauts remain frozen in confusion at the doorway. Hold the absurd reveal until the end while ordinary classroom room tone replaces the ship ambience.\n\noverall_soundscape: Radio static, a low ship hum, creaking metal, boot echoes and empty pod machinery fill the spacecraft. A door hiss gives way abruptly to ordinary classroom chair movement and quiet paper rustling.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "515b6f9c02a9aff5ebdd48c593434b27848f0a398afc1c0651103dfdd475c0da", + "duration": 15, "seed": 57, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic", "zh": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 17, @@ -473,15 +573,21 @@ }, "source_prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", "source_duration": "15s", - "prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", - "prompt_sha256": "bc2776b295ff7d9b6bb9f534b11b75dd1a94c4d7a5666701b6abc440e047f69b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds. Complete the described path in one uninterrupted 15-second flight, establishing the skyship first and reserving the final seconds for the sunlight reveal beyond the storm.\n\noverall_soundscape: High-altitude wind and engine thrust deepen as the skyship enters thunder, lashing rain and cracking rock. Waterfalls rush past the hull before the storm sound falls behind the ship in open sunlight.\n\nnon_diegetic_music: Broad sustained strings and steady low orchestral percussion build through the storm, then open into a long brass-and-string chord in sunlight.", + "prompt_sha256": "15ed03a4ff9ad4d170616c8220d15f8c7ddadb16a059a522c567a1cede541838", + "duration": 15, "seed": 58, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Skyship Through a Floating Kingdom", "zh": "飞空艇穿越浮空王国" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 18, @@ -498,15 +604,22 @@ }, "source_prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", "source_duration": "14s", - "prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", - "prompt_sha256": "7fd67989d4363071c6d2e8384ecf994ee181b236e5a2f03f81d32aaefc7d931c", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\" Use a single 15-second home-video take. Preserve the first several seconds of ordinary watering, begin the overhead crack after that peaceful lead-in, and end with the sky descending and camera falling.\n\noverall_soundscape: Ordinary backyard birds, distant traffic and a watering stream continue for several seconds. A sudden deep overhead crack becomes a heavy descending rumble, followed by startled breathing and the phone hitting the ground.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "42f8695e30223aeca06566bf0476f7882751c30d3db90ad0136585d517c6ba26", + "duration": 15, "seed": 59, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright", "zh": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 19, @@ -523,15 +636,22 @@ }, "source_prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", "source_duration": "15s", - "prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", - "prompt_sha256": "6545b609b933b4f6574b2aa774e0198a9e4380c5cff593db1e2cd3cfd429491f", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism. Nearby women bargaining in Javanese (S1) and vendors calling prices in Javanese (S2) remain overlapping background voices. Follow the vendor in one continuous 15-second handheld take, keeping the steaming bowl and his careful steps legible; finish with serving the bowl and the brief power flicker.\n\noverall_soundscape: Dense covered-market bustle includes footsteps, bowl clatter, steaming food and sacks brushing clothing. Slow motorcycle engines pass among the stalls, followed by a brief fluorescent power buzz and the uninterrupted market activity.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "644ee708e360b5247f1f86b9510bc54be46b1ede81be1dfd8e283227dfab466f", + "duration": 15, "seed": 60, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "also has an impressive understanding of different cultures", "zh": "also has an impressive understanding of different cultures" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" }, { "id": 20, @@ -548,15 +668,21 @@ }, "source_prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", "source_duration": "15s", - "prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", - "prompt_sha256": "1bece371d5c458f6ba5e2f3cfffa5837017493101b52b6a3ecb5ee037df9b894", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic wuxia mystery in a bamboo forest at night, in a restrained cold-blue, ink-green, charcoal and gray palette. Thin mist and fine drifting snow surround two period-clothed martial-arts investigators. Dense vertical bamboo fills the background; soft foreground leaves partially obscure a tight facial close-up of the first investigator. Cool, soft front-side light reveals natural skin texture while cold white mist-light glows in the distance. The investigator looks toward the other person with measured, wary attention. No modern objects, text, subtitles, watermarks or comic performance.\n[Shot 2] At 00:05.000, the camera cuts to the other investigator in matching close-up for a restrained reverse angle. Keep the same cold lighting, shallow depth of field and softly blurred falling snow. A small shift of the eyes toward the dark grove conveys the exchange of secret intelligence without adding invented dialogue. Clothing rustles quietly under the bamboo wind.\n[Shot 3] At 00:10.000, the camera cuts back to the first investigator seen through out-of-focus leaves. A slow, small push-in follows a controlled glance into the luminous mist. Hold the austere, tense expression until the end, with the distant forest remaining still except for snow and moving leaves. Photoreal period-drama lighting, no animation styling, bright daylight or over-smoothed skin.\n\noverall_soundscape: Thin wind passes through bamboo leaves with occasional dry stalk creaks. Soft period-clothing movement and faint footsteps remain close while falling snow and distant forest space stay quiet.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "11628027721bd267f57680bb10ff37aff3623b6ea6bbfb064178144173e3a828", + "duration": 15, "seed": 61, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Bamboo Forest Wuxia Mystery", "zh": "竹林武侠谜案" - } + }, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt" } ], "outputs": [ @@ -573,6 +699,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -582,7 +709,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -600,36 +727,45 @@ "name": "@albertgao", "url": "https://x.com/albertgao/status/2085137397134561487" }, + "source_prompt": "\"A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.\"\n\n- time: 15min ish\n- 768x768", "source_duration": "5s", - "prompt": "A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.", - "prompt_sha256": "6ede0fa934a1dc755b53a87f6747f8e9e39b87c912ac52370171dfac33815847", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cute little girl walks through a sunny forest when a small, friendly dinosaur with a light, cheerful English voice (S1) steps out, waves, and says: [English] hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style. Pace the meeting, sneeze and shared laughter over a continuous 15-second scene, allowing time to see the butterflies settle around both characters.\n\noverall_soundscape: Light forest wind, birds and small footsteps accompany the meeting. A comic sneeze releases a fluttering rush of butterfly wings, followed by the girl and dinosaur laughing together.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "954337465d95f9ab7b1fd512ba3879fa925f8e668f493e314c9d84a91e99fad3", + "duration": 15, "seed": 42, "adaptations": [ - "Removed source author runtime and resolution notes outside the quoted prompt." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 5s，本次明确扩展为 15 秒。" ], "source_title": { "en": "via MLX Serve by", "zh": "via MLX Serve by" }, - "pipeline_s": 3.3317544390447438, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.019709377083927, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/01.mp4", - "poster": "assets/fasth3-4/01.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/01.mp4", + "poster": "assets/v2-15s/fasth3-4/01.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2569868, - "sha256": "67a4b4187a3e296919fa0e199f7ed9f0551c73c46d51c0762a71fbac1a3dfbd4", + "duration_s": 15.166666, + "bytes": 6482287, + "sha256": "6541ab0ed98b8e1c74c21ca8be0e11d000c9577f519d91484def8aa4a66edeb2", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/01.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/01.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -644,6 +780,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -653,7 +790,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -671,34 +808,45 @@ "name": "@ivanfioravanti", "url": "https://x.com/ivanfioravanti/status/2086553101029290296" }, + "source_prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", "source_duration": "10s", - "prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", - "prompt_sha256": "713096bb780dfbaeb577baa49e60b44666cd2ce400e9ca9f97d8d00fcae796a0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] amateur handheld pov footage of a tourist with an excited, conversational English voice (S1) in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: [English] Wow look at that!, then they show back the room Use one uninterrupted 15-second handheld take. Switch off the room light around halfway through, hold the window view for the spoken reaction, then turn back to show the room.\n\noverall_soundscape: A steady enclosed-room ventilation hum and soft carpeted footsteps continue under slight handheld handling noise. A light-switch click marks the change to the darkened room.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "712f02d91f39b4d5a029cf4c3396609ac32d151511bf04949501f697139405e5", + "duration": 15, "seed": 43, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 10s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "The last thing you see in your first and last", "zh": "The last thing you see in your first and last" }, - "pipeline_s": 3.418869440909475, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.38888839399442, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/02.mp4", - "poster": "assets/fasth3-4/02.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/02.mp4", + "poster": "assets/v2-15s/fasth3-4/02.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2728494, - "sha256": "6d91d888d22af863873f9d8bb35d6e893ec383720504bf303aaccbfe90ae0bda", + "duration_s": 15.166666, + "bytes": 6844578, + "sha256": "1206f7816cd4c0b865d28738fead5c660828eb90406bdea3f1832c4ab82558a4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/02.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/02.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -713,6 +861,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -722,7 +871,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -740,34 +889,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2085622021812039920" }, + "source_prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", "source_duration": "15s", - "prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", - "prompt_sha256": "3de5bdf5968d3cb12d87fca7e975541ea97cee72e354a989fbe70ef0f19b7e10", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality. In one continuous 15-second camera move, establish the dawn lake first, raise and spiral the droplets in the middle, hold the completed readable word before its final liquid collapse.\n\noverall_soundscape: Quiet mountain wind, small lake ripples and faint distant dawn birds establish the lakeside. Delicate water movement grows into the rushing liquid-letter collapse and a crisp shower of droplets.\n\nnon_diegetic_music: Slow, sustained string notes and sparse high piano tones rise gently with the water lettering and resolve as the droplets fall.", + "prompt_sha256": "7abb51ba6a7327cb9266b13da9a8fa218928aeb0d97f444c82dc95ea21c245c2", + "duration": 15, "seed": 44, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Ultra cinematic macro shot: A calm mountain lake at dawn", "zh": "Ultra cinematic macro shot: A calm mountain lake at dawn" }, - "pipeline_s": 3.1660079597495496, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 17.910059693735093, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/03.mp4", - "poster": "assets/fasth3-4/03.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/03.mp4", + "poster": "assets/v2-15s/fasth3-4/03.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3028625, - "sha256": "43320ef9c3d5acb9a3ed99a7dcb4894fbac45498e730a3061ca598a79d7559b5", + "duration_s": 15.166666, + "bytes": 10485091, + "sha256": "2035c6a3eb0046e08b41001ff250eac42d80bd117e9e8c7fc240cd25ce08a133", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/03.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/03.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -782,6 +941,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -791,7 +951,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -809,34 +969,44 @@ "name": "@HBCoop_", "url": "https://x.com/HBCoop_/status/2082827829172117943" }, + "source_prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", "source_duration": "15s", - "prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", - "prompt_sha256": "0f456ed5bd7c35beb4dfc627e3b3082bc499140191221431d6a8fd6aa413e6ed", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Photorealistic premium food commercial, extreme macro high-speed product cinematography. A single ripe strawberry falls toward a white ceramic bowl in a sunlit spring greenhouse. Its red seeds and moist surface are sharply resolved against soft green glasshouse bokeh. Keep the strawberry's screen position, size, orientation and downward motion consistent through every seasonal match cut. No hands, labels or additional fruit.\n[Shot 2] At 00:03.000, the shot transitions to a hot summer picnic in a seamless seasonal match cut around the same falling strawberry and bowl. Warm sunlight replaces the greenhouse light; distant insects accompany its uninterrupted descent.\n[Shot 3] At 00:06.000, the shot transitions to an autumn kitchen with amber window light, preserving the strawberry and bowl alignment and the same downward motion. Soft indoor room tone replaces the outdoor insects.\n[Shot 4] At 00:09.000, the shot transitions to a candlelit winter dining room. The strawberry completes its fall into fresh cream in the white bowl, sending one graceful crown-shaped splash upward. Hold the detailed splash and its settling ripples through the final seconds. One continuous musical phrase connects all four seasons.\n\noverall_soundscape: Spring birds and a gentle greenhouse breeze change to summer insects, quiet autumn kitchen ambience, then the still winter dining room. The final cream splash and falling droplets remain clearly audible.\n\nnon_diegetic_music: One uninterrupted light piano-and-plucked-string phrase runs at a moderate tempo across all four seasons, resolving with the strawberry splash.", + "prompt_sha256": "0c4614d6b1c88a86974b1393eab7a574817abc11c37bf72e3cb53e2d9bbf718f", + "duration": 15, "seed": 45, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Strawberry seasonal match-cut food commercial", "zh": "草莓四季匹配剪辑食品广告" }, - "pipeline_s": 3.156159962993115, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 16.10258880071342, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/04.mp4", - "poster": "assets/fasth3-4/04.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/04.mp4", + "poster": "assets/v2-15s/fasth3-4/04.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3332012, - "sha256": "2e3fde490f935eba3dded5c9f13e702d3519b78801b4c168699d757c9322f915", + "duration_s": 15.166666, + "bytes": 4354758, + "sha256": "e46ad094d1dc471b260948562878526824f1475bb34ac6d4476b141a217f7d9a", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/04.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/04.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -851,6 +1021,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -860,7 +1031,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -878,34 +1049,44 @@ "name": "@AIandDesign", "url": "https://x.com/AIandDesign/status/2082522979339288619" }, + "source_prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", "source_duration": "15s", - "prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", - "prompt_sha256": "745f72b9452c89f647147db06593c1202b51e0af1a2f2b769f48d1bbdc3e0dfb", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown. Maintain one continuous 15-second rear chase shot with brief arcs that reveal the pursuers when the robot looks back; preserve the robot’s identity and forward motion.\n\noverall_soundscape: Rapid metallic footfalls and small robot servo movements run through the city ambience. Heavy pursuing footsteps, drone rotors and passing machinery track the chase without intelligible dialogue.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be75f61112b0d84692fb5ce91ced32ae50c3f1a3a1ad6b115fba2089944d4073", + "duration": 15, "seed": 46, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Little Robot Cyberpunk Escape", "zh": "小机器人赛博朋克逃亡" }, - "pipeline_s": 3.1928351931273937, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.840128422249109, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/05.mp4", - "poster": "assets/fasth3-4/05.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/05.mp4", + "poster": "assets/v2-15s/fasth3-4/05.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 4265976, - "sha256": "ae15d3e73b0095b932ccd5d5e6c5dce54d1fea4793b86500e2454ea749cddeaf", + "duration_s": 15.166666, + "bytes": 11493828, + "sha256": "53331ff4e26ca80e2abc8d3a063d5be56160a38a1866906f0385c27bb3a3df95", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/05.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/05.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -920,6 +1101,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -929,7 +1111,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -947,34 +1129,44 @@ "name": "@TaliaAariz", "url": "https://x.com/TaliaAariz/status/2095857062987120666" }, + "source_prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", "source_duration": "15s", - "prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", - "prompt_sha256": "99600269200a60ed5566c134a5fdb59a19a7543be9c1e1e8bb9d68661a1d8320", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field. Keep one continuous 15-second medium shot, with a very slow small push-in as she touches the flowers, smiles and finally closes her eyes to feel the breeze.\n\noverall_soundscape: Gentle shore wind moves the ribbons and reeds while small waves wash over the pebbles. Quiet bouquet and fabric rustles accompany her hand movements.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7e05bce3aa444d99e753af9bcc4865bd2934e2c9102f5a1ce63f952302516d15", + "duration": 15, "seed": 47, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3 Max", "zh": "Created with MiniMax H3 Max" }, - "pipeline_s": 3.2194683118723333, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 20.086923273745924, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/06.mp4", - "poster": "assets/fasth3-4/06.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/06.mp4", + "poster": "assets/v2-15s/fasth3-4/06.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1495457, - "sha256": "b95bb0edbe8792a1b84b88881478d91ab5b3c722a61e66ea3ca3b02e98d584f4", + "duration_s": 15.166666, + "bytes": 6770280, + "sha256": "da8a01f0702c2765cb7a26a2f08c10e188495570c29a6bcbdecfc9b241322b79", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/06.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/06.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -989,6 +1181,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -998,7 +1191,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1016,34 +1209,45 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087408671101989121" }, + "source_prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", "source_duration": "15s", - "prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", - "prompt_sha256": "022127583b887243e5d890afedd26196c9e3a7cf6665e714e0b8259b867c8c07", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western artist with a relaxed conversational English voice (S1) sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience. Her brief natural English remarks follow her painting actions. Keep the artist, easel and skyline consistent over the 15-second take. Begin with her positioning the easel, follow several visible finishing brush strokes, then give the final canvas-to-skyline comparison time to read.\n\noverall_soundscape: Soft city-park wind, distant traffic and lakeside birds surround the easel. Light wood movements, brush-on-canvas friction and cloth handling follow the artist’s actions.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "f39be188be70b922d6337aa1436e276c98592d5adf01475408b43bbd227e6cbc", + "duration": 15, "seed": 48, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "Created with MiniMax H3 on .", "zh": "Created with MiniMax H3 on ." }, - "pipeline_s": 3.151755423285067, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 16.880621903110296, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/07.mp4", - "poster": "assets/fasth3-4/07.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/07.mp4", + "poster": "assets/v2-15s/fasth3-4/07.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2489688, - "sha256": "d213041cc7b8a84da97d2a543c3db94f5d6259bc35f4525254c12e869ab6bd51", + "duration_s": 15.166666, + "bytes": 6960548, + "sha256": "cbe28a1d7749470547f2feef0e2f475d1c0b7c77f49c50ea00c4d906593b7a0f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/07.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/07.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1058,6 +1262,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1067,7 +1272,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1085,34 +1290,44 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087102541146522089" }, + "source_prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", "source_duration": "15s", - "prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", - "prompt_sha256": "1c1f768a47fd7db3bf3402c3e52f828f3de297d6a4ef32f449c7f75cb0cbb651", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western female street photographer with a warm conversational English voice (S1) walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says: [English] Look at that, then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism. Over one continuous 15-second handheld shot, follow her noticing and photographing the café scene, then turn with her as she shows the camera display, speaks and continues walking.\n\noverall_soundscape: Layered downtown footsteps, distant traffic and soft café room spill continue beneath a camera shutter click. The dog shifts and lightly pants beside the seated man.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be91933cc24647e7105ff883517384c8c83c2ccef42801f72555f509c467e319", + "duration": 15, "seed": 49, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3.", "zh": "Created with MiniMax H3." }, - "pipeline_s": 3.1507534091360867, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.759661377873272, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/08.mp4", - "poster": "assets/fasth3-4/08.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/08.mp4", + "poster": "assets/v2-15s/fasth3-4/08.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3884958, - "sha256": "979b8b4aee7facab4d13d116b0ad67e3423217bf933c1a296a51056c8fd317e4", + "duration_s": 15.166666, + "bytes": 14637787, + "sha256": "ecdf783c52abc0773192c9be51a0ae25e5579c7dd4de1db17ca0f33724785f9f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/08.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/08.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1127,6 +1342,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1136,7 +1352,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1154,34 +1370,45 @@ "name": "@Kiber_Alla", "url": "https://x.com/Kiber_Alla/status/2083583963512062019" }, + "source_prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", "source_duration": "15s", - "prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", - "prompt_sha256": "dd3b9a92fa180a0f865cf0f4e1869330f0aedf3b94bfa2a4df24ff87cd65ba1a", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. The woman, with a light melodic singing voice (S1), sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look. She performs a wordless vocal melody. Keep the low-angle backward tracking shot continuous for the full 15 seconds. Leave enough space for a full turn and return to eye contact without cutting.\n\noverall_soundscape: Light footsteps on stone, wind through the alley and natural fabric movement accompany the performance. Her soft laughter and breaths are synchronized to the visible movement.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7c2058cc5dc84a848b29169d61ecc607158e3ce4e68dd07d43622c10bfd34d91", + "duration": 15, "seed": 50, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词或歌词；未编造逐字台词，歌唱使用无词旋律。" + ], "source_title": { "en": "Low-Angle Fashion Tracking Film", "zh": "低机位时尚跟拍短片" }, - "pipeline_s": 3.189090142957866, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.1059017656371, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/09.mp4", - "poster": "assets/fasth3-4/09.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/09.mp4", + "poster": "assets/v2-15s/fasth3-4/09.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2526998, - "sha256": "7c852408c5a0dd2fcde2d5d2ea678935238332a6398096f5fd8adeb7c46d0b30", + "duration_s": 15.166666, + "bytes": 11326646, + "sha256": "8de629a836f83a636b3d5f822fa8e83260ecfd1204d6b274870e6ced79dc0282", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/09.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/09.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1196,6 +1423,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1205,7 +1433,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1223,34 +1451,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085448660612182507" }, + "source_prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", "source_duration": "14s", - "prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", - "prompt_sha256": "f778e6b31c8426ea25d455ac33f5b0767d095059e1a69fbc2d2b97f0969b7eb0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose. Follow the complete obstacle sequence in one 15-second handheld take, with the bell contact visible in the final seconds.\n\noverall_soundscape: Backyard birds and distant neighborhood ambience surround quick paw taps and excited dog panting. Plastic poles rattle, the fabric tunnel scrapes, and a small bell rings once at the finish.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "b441f7d70cc0c7a54efccb096492e34839955225bc31781700cf560af438bc8d", + "duration": 15, "seed": 51, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Generated a french bulldog with MiniMax H3", "zh": "Generated a french bulldog with MiniMax H3" }, - "pipeline_s": 3.141581412870437, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.046641102060676, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/10.mp4", - "poster": "assets/fasth3-4/10.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/10.mp4", + "poster": "assets/v2-15s/fasth3-4/10.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 12450056, - "sha256": "e5782a1cb953e86f5b415fb1a29ffb0d0b078e9238e61636add1762dc439bee7", + "duration_s": 15.166666, + "bytes": 13964920, + "sha256": "ff61266928916f5dd70406cefdb1a54d8a2ab6bc1fd6a6d8f7bc700443a71f5c", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/10.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/10.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1265,6 +1504,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1274,7 +1514,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1292,34 +1532,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2090740724803182619" }, + "source_prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", "source_duration": "15s", - "prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", - "prompt_sha256": "12a8a0c253466cbb68c3f4896f05c48ac47a2a4612fde6d00519fd3cf7813e0e", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX. Over 15 seconds, establish the glacier, follow the expanding fractures and airborne ice, hold the completed readable ice title, then finish on its violent shatter.\n\noverall_soundscape: Deep frozen-ocean wind surrounds spreading ice creaks and resonant fractures. Heavy ice collapse becomes a crystalline debris storm, followed by the sharp final title shatter.\n\nnon_diegetic_music: Sustained low strings swell with the ice fracture, joined by deep percussion at the shelf collapse and final title shatter.", + "prompt_sha256": "4450f406537c05f4eaf4263bdbb5e1e4d60f9b8e253dbe76aac8d0be5e2b24bd", + "duration": 15, "seed": 52, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with Hailuo H3.", "zh": "Created with Hailuo H3." }, - "pipeline_s": 3.184636514633894, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.46429255977273, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/11.mp4", - "poster": "assets/fasth3-4/11.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/11.mp4", + "poster": "assets/v2-15s/fasth3-4/11.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3752867, - "sha256": "02fef75cae5e15b31f2d95fc13d2ec17024922eefcb60a0a93820a50fa973394", + "duration_s": 15.166666, + "bytes": 8081341, + "sha256": "924c1eb52a9acb27a09a2e9615d51fa06e3a609ae0f18650e8742f151878e3ec", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/11.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/11.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1334,6 +1584,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1343,7 +1594,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1361,34 +1612,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2096554023809695817" }, + "source_prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", "source_duration": "15s", - "prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", - "prompt_sha256": "1622e4ef07a926746ba75db53791e1ee48aaa6d1da9cbdf512e40f27daf6cd0b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX. Keep the camera move continuous over 15 seconds: establish the floating monolith, enter the spreading fracture, orbit the suspended fragments, hold the completed title briefly, then finish on fragments expanding toward the lens.\n\noverall_soundscape: A projectile impact triggers fine traveling glass cracks and a swelling structural groan. Dense crystalline shattering and passing shards follow the orbit, with a second pressure-wave crack and outward burst at the end.\n\nnon_diegetic_music: A sparse low electronic pulse builds beneath bowed-metal tones, with two deep percussion hits at the main and final glass explosions.", + "prompt_sha256": "b198e1ddb0d1d5c79cab984ce06e7944ecd4efbca41f105960c7312c540dd572", + "duration": 15, "seed": 53, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "SHATTER.", "zh": "SHATTER." }, - "pipeline_s": 3.15117008658126, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.027102902997285, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/12.mp4", - "poster": "assets/fasth3-4/12.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/12.mp4", + "poster": "assets/v2-15s/fasth3-4/12.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 4778783, - "sha256": "a94d1d9b7a156d430ddab2155e2b49930d7013386f66fb1ba6d74149dfc1bc7b", + "duration_s": 15.166666, + "bytes": 8980980, + "sha256": "a1ea7fe2db44a233bf603bdf078ffce957e1fb18c7b03a3866d3e2a2b041dbb6", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/12.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/12.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1403,6 +1664,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1412,7 +1674,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1430,34 +1692,44 @@ "name": "@umesh_ai", "url": "https://x.com/umesh_ai/status/2082700637444452380" }, + "source_prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", "source_duration": "15s", - "prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", - "prompt_sha256": "543d21958bff63c8152dd8eafe797abae9972759bb8d0fe35a73fa0d456de3f4", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below. Perform the described camera path as one continuous 15-second move, ending on the high overhead view of the lone figure and exploding surf.\n\noverall_soundscape: Strong coastal wind drives the coat fabric. Deep ocean roar becomes clearer as the camera descends, with violent waves striking rock beneath the final overhead view.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "8a9fb882646cdb521b55ea777eea72c255e9b3863f257602d34d5bda4da3b879", + "duration": 15, "seed": 54, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Cliffside Descent to the Ocean", "zh": "悬崖下行至海边" }, - "pipeline_s": 3.137073096819222, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 12.714291729032993, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/13.mp4", - "poster": "assets/fasth3-4/13.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/13.mp4", + "poster": "assets/v2-15s/fasth3-4/13.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 6218104, - "sha256": "ce16ac32ead5f88e68da45cff83e42612bd066b458a2cdac2ba10d09ea579e09", + "duration_s": 15.166666, + "bytes": 11083182, + "sha256": "1269e1e7391bd501e314964b88b1dae6be6681b7b2d3157db789d92666297742", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/13.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/13.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1472,6 +1744,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1481,7 +1754,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1499,36 +1772,45 @@ "name": "@ZaraIrahh", "url": "https://x.com/ZaraIrahh/status/2083011066800242986" }, + "source_prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, vertical 9:16, social media ad quality, highly realistic natural skin texture, not overly retouched.", "source_duration": "15s", - "prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched.", - "prompt_sha256": "992264f4313ef249573cc318eb959dd1e4f84f8c9ef13d0a6dc4fda46e877d40", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] UGC-style skincare video featuring a realistic young woman with a clear conversational English voice (S1) speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched. Her short natural English presentation stays about her skincare routine. Use a continuous 15-second landscape smartphone shot framed to keep both her face and the bottle clearly visible. She presents the serum, applies it to her face, and returns her attention to the camera.\n\noverall_soundscape: Quiet bathroom room tone, small glass-bottle handling sounds and soft hand-on-skin movements accompany the presentation. Keep breaths and clothing movement natural and close.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "61f30fc7218593117f92f6c7e46d55e03de10b31642a4122a69a9eba43d40b4c", + "duration": 15, "seed": 55, "adaptations": [ - "Normalized vertical 9:16 to landscape 16:9 for the shared 1344x768 canvas." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" ], "source_title": { "en": "Bathroom Mirror Skincare UGC Ad", "zh": "浴室镜前护肤 UGC 广告" }, - "pipeline_s": 2.9761163759976625, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 12.73095204308629, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/14.mp4", - "poster": "assets/fasth3-4/14.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/14.mp4", + "poster": "assets/v2-15s/fasth3-4/14.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2466334, - "sha256": "bbf683c93395552f4e49626077ef5d7dd750a07240282eaa53e1fef0bad00941", + "duration_s": 15.166666, + "bytes": 4560139, + "sha256": "edbbf90d2ddec9845281f4991701ba21544dee90c35df2a82fb359ba45f8d6a7", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/14.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/14.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1543,6 +1825,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1552,7 +1835,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1570,34 +1853,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2084842607411335284" }, + "source_prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", "source_duration": "14s", - "prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", - "prompt_sha256": "11220e80abd1c920147a97e2b8ccc2436a5c0b8fecedb0b63751a2297c07f3af", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A low distorted English voice from the phone speaker (S1) says: [English] Finally. I’ve been trying to switch places with you. The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\" Keep the cheap boat-mounted camera viewpoint continuous for the 15-second scene, with only the specified white-flash glitch interrupting the image. Establish the fishing action, allow the phone line to finish, show the trapped image and then the wet double before the abrupt final black frame.\n\noverall_soundscape: Low lake water lapping, boat creaks and fishing-reel strain accompany the fisherman’s heavy breathing. Wet weeds scrape the hull, the phone glitches with clipped electronic noise, then water splashes as the soaked double enters the boat.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "778e3b3e9a6ca16d59312279338db624dd6281b72e5ade48402f63ed4b25f617", + "duration": 15, "seed": 56, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Horror Film Study 335284", "zh": "恐怖短片案例 335284" }, - "pipeline_s": 3.183855720795691, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.290331610944122, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/15.mp4", - "poster": "assets/fasth3-4/15.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/15.mp4", + "poster": "assets/v2-15s/fasth3-4/15.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3082601, - "sha256": "6a2cc520abe1833cf268f2f472485aff8da268646f2a5d095d800058ff90968d", + "duration_s": 15.166666, + "bytes": 9875469, + "sha256": "ed08b1518c4242806f80efac2b115a416ce7e4785ee1bc802b0c722d56b6849a", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/15.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/15.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1612,6 +1906,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1621,7 +1916,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1639,34 +1934,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085498501384024495" }, + "source_prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", "source_duration": "14s", - "prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", - "prompt_sha256": "03fa5bde85bff179c3e1b7bfcb9f378a3e9a1a931a80faff11dbf7ad57e1fc7d", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic science-fiction suspense, realistic spaceship crew in full futuristic uniforms boarding a dark abandoned spacecraft in response to a distress signal. A smooth tracking shot follows them through a narrow metallic corridor; flashlight beams cross flickering warning lights. One crew member with a restrained mid-pitched English voice (S1) says: [English] Signal's coming from deeper inside. Radio static, a low ship hum, creaking metal and hard boot echoes accompany their advance.\n[Shot 2] At 00:04.500, the camera cuts to a wide view of a large chamber with hundreds of open, empty cryogenic pods. A second crew member with an uneasy, slightly higher English voice (S2) looks from the abandoned pods toward a final closed door and says nervously: [English] Where did everyone go? The captain approaches that door as the camera pushes forward with the crew. Pod machinery resonates beneath their footsteps.\n[Shot 3] At 00:09.000, the camera cuts behind the captain as the door hisses open. Bright daylight reveals an ordinary present-day elementary school classroom, children at desks, backpacks on the floor and colorful posters on the walls. A teacher at the front with a clear, matter-of-fact English voice (S3) turns toward the astronauts and scolds them: [English] You're late again. Take your seats. The astronauts remain frozen in confusion at the doorway. Hold the absurd reveal until the end while ordinary classroom room tone replaces the ship ambience.\n\noverall_soundscape: Radio static, a low ship hum, creaking metal, boot echoes and empty pod machinery fill the spacecraft. A door hiss gives way abruptly to ordinary classroom chair movement and quiet paper rustling.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "515b6f9c02a9aff5ebdd48c593434b27848f0a398afc1c0651103dfdd475c0da", + "duration": 15, "seed": 57, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic", "zh": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic" }, - "pipeline_s": 3.1653734520077705, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.769144061021507, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/16.mp4", - "poster": "assets/fasth3-4/16.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/16.mp4", + "poster": "assets/v2-15s/fasth3-4/16.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3156874, - "sha256": "83a1051305114e1c17c869d9a673ee60b3f0307f0e330d60318f204cdbb89238", + "duration_s": 15.166666, + "bytes": 8488005, + "sha256": "453f8c0b801b032b06c1ff94d0b9e612cf141a7e2954f4edc59cd2f233e46854", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/16.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/16.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1681,6 +1987,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1690,7 +1997,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1708,34 +2015,44 @@ "name": "@Strength04_X", "url": "https://x.com/Strength04_X/status/2082712692159344891" }, + "source_prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", "source_duration": "15s", - "prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", - "prompt_sha256": "bc2776b295ff7d9b6bb9f534b11b75dd1a94c4d7a5666701b6abc440e047f69b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds. Complete the described path in one uninterrupted 15-second flight, establishing the skyship first and reserving the final seconds for the sunlight reveal beyond the storm.\n\noverall_soundscape: High-altitude wind and engine thrust deepen as the skyship enters thunder, lashing rain and cracking rock. Waterfalls rush past the hull before the storm sound falls behind the ship in open sunlight.\n\nnon_diegetic_music: Broad sustained strings and steady low orchestral percussion build through the storm, then open into a long brass-and-string chord in sunlight.", + "prompt_sha256": "15ed03a4ff9ad4d170616c8220d15f8c7ddadb16a059a522c567a1cede541838", + "duration": 15, "seed": 58, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Skyship Through a Floating Kingdom", "zh": "飞空艇穿越浮空王国" }, - "pipeline_s": 3.138898960314691, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.923790429718792, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/17.mp4", - "poster": "assets/fasth3-4/17.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/17.mp4", + "poster": "assets/v2-15s/fasth3-4/17.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 5520273, - "sha256": "8bc493a60f546bfd9588f8b513f38ae9f9bcd2e25e25eb74b3ab735c9a32ef14", + "duration_s": 15.166666, + "bytes": 13345460, + "sha256": "65ddfb7838bfc3f295707356b32649b7ba46196a62336e64cd15a72c7c717842", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/17.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/17.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1750,6 +2067,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1759,7 +2077,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1777,34 +2095,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2086879654116495564" }, + "source_prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", "source_duration": "14s", - "prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", - "prompt_sha256": "7fd67989d4363071c6d2e8384ecf994ee181b236e5a2f03f81d32aaefc7d931c", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\" Use a single 15-second home-video take. Preserve the first several seconds of ordinary watering, begin the overhead crack after that peaceful lead-in, and end with the sky descending and camera falling.\n\noverall_soundscape: Ordinary backyard birds, distant traffic and a watering stream continue for several seconds. A sudden deep overhead crack becomes a heavy descending rumble, followed by startled breathing and the phone hitting the ground.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "42f8695e30223aeca06566bf0476f7882751c30d3db90ad0136585d517c6ba26", + "duration": 15, "seed": 59, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright", "zh": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright" }, - "pipeline_s": 3.1646113111637533, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.091396818868816, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/18.mp4", - "poster": "assets/fasth3-4/18.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/18.mp4", + "poster": "assets/v2-15s/fasth3-4/18.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 4239183, - "sha256": "509270366e920eb8b7344879227fd1c7bf9009157bc64e67a893188367a31c38", + "duration_s": 15.166666, + "bytes": 10860207, + "sha256": "a809574802cfe4457e63a4f60ae5f5152adc11354db92c0512c324e5c77d08a7", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/18.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/18.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1819,6 +2148,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1828,7 +2158,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1846,34 +2176,45 @@ "name": "@junwatu", "url": "https://x.com/junwatu/status/2084840152715911311" }, + "source_prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", "source_duration": "15s", - "prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", - "prompt_sha256": "6545b609b933b4f6574b2aa774e0198a9e4380c5cff593db1e2cd3cfd429491f", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism. Nearby women bargaining in Javanese (S1) and vendors calling prices in Javanese (S2) remain overlapping background voices. Follow the vendor in one continuous 15-second handheld take, keeping the steaming bowl and his careful steps legible; finish with serving the bowl and the brief power flicker.\n\noverall_soundscape: Dense covered-market bustle includes footsteps, bowl clatter, steaming food and sacks brushing clothing. Slow motorcycle engines pass among the stalls, followed by a brief fluorescent power buzz and the uninterrupted market activity.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "644ee708e360b5247f1f86b9510bc54be46b1ede81be1dfd8e283227dfab466f", + "duration": 15, "seed": 60, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "also has an impressive understanding of different cultures", "zh": "also has an impressive understanding of different cultures" }, - "pipeline_s": 3.1769428960978985, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 12.90002251509577, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/19.mp4", - "poster": "assets/fasth3-4/19.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/19.mp4", + "poster": "assets/v2-15s/fasth3-4/19.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 4584395, - "sha256": "d6ee9cd0afff25b11b37630c151bad42b6127377f635386941319994cec5dc1d", + "duration_s": 15.166666, + "bytes": 10891060, + "sha256": "3353dc19545d8ea0de2ec192396a91216d4df4afb358b2d538ebb2e3151ec293", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/19.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/19.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "fasth3-4", @@ -1888,6 +2229,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1897,7 +2239,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 391.20716426102445, + "cold_load_s": 405.68627888010815, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1915,34 +2257,44 @@ "name": "@AIwithAliya", "url": "https://x.com/AIwithAliya/status/2083132770650571041" }, + "source_prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", "source_duration": "15s", - "prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", - "prompt_sha256": "1bece371d5c458f6ba5e2f3cfffa5837017493101b52b6a3ecb5ee037df9b894", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic wuxia mystery in a bamboo forest at night, in a restrained cold-blue, ink-green, charcoal and gray palette. Thin mist and fine drifting snow surround two period-clothed martial-arts investigators. Dense vertical bamboo fills the background; soft foreground leaves partially obscure a tight facial close-up of the first investigator. Cool, soft front-side light reveals natural skin texture while cold white mist-light glows in the distance. The investigator looks toward the other person with measured, wary attention. No modern objects, text, subtitles, watermarks or comic performance.\n[Shot 2] At 00:05.000, the camera cuts to the other investigator in matching close-up for a restrained reverse angle. Keep the same cold lighting, shallow depth of field and softly blurred falling snow. A small shift of the eyes toward the dark grove conveys the exchange of secret intelligence without adding invented dialogue. Clothing rustles quietly under the bamboo wind.\n[Shot 3] At 00:10.000, the camera cuts back to the first investigator seen through out-of-focus leaves. A slow, small push-in follows a controlled glance into the luminous mist. Hold the austere, tense expression until the end, with the distant forest remaining still except for snow and moving leaves. Photoreal period-drama lighting, no animation styling, bright daylight or over-smoothed skin.\n\noverall_soundscape: Thin wind passes through bamboo leaves with occasional dry stalk creaks. Soft period-clothing movement and faint footsteps remain close while falling snow and distant forest space stay quiet.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "11628027721bd267f57680bb10ff37aff3623b6ea6bbfb064178144173e3a828", + "duration": 15, "seed": 61, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Bamboo Forest Wuxia Mystery", "zh": "竹林武侠谜案" }, - "pipeline_s": 3.0060316137969494, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.855335155967623, "observed_sparse_calls": 144, - "video": "assets/fasth3-4/20.mp4", - "poster": "assets/fasth3-4/20.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/fasth3-4/20.mp4", + "poster": "assets/v2-15s/fasth3-4/20.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1568892, - "sha256": "f2a3d269549a62711b24c29f80dccb626fa234ec1f386059eddb3e59b79fdcd8", + "duration_s": 15.166666, + "bytes": 4361983, + "sha256": "f9e56708b01a7290fc25d84cf2aed2c80c8efce92b6fb836f5ccf62afcfd5c67", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/fasth3-4/20.mp4", + "dataset_path": "videos/v2-15s/fasth3-4/20.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -1957,6 +2309,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -1966,7 +2319,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -1984,36 +2337,45 @@ "name": "@albertgao", "url": "https://x.com/albertgao/status/2085137397134561487" }, + "source_prompt": "\"A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.\"\n\n- time: 15min ish\n- 768x768", "source_duration": "5s", - "prompt": "A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.", - "prompt_sha256": "6ede0fa934a1dc755b53a87f6747f8e9e39b87c912ac52370171dfac33815847", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cute little girl walks through a sunny forest when a small, friendly dinosaur with a light, cheerful English voice (S1) steps out, waves, and says: [English] hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style. Pace the meeting, sneeze and shared laughter over a continuous 15-second scene, allowing time to see the butterflies settle around both characters.\n\noverall_soundscape: Light forest wind, birds and small footsteps accompany the meeting. A comic sneeze releases a fluttering rush of butterfly wings, followed by the girl and dinosaur laughing together.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "954337465d95f9ab7b1fd512ba3879fa925f8e668f493e314c9d84a91e99fad3", + "duration": 15, "seed": 42, "adaptations": [ - "Removed source author runtime and resolution notes outside the quoted prompt." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 5s，本次明确扩展为 15 秒。" ], "source_title": { "en": "via MLX Serve by", "zh": "via MLX Serve by" }, - "pipeline_s": 5.8013729471713305, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 31.44722088892013, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/01.mp4", - "poster": "assets/hyperflow-8/01.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/01.mp4", + "poster": "assets/v2-15s/hyperflow-8/01.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1973152, - "sha256": "33368c3d9a62eae8fe86dc944d68d1f5ad0325b9609e44a84478f89642c1099f", + "duration_s": 15.166666, + "bytes": 3957369, + "sha256": "ec3d0204a2e0bc1d93061ca6d7635b2577babd6304e0ec69fef28f9ccbeb5c4f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/01.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/01.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2028,6 +2390,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2037,7 +2400,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2055,34 +2418,45 @@ "name": "@ivanfioravanti", "url": "https://x.com/ivanfioravanti/status/2086553101029290296" }, + "source_prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", "source_duration": "10s", - "prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", - "prompt_sha256": "713096bb780dfbaeb577baa49e60b44666cd2ce400e9ca9f97d8d00fcae796a0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] amateur handheld pov footage of a tourist with an excited, conversational English voice (S1) in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: [English] Wow look at that!, then they show back the room Use one uninterrupted 15-second handheld take. Switch off the room light around halfway through, hold the window view for the spoken reaction, then turn back to show the room.\n\noverall_soundscape: A steady enclosed-room ventilation hum and soft carpeted footsteps continue under slight handheld handling noise. A light-switch click marks the change to the darkened room.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "712f02d91f39b4d5a029cf4c3396609ac32d151511bf04949501f697139405e5", + "duration": 15, "seed": 43, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 10s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "The last thing you see in your first and last", "zh": "The last thing you see in your first and last" }, - "pipeline_s": 5.709236569702625, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 24.163682500831783, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/02.mp4", - "poster": "assets/hyperflow-8/02.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/02.mp4", + "poster": "assets/v2-15s/hyperflow-8/02.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2396457, - "sha256": "e5ca86bb480e131fc827d46305b0c55ce1212d2b3b638bd7b4b666557f57fc81", + "duration_s": 15.166666, + "bytes": 4973347, + "sha256": "29886ab08f788882031ddd88b26af0c705a34324f8308f921f76bece5ed6042c", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/02.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/02.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2097,6 +2471,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2106,7 +2481,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2124,34 +2499,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2085622021812039920" }, + "source_prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", "source_duration": "15s", - "prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", - "prompt_sha256": "3de5bdf5968d3cb12d87fca7e975541ea97cee72e354a989fbe70ef0f19b7e10", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality. In one continuous 15-second camera move, establish the dawn lake first, raise and spiral the droplets in the middle, hold the completed readable word before its final liquid collapse.\n\noverall_soundscape: Quiet mountain wind, small lake ripples and faint distant dawn birds establish the lakeside. Delicate water movement grows into the rushing liquid-letter collapse and a crisp shower of droplets.\n\nnon_diegetic_music: Slow, sustained string notes and sparse high piano tones rise gently with the water lettering and resolve as the droplets fall.", + "prompt_sha256": "7abb51ba6a7327cb9266b13da9a8fa218928aeb0d97f444c82dc95ea21c245c2", + "duration": 15, "seed": 44, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Ultra cinematic macro shot: A calm mountain lake at dawn", "zh": "Ultra cinematic macro shot: A calm mountain lake at dawn" }, - "pipeline_s": 5.709947870112956, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.701676580123603, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/03.mp4", - "poster": "assets/hyperflow-8/03.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/03.mp4", + "poster": "assets/v2-15s/hyperflow-8/03.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2542536, - "sha256": "836248b4dbfdfff0043948ab3f3d2b99efa63ab07bd36a860831a2570624f52f", + "duration_s": 15.166666, + "bytes": 4840961, + "sha256": "3326242d298f719888ba2075ef34d2b26721d5fa21b8b890f856c30a3735a9c6", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/03.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/03.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2166,6 +2551,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2175,7 +2561,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2193,34 +2579,44 @@ "name": "@HBCoop_", "url": "https://x.com/HBCoop_/status/2082827829172117943" }, + "source_prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", "source_duration": "15s", - "prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", - "prompt_sha256": "0f456ed5bd7c35beb4dfc627e3b3082bc499140191221431d6a8fd6aa413e6ed", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Photorealistic premium food commercial, extreme macro high-speed product cinematography. A single ripe strawberry falls toward a white ceramic bowl in a sunlit spring greenhouse. Its red seeds and moist surface are sharply resolved against soft green glasshouse bokeh. Keep the strawberry's screen position, size, orientation and downward motion consistent through every seasonal match cut. No hands, labels or additional fruit.\n[Shot 2] At 00:03.000, the shot transitions to a hot summer picnic in a seamless seasonal match cut around the same falling strawberry and bowl. Warm sunlight replaces the greenhouse light; distant insects accompany its uninterrupted descent.\n[Shot 3] At 00:06.000, the shot transitions to an autumn kitchen with amber window light, preserving the strawberry and bowl alignment and the same downward motion. Soft indoor room tone replaces the outdoor insects.\n[Shot 4] At 00:09.000, the shot transitions to a candlelit winter dining room. The strawberry completes its fall into fresh cream in the white bowl, sending one graceful crown-shaped splash upward. Hold the detailed splash and its settling ripples through the final seconds. One continuous musical phrase connects all four seasons.\n\noverall_soundscape: Spring birds and a gentle greenhouse breeze change to summer insects, quiet autumn kitchen ambience, then the still winter dining room. The final cream splash and falling droplets remain clearly audible.\n\nnon_diegetic_music: One uninterrupted light piano-and-plucked-string phrase runs at a moderate tempo across all four seasons, resolving with the strawberry splash.", + "prompt_sha256": "0c4614d6b1c88a86974b1393eab7a574817abc11c37bf72e3cb53e2d9bbf718f", + "duration": 15, "seed": 45, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Strawberry seasonal match-cut food commercial", "zh": "草莓四季匹配剪辑食品广告" }, - "pipeline_s": 5.675561238080263, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.846177545841783, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/04.mp4", - "poster": "assets/hyperflow-8/04.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/04.mp4", + "poster": "assets/v2-15s/hyperflow-8/04.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1214592, - "sha256": "bd55539944a9cafe189ddef7395c4332ba9ce55548a4e77d21affb875600a52c", + "duration_s": 15.166666, + "bytes": 1926025, + "sha256": "0e48a5923fb47456d97f0d7c8ede826a0847e77f590c38751e34469e1c23d0d9", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/04.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/04.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2235,6 +2631,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2244,7 +2641,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2262,34 +2659,44 @@ "name": "@AIandDesign", "url": "https://x.com/AIandDesign/status/2082522979339288619" }, + "source_prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", "source_duration": "15s", - "prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", - "prompt_sha256": "745f72b9452c89f647147db06593c1202b51e0af1a2f2b769f48d1bbdc3e0dfb", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown. Maintain one continuous 15-second rear chase shot with brief arcs that reveal the pursuers when the robot looks back; preserve the robot’s identity and forward motion.\n\noverall_soundscape: Rapid metallic footfalls and small robot servo movements run through the city ambience. Heavy pursuing footsteps, drone rotors and passing machinery track the chase without intelligible dialogue.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be75f61112b0d84692fb5ce91ced32ae50c3f1a3a1ad6b115fba2089944d4073", + "duration": 15, "seed": 46, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Little Robot Cyberpunk Escape", "zh": "小机器人赛博朋克逃亡" }, - "pipeline_s": 5.751464718952775, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 24.001835828181356, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/05.mp4", - "poster": "assets/hyperflow-8/05.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/05.mp4", + "poster": "assets/v2-15s/hyperflow-8/05.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2618700, - "sha256": "8b3b57456a54c715c4cf5ba5b584074dc4828c8ced027677fbc2f672ff612054", + "duration_s": 15.166666, + "bytes": 8365151, + "sha256": "5a5f5387a08887839fd19e863855fdad094eeea60f17b1c073bda1c73808f122", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/05.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/05.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2304,6 +2711,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2313,7 +2721,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2331,34 +2739,44 @@ "name": "@TaliaAariz", "url": "https://x.com/TaliaAariz/status/2095857062987120666" }, + "source_prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", "source_duration": "15s", - "prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", - "prompt_sha256": "99600269200a60ed5566c134a5fdb59a19a7543be9c1e1e8bb9d68661a1d8320", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field. Keep one continuous 15-second medium shot, with a very slow small push-in as she touches the flowers, smiles and finally closes her eyes to feel the breeze.\n\noverall_soundscape: Gentle shore wind moves the ribbons and reeds while small waves wash over the pebbles. Quiet bouquet and fabric rustles accompany her hand movements.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7e05bce3aa444d99e753af9bcc4865bd2934e2c9102f5a1ce63f952302516d15", + "duration": 15, "seed": 47, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3 Max", "zh": "Created with MiniMax H3 Max" }, - "pipeline_s": 5.828220264986157, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.892320454120636, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/06.mp4", - "poster": "assets/hyperflow-8/06.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/06.mp4", + "poster": "assets/v2-15s/hyperflow-8/06.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1211131, - "sha256": "5565b103c8542e0b368de6f557107bc166630a102bb19bb11d0a76b841eb5547", + "duration_s": 15.166666, + "bytes": 3771535, + "sha256": "cf4bbb52bf1b4201bcc754a303d362d99a39385fd6c3fb0b44fb542bea1503e7", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/06.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/06.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2373,6 +2791,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2382,7 +2801,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2400,34 +2819,45 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087408671101989121" }, + "source_prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", "source_duration": "15s", - "prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", - "prompt_sha256": "022127583b887243e5d890afedd26196c9e3a7cf6665e714e0b8259b867c8c07", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western artist with a relaxed conversational English voice (S1) sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience. Her brief natural English remarks follow her painting actions. Keep the artist, easel and skyline consistent over the 15-second take. Begin with her positioning the easel, follow several visible finishing brush strokes, then give the final canvas-to-skyline comparison time to read.\n\noverall_soundscape: Soft city-park wind, distant traffic and lakeside birds surround the easel. Light wood movements, brush-on-canvas friction and cloth handling follow the artist’s actions.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "f39be188be70b922d6337aa1436e276c98592d5adf01475408b43bbd227e6cbc", + "duration": 15, "seed": 48, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "Created with MiniMax H3 on .", "zh": "Created with MiniMax H3 on ." }, - "pipeline_s": 5.682192385196686, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.84398057172075, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/07.mp4", - "poster": "assets/hyperflow-8/07.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/07.mp4", + "poster": "assets/v2-15s/hyperflow-8/07.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1687704, - "sha256": "0f986aeb1d2c4d974c87b90816b41d7e0d3c9cae395498bace039db98af7757c", + "duration_s": 15.166666, + "bytes": 4056216, + "sha256": "117ab831498a86b8358e40cebc9f29411d5128bbe55789ba448f97f8ed0a273f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/07.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/07.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2442,6 +2872,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2451,7 +2882,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2469,34 +2900,44 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087102541146522089" }, + "source_prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", "source_duration": "15s", - "prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", - "prompt_sha256": "1c1f768a47fd7db3bf3402c3e52f828f3de297d6a4ef32f449c7f75cb0cbb651", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western female street photographer with a warm conversational English voice (S1) walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says: [English] Look at that, then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism. Over one continuous 15-second handheld shot, follow her noticing and photographing the café scene, then turn with her as she shows the camera display, speaks and continues walking.\n\noverall_soundscape: Layered downtown footsteps, distant traffic and soft café room spill continue beneath a camera shutter click. The dog shifts and lightly pants beside the seated man.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be91933cc24647e7105ff883517384c8c83c2ccef42801f72555f509c467e319", + "duration": 15, "seed": 49, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3.", "zh": "Created with MiniMax H3." }, - "pipeline_s": 5.685499877203256, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 24.031339339911938, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/08.mp4", - "poster": "assets/hyperflow-8/08.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/08.mp4", + "poster": "assets/v2-15s/hyperflow-8/08.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2361223, - "sha256": "5698d0b112d5706d10da11573b94417b2075bfd25a8787f2777f5817c52be08b", + "duration_s": 15.166666, + "bytes": 11448912, + "sha256": "5fb8dd1d15ec864526e9c57e5e799bf2fcd2c2f719473609d84b533754ddc97a", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/08.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/08.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2511,6 +2952,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2520,7 +2962,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2538,34 +2980,45 @@ "name": "@Kiber_Alla", "url": "https://x.com/Kiber_Alla/status/2083583963512062019" }, + "source_prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", "source_duration": "15s", - "prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", - "prompt_sha256": "dd3b9a92fa180a0f865cf0f4e1869330f0aedf3b94bfa2a4df24ff87cd65ba1a", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. The woman, with a light melodic singing voice (S1), sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look. She performs a wordless vocal melody. Keep the low-angle backward tracking shot continuous for the full 15 seconds. Leave enough space for a full turn and return to eye contact without cutting.\n\noverall_soundscape: Light footsteps on stone, wind through the alley and natural fabric movement accompany the performance. Her soft laughter and breaths are synchronized to the visible movement.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7c2058cc5dc84a848b29169d61ecc607158e3ce4e68dd07d43622c10bfd34d91", + "duration": 15, "seed": 50, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词或歌词；未编造逐字台词，歌唱使用无词旋律。" + ], "source_title": { "en": "Low-Angle Fashion Tracking Film", "zh": "低机位时尚跟拍短片" }, - "pipeline_s": 5.714481103699654, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.710000686813146, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/09.mp4", - "poster": "assets/hyperflow-8/09.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/09.mp4", + "poster": "assets/v2-15s/hyperflow-8/09.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2280833, - "sha256": "0da62c87076c24f4f71fc50affc565e47993892d6d8e5ce22697d70798ec233a", + "duration_s": 15.166666, + "bytes": 9387986, + "sha256": "b776b84381fd14e090c6f68370a596687be78c1a3ff8ed7980f75bd5c59ee001", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/09.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/09.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2580,6 +3033,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2589,7 +3043,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2607,34 +3061,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085448660612182507" }, + "source_prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", "source_duration": "14s", - "prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", - "prompt_sha256": "f778e6b31c8426ea25d455ac33f5b0767d095059e1a69fbc2d2b97f0969b7eb0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose. Follow the complete obstacle sequence in one 15-second handheld take, with the bell contact visible in the final seconds.\n\noverall_soundscape: Backyard birds and distant neighborhood ambience surround quick paw taps and excited dog panting. Plastic poles rattle, the fabric tunnel scrapes, and a small bell rings once at the finish.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "b441f7d70cc0c7a54efccb096492e34839955225bc31781700cf560af438bc8d", + "duration": 15, "seed": 51, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Generated a french bulldog with MiniMax H3", "zh": "Generated a french bulldog with MiniMax H3" }, - "pipeline_s": 5.676533000078052, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.865956634748727, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/10.mp4", - "poster": "assets/hyperflow-8/10.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/10.mp4", + "poster": "assets/v2-15s/hyperflow-8/10.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 9921825, - "sha256": "3fac0cdede6603e85ca50e5b8442754d759cd93bd1991cd700b281958291e11c", + "duration_s": 15.166666, + "bytes": 16751851, + "sha256": "088ee1cefac41673416633b823644e1235cece97d406c905f7c1da686648af22", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/10.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/10.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2649,6 +3114,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2658,7 +3124,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2676,34 +3142,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2090740724803182619" }, + "source_prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", "source_duration": "15s", - "prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", - "prompt_sha256": "12a8a0c253466cbb68c3f4896f05c48ac47a2a4612fde6d00519fd3cf7813e0e", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX. Over 15 seconds, establish the glacier, follow the expanding fractures and airborne ice, hold the completed readable ice title, then finish on its violent shatter.\n\noverall_soundscape: Deep frozen-ocean wind surrounds spreading ice creaks and resonant fractures. Heavy ice collapse becomes a crystalline debris storm, followed by the sharp final title shatter.\n\nnon_diegetic_music: Sustained low strings swell with the ice fracture, joined by deep percussion at the shelf collapse and final title shatter.", + "prompt_sha256": "4450f406537c05f4eaf4263bdbb5e1e4d60f9b8e253dbe76aac8d0be5e2b24bd", + "duration": 15, "seed": 52, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with Hailuo H3.", "zh": "Created with Hailuo H3." }, - "pipeline_s": 5.724580345209688, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.703748255968094, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/11.mp4", - "poster": "assets/hyperflow-8/11.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/11.mp4", + "poster": "assets/v2-15s/hyperflow-8/11.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3410336, - "sha256": "c1871fe00ede709df88e20551d9a4b4062425838bdcd4800ce6fb671755aadad", + "duration_s": 15.166666, + "bytes": 6991570, + "sha256": "be32be7a2707097a2b5ef7e0147dbf2e25c5958f91a18aaf044c399192b37708", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/11.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/11.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2718,6 +3194,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2727,7 +3204,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2745,34 +3222,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2096554023809695817" }, + "source_prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", "source_duration": "15s", - "prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", - "prompt_sha256": "1622e4ef07a926746ba75db53791e1ee48aaa6d1da9cbdf512e40f27daf6cd0b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX. Keep the camera move continuous over 15 seconds: establish the floating monolith, enter the spreading fracture, orbit the suspended fragments, hold the completed title briefly, then finish on fragments expanding toward the lens.\n\noverall_soundscape: A projectile impact triggers fine traveling glass cracks and a swelling structural groan. Dense crystalline shattering and passing shards follow the orbit, with a second pressure-wave crack and outward burst at the end.\n\nnon_diegetic_music: A sparse low electronic pulse builds beneath bowed-metal tones, with two deep percussion hits at the main and final glass explosions.", + "prompt_sha256": "b198e1ddb0d1d5c79cab984ce06e7944ecd4efbca41f105960c7312c540dd572", + "duration": 15, "seed": 53, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "SHATTER.", "zh": "SHATTER." }, - "pipeline_s": 5.709502519108355, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.733664467930794, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/12.mp4", - "poster": "assets/hyperflow-8/12.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/12.mp4", + "poster": "assets/v2-15s/hyperflow-8/12.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2294823, - "sha256": "a185f4abe3c2a88b3662b9ff3557df57e19f2109a7a4515df76e0ed7ecc5f6dd", + "duration_s": 15.166666, + "bytes": 7373635, + "sha256": "a062da968a7a60bb9edfb346bef1ac22f146742ec85805717618a9fb0cf9f7f8", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/12.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/12.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2787,6 +3274,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2796,7 +3284,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2814,34 +3302,44 @@ "name": "@umesh_ai", "url": "https://x.com/umesh_ai/status/2082700637444452380" }, + "source_prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", "source_duration": "15s", - "prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", - "prompt_sha256": "543d21958bff63c8152dd8eafe797abae9972759bb8d0fe35a73fa0d456de3f4", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below. Perform the described camera path as one continuous 15-second move, ending on the high overhead view of the lone figure and exploding surf.\n\noverall_soundscape: Strong coastal wind drives the coat fabric. Deep ocean roar becomes clearer as the camera descends, with violent waves striking rock beneath the final overhead view.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "8a9fb882646cdb521b55ea777eea72c255e9b3863f257602d34d5bda4da3b879", + "duration": 15, "seed": 54, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Cliffside Descent to the Ocean", "zh": "悬崖下行至海边" }, - "pipeline_s": 5.6993877668865025, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.51624410180375, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/13.mp4", - "poster": "assets/hyperflow-8/13.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/13.mp4", + "poster": "assets/v2-15s/hyperflow-8/13.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3308591, - "sha256": "66320eb9db560183cd9ee0ec713bb435f6b437dc34702903c5d89d3e6b948e35", + "duration_s": 15.166666, + "bytes": 8705354, + "sha256": "30147ab7e53a46f154887e2dc9088eaa2cee82ebaad202295d5ca01aed313f6f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/13.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/13.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2856,6 +3354,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2865,7 +3364,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2883,36 +3382,45 @@ "name": "@ZaraIrahh", "url": "https://x.com/ZaraIrahh/status/2083011066800242986" }, + "source_prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, vertical 9:16, social media ad quality, highly realistic natural skin texture, not overly retouched.", "source_duration": "15s", - "prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched.", - "prompt_sha256": "992264f4313ef249573cc318eb959dd1e4f84f8c9ef13d0a6dc4fda46e877d40", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] UGC-style skincare video featuring a realistic young woman with a clear conversational English voice (S1) speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched. Her short natural English presentation stays about her skincare routine. Use a continuous 15-second landscape smartphone shot framed to keep both her face and the bottle clearly visible. She presents the serum, applies it to her face, and returns her attention to the camera.\n\noverall_soundscape: Quiet bathroom room tone, small glass-bottle handling sounds and soft hand-on-skin movements accompany the presentation. Keep breaths and clothing movement natural and close.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "61f30fc7218593117f92f6c7e46d55e03de10b31642a4122a69a9eba43d40b4c", + "duration": 15, "seed": 55, "adaptations": [ - "Normalized vertical 9:16 to landscape 16:9 for the shared 1344x768 canvas." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" ], "source_title": { "en": "Bathroom Mirror Skincare UGC Ad", "zh": "浴室镜前护肤 UGC 广告" }, - "pipeline_s": 5.551956397946924, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.412956648040563, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/14.mp4", - "poster": "assets/hyperflow-8/14.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/14.mp4", + "poster": "assets/v2-15s/hyperflow-8/14.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1036748, - "sha256": "76c06197c8a4b5b16ce4fba364b7561bced5f56b7b94ae3f10f44d9571e89d10", + "duration_s": 15.166666, + "bytes": 3956592, + "sha256": "c6efbe00f55c5e7736ebcc6ac56f5a4e4603300f667528537d64c3e047b1f1f6", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/14.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/14.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2927,6 +3435,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -2936,7 +3445,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -2954,34 +3463,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2084842607411335284" }, + "source_prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", "source_duration": "14s", - "prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", - "prompt_sha256": "11220e80abd1c920147a97e2b8ccc2436a5c0b8fecedb0b63751a2297c07f3af", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A low distorted English voice from the phone speaker (S1) says: [English] Finally. I’ve been trying to switch places with you. The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\" Keep the cheap boat-mounted camera viewpoint continuous for the 15-second scene, with only the specified white-flash glitch interrupting the image. Establish the fishing action, allow the phone line to finish, show the trapped image and then the wet double before the abrupt final black frame.\n\noverall_soundscape: Low lake water lapping, boat creaks and fishing-reel strain accompany the fisherman’s heavy breathing. Wet weeds scrape the hull, the phone glitches with clipped electronic noise, then water splashes as the soaked double enters the boat.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "778e3b3e9a6ca16d59312279338db624dd6281b72e5ade48402f63ed4b25f617", + "duration": 15, "seed": 56, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Horror Film Study 335284", "zh": "恐怖短片案例 335284" }, - "pipeline_s": 5.701896829996258, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.726625872775912, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/15.mp4", - "poster": "assets/hyperflow-8/15.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/15.mp4", + "poster": "assets/v2-15s/hyperflow-8/15.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3016883, - "sha256": "ea607cb04b31f8494c20d13e6e996c53bc652096a5c37b1c6d1ed631ce7b16b7", + "duration_s": 15.166666, + "bytes": 5922903, + "sha256": "4de8b4c7568ff123898736c87dcd8b8f9454877b6ef358b8db58be8a091089ce", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/15.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/15.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -2996,6 +3516,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3005,7 +3526,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3023,34 +3544,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085498501384024495" }, + "source_prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", "source_duration": "14s", - "prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", - "prompt_sha256": "03fa5bde85bff179c3e1b7bfcb9f378a3e9a1a931a80faff11dbf7ad57e1fc7d", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic science-fiction suspense, realistic spaceship crew in full futuristic uniforms boarding a dark abandoned spacecraft in response to a distress signal. A smooth tracking shot follows them through a narrow metallic corridor; flashlight beams cross flickering warning lights. One crew member with a restrained mid-pitched English voice (S1) says: [English] Signal's coming from deeper inside. Radio static, a low ship hum, creaking metal and hard boot echoes accompany their advance.\n[Shot 2] At 00:04.500, the camera cuts to a wide view of a large chamber with hundreds of open, empty cryogenic pods. A second crew member with an uneasy, slightly higher English voice (S2) looks from the abandoned pods toward a final closed door and says nervously: [English] Where did everyone go? The captain approaches that door as the camera pushes forward with the crew. Pod machinery resonates beneath their footsteps.\n[Shot 3] At 00:09.000, the camera cuts behind the captain as the door hisses open. Bright daylight reveals an ordinary present-day elementary school classroom, children at desks, backpacks on the floor and colorful posters on the walls. A teacher at the front with a clear, matter-of-fact English voice (S3) turns toward the astronauts and scolds them: [English] You're late again. Take your seats. The astronauts remain frozen in confusion at the doorway. Hold the absurd reveal until the end while ordinary classroom room tone replaces the ship ambience.\n\noverall_soundscape: Radio static, a low ship hum, creaking metal, boot echoes and empty pod machinery fill the spacecraft. A door hiss gives way abruptly to ordinary classroom chair movement and quiet paper rustling.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "515b6f9c02a9aff5ebdd48c593434b27848f0a398afc1c0651103dfdd475c0da", + "duration": 15, "seed": 57, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic", "zh": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic" }, - "pipeline_s": 5.803852798882872, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.7857776992023, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/16.mp4", - "poster": "assets/hyperflow-8/16.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/16.mp4", + "poster": "assets/v2-15s/hyperflow-8/16.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2384271, - "sha256": "03d752dfcf92b0624fabb96ae03da2729b768c19ec85bf86549522566540f6fb", + "duration_s": 15.166666, + "bytes": 5037351, + "sha256": "1c859affd1d6a1fbdba93bf8b545553ccefabcb063b2a6b3015742d037b6eb89", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/16.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/16.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -3065,6 +3597,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3074,7 +3607,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3092,34 +3625,44 @@ "name": "@Strength04_X", "url": "https://x.com/Strength04_X/status/2082712692159344891" }, + "source_prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", "source_duration": "15s", - "prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", - "prompt_sha256": "bc2776b295ff7d9b6bb9f534b11b75dd1a94c4d7a5666701b6abc440e047f69b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds. Complete the described path in one uninterrupted 15-second flight, establishing the skyship first and reserving the final seconds for the sunlight reveal beyond the storm.\n\noverall_soundscape: High-altitude wind and engine thrust deepen as the skyship enters thunder, lashing rain and cracking rock. Waterfalls rush past the hull before the storm sound falls behind the ship in open sunlight.\n\nnon_diegetic_music: Broad sustained strings and steady low orchestral percussion build through the storm, then open into a long brass-and-string chord in sunlight.", + "prompt_sha256": "15ed03a4ff9ad4d170616c8220d15f8c7ddadb16a059a522c567a1cede541838", + "duration": 15, "seed": 58, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Skyship Through a Floating Kingdom", "zh": "飞空艇穿越浮空王国" }, - "pipeline_s": 5.929556554183364, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.475021732971072, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/17.mp4", - "poster": "assets/hyperflow-8/17.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/17.mp4", + "poster": "assets/v2-15s/hyperflow-8/17.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2964140, - "sha256": "139322d6b17f8e981f066b9a4292b09009fc166278dd4846d32f0796662f6135", + "duration_s": 15.166666, + "bytes": 7945892, + "sha256": "b88420a0c61557f419b201fdf4d705aea052bcd62ce4650db2823e0d738c0265", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/17.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/17.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -3134,6 +3677,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3143,7 +3687,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3161,34 +3705,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2086879654116495564" }, + "source_prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", "source_duration": "14s", - "prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", - "prompt_sha256": "7fd67989d4363071c6d2e8384ecf994ee181b236e5a2f03f81d32aaefc7d931c", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\" Use a single 15-second home-video take. Preserve the first several seconds of ordinary watering, begin the overhead crack after that peaceful lead-in, and end with the sky descending and camera falling.\n\noverall_soundscape: Ordinary backyard birds, distant traffic and a watering stream continue for several seconds. A sudden deep overhead crack becomes a heavy descending rumble, followed by startled breathing and the phone hitting the ground.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "42f8695e30223aeca06566bf0476f7882751c30d3db90ad0136585d517c6ba26", + "duration": 15, "seed": 59, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright", "zh": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright" }, - "pipeline_s": 5.886082746088505, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 24.25950285512954, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/18.mp4", - "poster": "assets/hyperflow-8/18.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/18.mp4", + "poster": "assets/v2-15s/hyperflow-8/18.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3684855, - "sha256": "99c8aee0ef953601ba9111dd6473b0b2faa3d49f348fa7e4562d69fb0ed101a8", + "duration_s": 15.166666, + "bytes": 9023305, + "sha256": "eb7a6df9d2979f591d1850c4697b5f7c1a645824a67252e5ba831859e2df1bd0", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/18.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/18.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -3203,6 +3758,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3212,7 +3768,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3230,34 +3786,45 @@ "name": "@junwatu", "url": "https://x.com/junwatu/status/2084840152715911311" }, + "source_prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", "source_duration": "15s", - "prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", - "prompt_sha256": "6545b609b933b4f6574b2aa774e0198a9e4380c5cff593db1e2cd3cfd429491f", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism. Nearby women bargaining in Javanese (S1) and vendors calling prices in Javanese (S2) remain overlapping background voices. Follow the vendor in one continuous 15-second handheld take, keeping the steaming bowl and his careful steps legible; finish with serving the bowl and the brief power flicker.\n\noverall_soundscape: Dense covered-market bustle includes footsteps, bowl clatter, steaming food and sacks brushing clothing. Slow motorcycle engines pass among the stalls, followed by a brief fluorescent power buzz and the uninterrupted market activity.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "644ee708e360b5247f1f86b9510bc54be46b1ede81be1dfd8e283227dfab466f", + "duration": 15, "seed": 60, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "also has an impressive understanding of different cultures", "zh": "also has an impressive understanding of different cultures" }, - "pipeline_s": 5.720277740154415, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.886915264185518, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/19.mp4", - "poster": "assets/hyperflow-8/19.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/19.mp4", + "poster": "assets/v2-15s/hyperflow-8/19.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3050276, - "sha256": "22e24507e6e317538f1ae4b491bb0bccd5394c8cdfdf3f5ec5db5777ab87f6dd", + "duration_s": 15.166666, + "bytes": 7437662, + "sha256": "71ebf0e6394afd8de29ada1beaccd0c2f8f1c228abd4c576f99e38855971c08f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/19.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/19.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "hyperflow-8", @@ -3272,6 +3839,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3281,7 +3849,7 @@ "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 424.0454792641103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3299,34 +3867,44 @@ "name": "@AIwithAliya", "url": "https://x.com/AIwithAliya/status/2083132770650571041" }, + "source_prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", "source_duration": "15s", - "prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", - "prompt_sha256": "1bece371d5c458f6ba5e2f3cfffa5837017493101b52b6a3ecb5ee037df9b894", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic wuxia mystery in a bamboo forest at night, in a restrained cold-blue, ink-green, charcoal and gray palette. Thin mist and fine drifting snow surround two period-clothed martial-arts investigators. Dense vertical bamboo fills the background; soft foreground leaves partially obscure a tight facial close-up of the first investigator. Cool, soft front-side light reveals natural skin texture while cold white mist-light glows in the distance. The investigator looks toward the other person with measured, wary attention. No modern objects, text, subtitles, watermarks or comic performance.\n[Shot 2] At 00:05.000, the camera cuts to the other investigator in matching close-up for a restrained reverse angle. Keep the same cold lighting, shallow depth of field and softly blurred falling snow. A small shift of the eyes toward the dark grove conveys the exchange of secret intelligence without adding invented dialogue. Clothing rustles quietly under the bamboo wind.\n[Shot 3] At 00:10.000, the camera cuts back to the first investigator seen through out-of-focus leaves. A slow, small push-in follows a controlled glance into the luminous mist. Hold the austere, tense expression until the end, with the distant forest remaining still except for snow and moving leaves. Photoreal period-drama lighting, no animation styling, bright daylight or over-smoothed skin.\n\noverall_soundscape: Thin wind passes through bamboo leaves with occasional dry stalk creaks. Soft period-clothing movement and faint footsteps remain close while falling snow and distant forest space stay quiet.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "11628027721bd267f57680bb10ff37aff3623b6ea6bbfb064178144173e3a828", + "duration": 15, "seed": 61, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Bamboo Forest Wuxia Mystery", "zh": "竹林武侠谜案" }, - "pipeline_s": 5.552706141024828, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 23.756963307969272, "observed_sparse_calls": 288, - "video": "assets/hyperflow-8/20.mp4", - "poster": "assets/hyperflow-8/20.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/hyperflow-8/20.mp4", + "poster": "assets/v2-15s/hyperflow-8/20.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1208853, - "sha256": "54d49bdf145141099336d5d91d65bbcedf34509e285aa2aa85b0dafcd8094d13", + "duration_s": 15.166666, + "bytes": 2224088, + "sha256": "da0a26f5047f011dea91c347b8300709f458951520b92f8142f96634e3764226", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/hyperflow-8/20.mp4", + "dataset_path": "videos/v2-15s/hyperflow-8/20.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3341,6 +3919,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3350,7 +3929,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3368,36 +3947,45 @@ "name": "@albertgao", "url": "https://x.com/albertgao/status/2085137397134561487" }, + "source_prompt": "\"A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.\"\n\n- time: 15min ish\n- 768x768", "source_duration": "5s", - "prompt": "A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.", - "prompt_sha256": "6ede0fa934a1dc755b53a87f6747f8e9e39b87c912ac52370171dfac33815847", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cute little girl walks through a sunny forest when a small, friendly dinosaur with a light, cheerful English voice (S1) steps out, waves, and says: [English] hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style. Pace the meeting, sneeze and shared laughter over a continuous 15-second scene, allowing time to see the butterflies settle around both characters.\n\noverall_soundscape: Light forest wind, birds and small footsteps accompany the meeting. A comic sneeze releases a fluttering rush of butterfly wings, followed by the girl and dinosaur laughing together.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "954337465d95f9ab7b1fd512ba3879fa925f8e668f493e314c9d84a91e99fad3", + "duration": 15, "seed": 42, "adaptations": [ - "Removed source author runtime and resolution notes outside the quoted prompt." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 5s，本次明确扩展为 15 秒。" ], "source_title": { "en": "via MLX Serve by", "zh": "via MLX Serve by" }, - "pipeline_s": 3.3090272787958384, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 20.255016945768148, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/01.mp4", - "poster": "assets/lightx2v-4/01.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/01.mp4", + "poster": "assets/v2-15s/lightx2v-4/01.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1828307, - "sha256": "3c913c8333f35a81c100e47ba3de3316bc79b53ea3d64ad865b1f45190c4eeda", + "duration_s": 15.166666, + "bytes": 5142008, + "sha256": "4da69e6aefa643d09e4f1dc828d000595704bf47c8c9405702280d7eee435c2b", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/01.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/01.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3412,6 +4000,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3421,7 +4010,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3439,34 +4028,45 @@ "name": "@ivanfioravanti", "url": "https://x.com/ivanfioravanti/status/2086553101029290296" }, + "source_prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", "source_duration": "10s", - "prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", - "prompt_sha256": "713096bb780dfbaeb577baa49e60b44666cd2ce400e9ca9f97d8d00fcae796a0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] amateur handheld pov footage of a tourist with an excited, conversational English voice (S1) in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: [English] Wow look at that!, then they show back the room Use one uninterrupted 15-second handheld take. Switch off the room light around halfway through, hold the window view for the spoken reaction, then turn back to show the room.\n\noverall_soundscape: A steady enclosed-room ventilation hum and soft carpeted footsteps continue under slight handheld handling noise. A light-switch click marks the change to the darkened room.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "712f02d91f39b4d5a029cf4c3396609ac32d151511bf04949501f697139405e5", + "duration": 15, "seed": 43, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 10s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "The last thing you see in your first and last", "zh": "The last thing you see in your first and last" }, - "pipeline_s": 3.1597913149744272, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 13.716562563087791, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/02.mp4", - "poster": "assets/lightx2v-4/02.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/02.mp4", + "poster": "assets/v2-15s/lightx2v-4/02.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2910744, - "sha256": "d78f0bb612dd344b27377831207b8b246ee07929eee5344a06eecbd118d4d81b", + "duration_s": 15.166666, + "bytes": 6542142, + "sha256": "581de6abe2dfddc1e4e6ddaaa79866d878d39805279caa989bab1866e72c9cbe", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/02.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/02.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3481,6 +4081,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3490,7 +4091,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3508,34 +4109,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2085622021812039920" }, + "source_prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", "source_duration": "15s", - "prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", - "prompt_sha256": "3de5bdf5968d3cb12d87fca7e975541ea97cee72e354a989fbe70ef0f19b7e10", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality. In one continuous 15-second camera move, establish the dawn lake first, raise and spiral the droplets in the middle, hold the completed readable word before its final liquid collapse.\n\noverall_soundscape: Quiet mountain wind, small lake ripples and faint distant dawn birds establish the lakeside. Delicate water movement grows into the rushing liquid-letter collapse and a crisp shower of droplets.\n\nnon_diegetic_music: Slow, sustained string notes and sparse high piano tones rise gently with the water lettering and resolve as the droplets fall.", + "prompt_sha256": "7abb51ba6a7327cb9266b13da9a8fa218928aeb0d97f444c82dc95ea21c245c2", + "duration": 15, "seed": 44, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Ultra cinematic macro shot: A calm mountain lake at dawn", "zh": "Ultra cinematic macro shot: A calm mountain lake at dawn" }, - "pipeline_s": 3.1653449540026486, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 18.244925754144788, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/03.mp4", - "poster": "assets/lightx2v-4/03.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/03.mp4", + "poster": "assets/v2-15s/lightx2v-4/03.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3129221, - "sha256": "1361c798af4785ba4afa04cf4d2bd1cb2ae9357edee2add96c9f7b164dc1475a", + "duration_s": 15.166666, + "bytes": 6477081, + "sha256": "e8c26039001c84bd402131b8f61cf482862bc84477e481015dcbb8f674203fba", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/03.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/03.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3550,6 +4161,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3559,7 +4171,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3577,34 +4189,44 @@ "name": "@HBCoop_", "url": "https://x.com/HBCoop_/status/2082827829172117943" }, + "source_prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", "source_duration": "15s", - "prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", - "prompt_sha256": "0f456ed5bd7c35beb4dfc627e3b3082bc499140191221431d6a8fd6aa413e6ed", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Photorealistic premium food commercial, extreme macro high-speed product cinematography. A single ripe strawberry falls toward a white ceramic bowl in a sunlit spring greenhouse. Its red seeds and moist surface are sharply resolved against soft green glasshouse bokeh. Keep the strawberry's screen position, size, orientation and downward motion consistent through every seasonal match cut. No hands, labels or additional fruit.\n[Shot 2] At 00:03.000, the shot transitions to a hot summer picnic in a seamless seasonal match cut around the same falling strawberry and bowl. Warm sunlight replaces the greenhouse light; distant insects accompany its uninterrupted descent.\n[Shot 3] At 00:06.000, the shot transitions to an autumn kitchen with amber window light, preserving the strawberry and bowl alignment and the same downward motion. Soft indoor room tone replaces the outdoor insects.\n[Shot 4] At 00:09.000, the shot transitions to a candlelit winter dining room. The strawberry completes its fall into fresh cream in the white bowl, sending one graceful crown-shaped splash upward. Hold the detailed splash and its settling ripples through the final seconds. One continuous musical phrase connects all four seasons.\n\noverall_soundscape: Spring birds and a gentle greenhouse breeze change to summer insects, quiet autumn kitchen ambience, then the still winter dining room. The final cream splash and falling droplets remain clearly audible.\n\nnon_diegetic_music: One uninterrupted light piano-and-plucked-string phrase runs at a moderate tempo across all four seasons, resolving with the strawberry splash.", + "prompt_sha256": "0c4614d6b1c88a86974b1393eab7a574817abc11c37bf72e3cb53e2d9bbf718f", + "duration": 15, "seed": 45, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Strawberry seasonal match-cut food commercial", "zh": "草莓四季匹配剪辑食品广告" }, - "pipeline_s": 3.151411049067974, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 16.116741803009063, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/04.mp4", - "poster": "assets/lightx2v-4/04.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/04.mp4", + "poster": "assets/v2-15s/lightx2v-4/04.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1557955, - "sha256": "7ea05bd839d8cb39a176a23f93fe52df549ea9ff96e0a04d4d94130c799aadc1", + "duration_s": 15.166666, + "bytes": 2238263, + "sha256": "2feb84ee84b23fd71ba578bcd4b6bc098a083f400b0810bc3370d61c04ff5116", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/04.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/04.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3619,6 +4241,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3628,7 +4251,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3646,34 +4269,44 @@ "name": "@AIandDesign", "url": "https://x.com/AIandDesign/status/2082522979339288619" }, + "source_prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", "source_duration": "15s", - "prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", - "prompt_sha256": "745f72b9452c89f647147db06593c1202b51e0af1a2f2b769f48d1bbdc3e0dfb", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown. Maintain one continuous 15-second rear chase shot with brief arcs that reveal the pursuers when the robot looks back; preserve the robot’s identity and forward motion.\n\noverall_soundscape: Rapid metallic footfalls and small robot servo movements run through the city ambience. Heavy pursuing footsteps, drone rotors and passing machinery track the chase without intelligible dialogue.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be75f61112b0d84692fb5ce91ced32ae50c3f1a3a1ad6b115fba2089944d4073", + "duration": 15, "seed": 46, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Little Robot Cyberpunk Escape", "zh": "小机器人赛博朋克逃亡" }, - "pipeline_s": 3.1433526701293886, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.297106197103858, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/05.mp4", - "poster": "assets/lightx2v-4/05.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/05.mp4", + "poster": "assets/v2-15s/lightx2v-4/05.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3243703, - "sha256": "51b0e44754d29be628e1c06ad0e7ba387e4989f506f4b5575c342fd330eb7a04", + "duration_s": 15.166666, + "bytes": 10620006, + "sha256": "8f16f00bbb3e6ebb7d9f7f704272365dddff319274fd9d826d68488cd43e6a18", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/05.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/05.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3688,6 +4321,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3697,7 +4331,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3715,34 +4349,44 @@ "name": "@TaliaAariz", "url": "https://x.com/TaliaAariz/status/2095857062987120666" }, + "source_prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", "source_duration": "15s", - "prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", - "prompt_sha256": "99600269200a60ed5566c134a5fdb59a19a7543be9c1e1e8bb9d68661a1d8320", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field. Keep one continuous 15-second medium shot, with a very slow small push-in as she touches the flowers, smiles and finally closes her eyes to feel the breeze.\n\noverall_soundscape: Gentle shore wind moves the ribbons and reeds while small waves wash over the pebbles. Quiet bouquet and fabric rustles accompany her hand movements.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7e05bce3aa444d99e753af9bcc4865bd2934e2c9102f5a1ce63f952302516d15", + "duration": 15, "seed": 47, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3 Max", "zh": "Created with MiniMax H3 Max" }, - "pipeline_s": 3.2213578899390996, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 20.539152700453997, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/06.mp4", - "poster": "assets/lightx2v-4/06.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/06.mp4", + "poster": "assets/v2-15s/lightx2v-4/06.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1447850, - "sha256": "c933535ef36b928a284bf6eed943edc812ce89d41e9f15ef1df10bb56497d760", + "duration_s": 15.166666, + "bytes": 4285643, + "sha256": "a5a4d1c83151ed98f369117c5fb540b6bab7e94e4b312cf8cb2435e607437d3b", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/06.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/06.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3757,6 +4401,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3766,7 +4411,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3784,34 +4429,45 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087408671101989121" }, + "source_prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", "source_duration": "15s", - "prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", - "prompt_sha256": "022127583b887243e5d890afedd26196c9e3a7cf6665e714e0b8259b867c8c07", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western artist with a relaxed conversational English voice (S1) sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience. Her brief natural English remarks follow her painting actions. Keep the artist, easel and skyline consistent over the 15-second take. Begin with her positioning the easel, follow several visible finishing brush strokes, then give the final canvas-to-skyline comparison time to read.\n\noverall_soundscape: Soft city-park wind, distant traffic and lakeside birds surround the easel. Light wood movements, brush-on-canvas friction and cloth handling follow the artist’s actions.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "f39be188be70b922d6337aa1436e276c98592d5adf01475408b43bbd227e6cbc", + "duration": 15, "seed": 48, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "Created with MiniMax H3 on .", "zh": "Created with MiniMax H3 on ." }, - "pipeline_s": 3.1336171398870647, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 17.762425834778696, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/07.mp4", - "poster": "assets/lightx2v-4/07.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/07.mp4", + "poster": "assets/v2-15s/lightx2v-4/07.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1482011, - "sha256": "b9a2a9638ff36335f8047519a37bde655d8b639e6a52c64fa110b009f9556f2c", + "duration_s": 15.166666, + "bytes": 3574357, + "sha256": "5d0518131fee9833d813d3a67e140ef81b3223837410bfc8d9cd27cbb6f799f1", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/07.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/07.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3826,6 +4482,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3835,7 +4492,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3853,34 +4510,44 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087102541146522089" }, + "source_prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", "source_duration": "15s", - "prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", - "prompt_sha256": "1c1f768a47fd7db3bf3402c3e52f828f3de297d6a4ef32f449c7f75cb0cbb651", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western female street photographer with a warm conversational English voice (S1) walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says: [English] Look at that, then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism. Over one continuous 15-second handheld shot, follow her noticing and photographing the café scene, then turn with her as she shows the camera display, speaks and continues walking.\n\noverall_soundscape: Layered downtown footsteps, distant traffic and soft café room spill continue beneath a camera shutter click. The dog shifts and lightly pants beside the seated man.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be91933cc24647e7105ff883517384c8c83c2ccef42801f72555f509c467e319", + "duration": 15, "seed": 49, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3.", "zh": "Created with MiniMax H3." }, - "pipeline_s": 3.128684302326292, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.008996727876365, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/08.mp4", - "poster": "assets/lightx2v-4/08.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/08.mp4", + "poster": "assets/v2-15s/lightx2v-4/08.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3114586, - "sha256": "022c5aa7b91a77f3e00ce6a746b466f41288217d39cdf1e74cae650ca35b4a4d", + "duration_s": 15.166666, + "bytes": 13065707, + "sha256": "9765bf4815257fb4c0160762b9d542f0d6c0b4ae33d8d855a6764563e4a1336f", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/08.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/08.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3895,6 +4562,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3904,7 +4572,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3922,34 +4590,45 @@ "name": "@Kiber_Alla", "url": "https://x.com/Kiber_Alla/status/2083583963512062019" }, + "source_prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", "source_duration": "15s", - "prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", - "prompt_sha256": "dd3b9a92fa180a0f865cf0f4e1869330f0aedf3b94bfa2a4df24ff87cd65ba1a", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. The woman, with a light melodic singing voice (S1), sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look. She performs a wordless vocal melody. Keep the low-angle backward tracking shot continuous for the full 15 seconds. Leave enough space for a full turn and return to eye contact without cutting.\n\noverall_soundscape: Light footsteps on stone, wind through the alley and natural fabric movement accompany the performance. Her soft laughter and breaths are synchronized to the visible movement.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7c2058cc5dc84a848b29169d61ecc607158e3ce4e68dd07d43622c10bfd34d91", + "duration": 15, "seed": 50, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词或歌词；未编造逐字台词，歌唱使用无词旋律。" + ], "source_title": { "en": "Low-Angle Fashion Tracking Film", "zh": "低机位时尚跟拍短片" }, - "pipeline_s": 3.1760036889463663, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 16.49164312472567, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/09.mp4", - "poster": "assets/lightx2v-4/09.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/09.mp4", + "poster": "assets/v2-15s/lightx2v-4/09.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3004500, - "sha256": "57356d28cb2304d30612081c57613282a3819ed7ff1a21361ef0dbedf4085958", + "duration_s": 15.166666, + "bytes": 9378621, + "sha256": "678760472fde5af4fca5d0ec8b3bcae07923cb356f22dd082feb0b44a463d410", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/09.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/09.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -3964,6 +4643,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -3973,7 +4653,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -3991,34 +4671,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085448660612182507" }, + "source_prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", "source_duration": "14s", - "prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", - "prompt_sha256": "f778e6b31c8426ea25d455ac33f5b0767d095059e1a69fbc2d2b97f0969b7eb0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose. Follow the complete obstacle sequence in one 15-second handheld take, with the bell contact visible in the final seconds.\n\noverall_soundscape: Backyard birds and distant neighborhood ambience surround quick paw taps and excited dog panting. Plastic poles rattle, the fabric tunnel scrapes, and a small bell rings once at the finish.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "b441f7d70cc0c7a54efccb096492e34839955225bc31781700cf560af438bc8d", + "duration": 15, "seed": 51, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Generated a french bulldog with MiniMax H3", "zh": "Generated a french bulldog with MiniMax H3" }, - "pipeline_s": 3.1459635561332107, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.970872349105775, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/10.mp4", - "poster": "assets/lightx2v-4/10.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/10.mp4", + "poster": "assets/v2-15s/lightx2v-4/10.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 14278267, - "sha256": "d1789c123ebadd8066e217f2452c3b9d721e06eb43534192c12e1feb83b15ed0", + "duration_s": 15.166666, + "bytes": 30330339, + "sha256": "2973b567354c664fa0787a4f6fc7839664cd48a2a53e5f8c964cf71846a261f4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/10.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/10.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4033,6 +4724,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4042,7 +4734,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4060,34 +4752,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2090740724803182619" }, + "source_prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", "source_duration": "15s", - "prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", - "prompt_sha256": "12a8a0c253466cbb68c3f4896f05c48ac47a2a4612fde6d00519fd3cf7813e0e", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX. Over 15 seconds, establish the glacier, follow the expanding fractures and airborne ice, hold the completed readable ice title, then finish on its violent shatter.\n\noverall_soundscape: Deep frozen-ocean wind surrounds spreading ice creaks and resonant fractures. Heavy ice collapse becomes a crystalline debris storm, followed by the sharp final title shatter.\n\nnon_diegetic_music: Sustained low strings swell with the ice fracture, joined by deep percussion at the shelf collapse and final title shatter.", + "prompt_sha256": "4450f406537c05f4eaf4263bdbb5e1e4d60f9b8e253dbe76aac8d0be5e2b24bd", + "duration": 15, "seed": 52, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with Hailuo H3.", "zh": "Created with Hailuo H3." }, - "pipeline_s": 3.2079184181056917, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.799380616284907, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/11.mp4", - "poster": "assets/lightx2v-4/11.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/11.mp4", + "poster": "assets/v2-15s/lightx2v-4/11.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3532456, - "sha256": "2c7bcabb9a11b94a880b70f45a947d5602b9610e280e90e93db432fe768b5890", + "duration_s": 15.166666, + "bytes": 8639754, + "sha256": "d5c4831cc35b069ed7127693276f1e630f1d3fcc162d6709f93d35839f494ba3", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/11.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/11.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4102,6 +4804,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4111,7 +4814,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4129,34 +4832,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2096554023809695817" }, + "source_prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", "source_duration": "15s", - "prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", - "prompt_sha256": "1622e4ef07a926746ba75db53791e1ee48aaa6d1da9cbdf512e40f27daf6cd0b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX. Keep the camera move continuous over 15 seconds: establish the floating monolith, enter the spreading fracture, orbit the suspended fragments, hold the completed title briefly, then finish on fragments expanding toward the lens.\n\noverall_soundscape: A projectile impact triggers fine traveling glass cracks and a swelling structural groan. Dense crystalline shattering and passing shards follow the orbit, with a second pressure-wave crack and outward burst at the end.\n\nnon_diegetic_music: A sparse low electronic pulse builds beneath bowed-metal tones, with two deep percussion hits at the main and final glass explosions.", + "prompt_sha256": "b198e1ddb0d1d5c79cab984ce06e7944ecd4efbca41f105960c7312c540dd572", + "duration": 15, "seed": 53, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "SHATTER.", "zh": "SHATTER." }, - "pipeline_s": 3.1384907788597047, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.225255922880024, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/12.mp4", - "poster": "assets/lightx2v-4/12.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/12.mp4", + "poster": "assets/v2-15s/lightx2v-4/12.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3282881, - "sha256": "f75e7de1f975792f9a993fea5363439078b1a6059318249936460919a57cd5cf", + "duration_s": 15.166666, + "bytes": 9688152, + "sha256": "79d2e7b5fd8cb7a542ae05d2bd8a6eee266dc487ccf6e1df09f2ff95de502ecb", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/12.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/12.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4171,6 +4884,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4180,7 +4894,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4198,34 +4912,44 @@ "name": "@umesh_ai", "url": "https://x.com/umesh_ai/status/2082700637444452380" }, + "source_prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", "source_duration": "15s", - "prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", - "prompt_sha256": "543d21958bff63c8152dd8eafe797abae9972759bb8d0fe35a73fa0d456de3f4", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below. Perform the described camera path as one continuous 15-second move, ending on the high overhead view of the lone figure and exploding surf.\n\noverall_soundscape: Strong coastal wind drives the coat fabric. Deep ocean roar becomes clearer as the camera descends, with violent waves striking rock beneath the final overhead view.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "8a9fb882646cdb521b55ea777eea72c255e9b3863f257602d34d5bda4da3b879", + "duration": 15, "seed": 54, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Cliffside Descent to the Ocean", "zh": "悬崖下行至海边" }, - "pipeline_s": 11.004745036363602, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 12.783019708935171, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/13.mp4", - "poster": "assets/lightx2v-4/13.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/13.mp4", + "poster": "assets/v2-15s/lightx2v-4/13.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3824470, - "sha256": "3bc2eaa4ee3067c9d88dff935569b59df4e744b6f544a602e136f6380bf32fbf", + "duration_s": 15.166666, + "bytes": 11382393, + "sha256": "8f04be08d61fb0eeac292a1e50db564cb6a5c4f6a533e2532e9cc05fe2e3173c", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/13.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/13.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4240,6 +4964,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4249,7 +4974,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4267,36 +4992,45 @@ "name": "@ZaraIrahh", "url": "https://x.com/ZaraIrahh/status/2083011066800242986" }, + "source_prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, vertical 9:16, social media ad quality, highly realistic natural skin texture, not overly retouched.", "source_duration": "15s", - "prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched.", - "prompt_sha256": "992264f4313ef249573cc318eb959dd1e4f84f8c9ef13d0a6dc4fda46e877d40", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] UGC-style skincare video featuring a realistic young woman with a clear conversational English voice (S1) speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched. Her short natural English presentation stays about her skincare routine. Use a continuous 15-second landscape smartphone shot framed to keep both her face and the bottle clearly visible. She presents the serum, applies it to her face, and returns her attention to the camera.\n\noverall_soundscape: Quiet bathroom room tone, small glass-bottle handling sounds and soft hand-on-skin movements accompany the presentation. Keep breaths and clothing movement natural and close.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "61f30fc7218593117f92f6c7e46d55e03de10b31642a4122a69a9eba43d40b4c", + "duration": 15, "seed": 55, "adaptations": [ - "Normalized vertical 9:16 to landscape 16:9 for the shared 1344x768 canvas." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" ], "source_title": { "en": "Bathroom Mirror Skincare UGC Ad", "zh": "浴室镜前护肤 UGC 广告" }, - "pipeline_s": 2.9658619118854403, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 12.813158273231238, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/14.mp4", - "poster": "assets/lightx2v-4/14.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/14.mp4", + "poster": "assets/v2-15s/lightx2v-4/14.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 983619, - "sha256": "5b3902225630f3456239fa197a2a0421211386c3f297d49643bd35b772986e8a", + "duration_s": 15.166666, + "bytes": 3760929, + "sha256": "f12345b2b2879ccf185824d6005ba44e9441d41c5ddf2497384b58d98c2ae188", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/14.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/14.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4311,6 +5045,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4320,7 +5055,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4338,34 +5073,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2084842607411335284" }, + "source_prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", "source_duration": "14s", - "prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", - "prompt_sha256": "11220e80abd1c920147a97e2b8ccc2436a5c0b8fecedb0b63751a2297c07f3af", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A low distorted English voice from the phone speaker (S1) says: [English] Finally. I’ve been trying to switch places with you. The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\" Keep the cheap boat-mounted camera viewpoint continuous for the 15-second scene, with only the specified white-flash glitch interrupting the image. Establish the fishing action, allow the phone line to finish, show the trapped image and then the wet double before the abrupt final black frame.\n\noverall_soundscape: Low lake water lapping, boat creaks and fishing-reel strain accompany the fisherman’s heavy breathing. Wet weeds scrape the hull, the phone glitches with clipped electronic noise, then water splashes as the soaked double enters the boat.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "778e3b3e9a6ca16d59312279338db624dd6281b72e5ade48402f63ed4b25f617", + "duration": 15, "seed": 56, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Horror Film Study 335284", "zh": "恐怖短片案例 335284" }, - "pipeline_s": 5.829288989305496, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.542793967295438, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/15.mp4", - "poster": "assets/lightx2v-4/15.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/15.mp4", + "poster": "assets/v2-15s/lightx2v-4/15.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3201553, - "sha256": "3f364caf272c7d8128559377dff26fcc5140fd2ed19cd277506b1706279b6fe6", + "duration_s": 15.166666, + "bytes": 7156287, + "sha256": "49106d72a9706c94dc771bfbcd63f14177fcd0e1b05b8deb47ac1f5bf9f877cc", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/15.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/15.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4380,6 +5126,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4389,7 +5136,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4407,34 +5154,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085498501384024495" }, + "source_prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", "source_duration": "14s", - "prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", - "prompt_sha256": "03fa5bde85bff179c3e1b7bfcb9f378a3e9a1a931a80faff11dbf7ad57e1fc7d", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic science-fiction suspense, realistic spaceship crew in full futuristic uniforms boarding a dark abandoned spacecraft in response to a distress signal. A smooth tracking shot follows them through a narrow metallic corridor; flashlight beams cross flickering warning lights. One crew member with a restrained mid-pitched English voice (S1) says: [English] Signal's coming from deeper inside. Radio static, a low ship hum, creaking metal and hard boot echoes accompany their advance.\n[Shot 2] At 00:04.500, the camera cuts to a wide view of a large chamber with hundreds of open, empty cryogenic pods. A second crew member with an uneasy, slightly higher English voice (S2) looks from the abandoned pods toward a final closed door and says nervously: [English] Where did everyone go? The captain approaches that door as the camera pushes forward with the crew. Pod machinery resonates beneath their footsteps.\n[Shot 3] At 00:09.000, the camera cuts behind the captain as the door hisses open. Bright daylight reveals an ordinary present-day elementary school classroom, children at desks, backpacks on the floor and colorful posters on the walls. A teacher at the front with a clear, matter-of-fact English voice (S3) turns toward the astronauts and scolds them: [English] You're late again. Take your seats. The astronauts remain frozen in confusion at the doorway. Hold the absurd reveal until the end while ordinary classroom room tone replaces the ship ambience.\n\noverall_soundscape: Radio static, a low ship hum, creaking metal, boot echoes and empty pod machinery fill the spacecraft. A door hiss gives way abruptly to ordinary classroom chair movement and quiet paper rustling.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "515b6f9c02a9aff5ebdd48c593434b27848f0a398afc1c0651103dfdd475c0da", + "duration": 15, "seed": 57, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic", "zh": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic" }, - "pipeline_s": 5.25302194012329, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.019365294836462, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/16.mp4", - "poster": "assets/lightx2v-4/16.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/16.mp4", + "poster": "assets/v2-15s/lightx2v-4/16.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3110896, - "sha256": "f65f0cb71f57aff3852776153642385f97a348e7f3e1f650450ada8838563051", + "duration_s": 15.166666, + "bytes": 6812769, + "sha256": "74f38fb47041f298c2839e70c5dcbd72ecbb7324d1f313abff9009167998a207", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/16.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/16.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4449,6 +5207,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4458,7 +5217,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4476,34 +5235,44 @@ "name": "@Strength04_X", "url": "https://x.com/Strength04_X/status/2082712692159344891" }, + "source_prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", "source_duration": "15s", - "prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", - "prompt_sha256": "bc2776b295ff7d9b6bb9f534b11b75dd1a94c4d7a5666701b6abc440e047f69b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds. Complete the described path in one uninterrupted 15-second flight, establishing the skyship first and reserving the final seconds for the sunlight reveal beyond the storm.\n\noverall_soundscape: High-altitude wind and engine thrust deepen as the skyship enters thunder, lashing rain and cracking rock. Waterfalls rush past the hull before the storm sound falls behind the ship in open sunlight.\n\nnon_diegetic_music: Broad sustained strings and steady low orchestral percussion build through the storm, then open into a long brass-and-string chord in sunlight.", + "prompt_sha256": "15ed03a4ff9ad4d170616c8220d15f8c7ddadb16a059a522c567a1cede541838", + "duration": 15, "seed": 58, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Skyship Through a Floating Kingdom", "zh": "飞空艇穿越浮空王国" }, - "pipeline_s": 8.276164689101279, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.037491537630558, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/17.mp4", - "poster": "assets/lightx2v-4/17.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/17.mp4", + "poster": "assets/v2-15s/lightx2v-4/17.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 4247857, - "sha256": "f6ee95186f7bf2255d1be4c5a28ebaca20a7899cc91163177f628b72005883b8", + "duration_s": 15.166666, + "bytes": 10047621, + "sha256": "087ed14d9efdd260f5d658f36b9323ce34f87ca864f119a92052a7f02cb7f5fa", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/17.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/17.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4518,6 +5287,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4527,7 +5297,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4545,34 +5315,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2086879654116495564" }, + "source_prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", "source_duration": "14s", - "prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", - "prompt_sha256": "7fd67989d4363071c6d2e8384ecf994ee181b236e5a2f03f81d32aaefc7d931c", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\" Use a single 15-second home-video take. Preserve the first several seconds of ordinary watering, begin the overhead crack after that peaceful lead-in, and end with the sky descending and camera falling.\n\noverall_soundscape: Ordinary backyard birds, distant traffic and a watering stream continue for several seconds. A sudden deep overhead crack becomes a heavy descending rumble, followed by startled breathing and the phone hitting the ground.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "42f8695e30223aeca06566bf0476f7882751c30d3db90ad0136585d517c6ba26", + "duration": 15, "seed": 59, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright", "zh": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright" }, - "pipeline_s": 6.300810041837394, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.118513136170805, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/18.mp4", - "poster": "assets/lightx2v-4/18.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/18.mp4", + "poster": "assets/v2-15s/lightx2v-4/18.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3520616, - "sha256": "c586ee01c5edfc3ed8e2e5a86dd3fe4132e16f56192c95e0c7f1aea18866e39c", + "duration_s": 15.166666, + "bytes": 8065659, + "sha256": "37441c49a03fb06169ddbfc0ec0fb8180e900f1faf59131d8805862a64b300d4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/18.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/18.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4587,6 +5368,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4596,7 +5378,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4614,34 +5396,45 @@ "name": "@junwatu", "url": "https://x.com/junwatu/status/2084840152715911311" }, + "source_prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", "source_duration": "15s", - "prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", - "prompt_sha256": "6545b609b933b4f6574b2aa774e0198a9e4380c5cff593db1e2cd3cfd429491f", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism. Nearby women bargaining in Javanese (S1) and vendors calling prices in Javanese (S2) remain overlapping background voices. Follow the vendor in one continuous 15-second handheld take, keeping the steaming bowl and his careful steps legible; finish with serving the bowl and the brief power flicker.\n\noverall_soundscape: Dense covered-market bustle includes footsteps, bowl clatter, steaming food and sacks brushing clothing. Slow motorcycle engines pass among the stalls, followed by a brief fluorescent power buzz and the uninterrupted market activity.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "644ee708e360b5247f1f86b9510bc54be46b1ede81be1dfd8e283227dfab466f", + "duration": 15, "seed": 60, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "also has an impressive understanding of different cultures", "zh": "also has an impressive understanding of different cultures" }, - "pipeline_s": 9.603563416749239, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 14.480799765791744, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/19.mp4", - "poster": "assets/lightx2v-4/19.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/19.mp4", + "poster": "assets/v2-15s/lightx2v-4/19.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3496951, - "sha256": "ce6e342ec4e9cfe292cc9d61fd3cd842b3f8542bb741ac32341003789a781d66", + "duration_s": 15.166666, + "bytes": 9210509, + "sha256": "e92844ee9bcaed12bee7b32749331af0eef81bb85ecc3ac0d8df48a1ee2a8242", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/19.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/19.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-4", @@ -4656,6 +5449,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4665,7 +5459,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "cold_load_s": 411.3973352322355, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4683,34 +5477,44 @@ "name": "@AIwithAliya", "url": "https://x.com/AIwithAliya/status/2083132770650571041" }, + "source_prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", "source_duration": "15s", - "prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", - "prompt_sha256": "1bece371d5c458f6ba5e2f3cfffa5837017493101b52b6a3ecb5ee037df9b894", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic wuxia mystery in a bamboo forest at night, in a restrained cold-blue, ink-green, charcoal and gray palette. Thin mist and fine drifting snow surround two period-clothed martial-arts investigators. Dense vertical bamboo fills the background; soft foreground leaves partially obscure a tight facial close-up of the first investigator. Cool, soft front-side light reveals natural skin texture while cold white mist-light glows in the distance. The investigator looks toward the other person with measured, wary attention. No modern objects, text, subtitles, watermarks or comic performance.\n[Shot 2] At 00:05.000, the camera cuts to the other investigator in matching close-up for a restrained reverse angle. Keep the same cold lighting, shallow depth of field and softly blurred falling snow. A small shift of the eyes toward the dark grove conveys the exchange of secret intelligence without adding invented dialogue. Clothing rustles quietly under the bamboo wind.\n[Shot 3] At 00:10.000, the camera cuts back to the first investigator seen through out-of-focus leaves. A slow, small push-in follows a controlled glance into the luminous mist. Hold the austere, tense expression until the end, with the distant forest remaining still except for snow and moving leaves. Photoreal period-drama lighting, no animation styling, bright daylight or over-smoothed skin.\n\noverall_soundscape: Thin wind passes through bamboo leaves with occasional dry stalk creaks. Soft period-clothing movement and faint footsteps remain close while falling snow and distant forest space stay quiet.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "11628027721bd267f57680bb10ff37aff3623b6ea6bbfb064178144173e3a828", + "duration": 15, "seed": 61, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Bamboo Forest Wuxia Mystery", "zh": "竹林武侠谜案" }, - "pipeline_s": 3.0142150623723865, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 15.904188649728894, "observed_sparse_calls": 144, - "video": "assets/lightx2v-4/20.mp4", - "poster": "assets/lightx2v-4/20.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-4/20.mp4", + "poster": "assets/v2-15s/lightx2v-4/20.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1250788, - "sha256": "d83f05a1cf0d139e6691b756cad2be9ddc9e65c12cabfe2380e3961162d5d267", + "duration_s": 15.166666, + "bytes": 2621884, + "sha256": "f7e0b03b55e9147092525f1e3ec0e4d69bdbb44cc0fe5a0e8ab160068ca904aa", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-4/20.mp4", + "dataset_path": "videos/v2-15s/lightx2v-4/20.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -4725,6 +5529,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4734,7 +5539,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4752,36 +5557,45 @@ "name": "@albertgao", "url": "https://x.com/albertgao/status/2085137397134561487" }, + "source_prompt": "\"A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.\"\n\n- time: 15min ish\n- 768x768", "source_duration": "5s", - "prompt": "A cute little girl walks through a sunny forest when a small, friendly dinosaur steps out, waves, and says hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style.", - "prompt_sha256": "6ede0fa934a1dc755b53a87f6747f8e9e39b87c912ac52370171dfac33815847", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cute little girl walks through a sunny forest when a small, friendly dinosaur with a light, cheerful English voice (S1) steps out, waves, and says: [English] hi. Suddenly, the dinosaur sneezes—and a shower of colorful butterflies bursts from its nose. They both laugh. Cute, colorful animated style. Pace the meeting, sneeze and shared laughter over a continuous 15-second scene, allowing time to see the butterflies settle around both characters.\n\noverall_soundscape: Light forest wind, birds and small footsteps accompany the meeting. A comic sneeze releases a fluttering rush of butterfly wings, followed by the girl and dinosaur laughing together.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "954337465d95f9ab7b1fd512ba3879fa925f8e668f493e314c9d84a91e99fad3", + "duration": 15, "seed": 42, "adaptations": [ - "Removed source author runtime and resolution notes outside the quoted prompt." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 5s，本次明确扩展为 15 秒。" ], "source_title": { "en": "via MLX Serve by", "zh": "via MLX Serve by" }, - "pipeline_s": 12.501029765699059, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 29.708924832753837, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/01.mp4", - "poster": "assets/lightx2v-8/01.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/01.mp4", + "poster": "assets/v2-15s/lightx2v-8/01.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1851520, - "sha256": "c68d94ef37bc81e16af7d705c3b91176b1ad0f54c2de13c7039947c721473318", + "duration_s": 15.166666, + "bytes": 4444968, + "sha256": "bb300ec0825a2ac8430d390e730d646e7caf697ed827934d120006de63060745", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/01.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/01.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -4796,6 +5610,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4805,7 +5620,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4823,34 +5638,45 @@ "name": "@ivanfioravanti", "url": "https://x.com/ivanfioravanti/status/2086553101029290296" }, + "source_prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", "source_duration": "10s", - "prompt": "amateur handheld pov footage of a tourist in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: \"Wow look at that!\", then they show back the room", - "prompt_sha256": "713096bb780dfbaeb577baa49e60b44666cd2ce400e9ca9f97d8d00fcae796a0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] amateur handheld pov footage of a tourist with an excited, conversational English voice (S1) in their plush and comfortable room of a space cruise, carpeted floor and comfy bed, large window shows a view of Gargantua blackhole, you can see their reflection in the window, they turn off the light half way through so we can see outside better and say: [English] Wow look at that!, then they show back the room Use one uninterrupted 15-second handheld take. Switch off the room light around halfway through, hold the window view for the spoken reaction, then turn back to show the room.\n\noverall_soundscape: A steady enclosed-room ventilation hum and soft carpeted footsteps continue under slight handheld handling noise. A light-switch click marks the change to the darkened room.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "712f02d91f39b4d5a029cf4c3396609ac32d151511bf04949501f697139405e5", + "duration": 15, "seed": 43, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 10s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "The last thing you see in your first and last", "zh": "The last thing you see in your first and last" }, - "pipeline_s": 7.9720315909944475, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.4092057608068, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/02.mp4", - "poster": "assets/lightx2v-8/02.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/02.mp4", + "poster": "assets/v2-15s/lightx2v-8/02.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2704950, - "sha256": "ad818a0ed33a6e6b9a30e2e5ad352649f714883a46f6271f448afe97b8477275", + "duration_s": 15.166666, + "bytes": 5588386, + "sha256": "93f9f21c99b7db75349f9ad37f089de929cee64a39fa46033c6b3c20760d9a20", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/02.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/02.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -4865,6 +5691,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4874,7 +5701,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4892,34 +5719,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2085622021812039920" }, + "source_prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", "source_duration": "15s", - "prompt": "Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality.", - "prompt_sha256": "3de5bdf5968d3cb12d87fca7e975541ea97cee72e354a989fbe70ef0f19b7e10", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic macro shot: A calm mountain lake at dawn reflects the first golden rays of sunlight. Tiny ripples spread across the mirror-like surface as morning mist drifts gracefully above the water. Billions of sparkling droplets slowly rise into the air, suspended as if gravity has stopped. They spiral together with incredible fluid realism, naturally sculpting the word \"MORNING\" entirely from crystal-clear water. The liquid letters shimmer with golden reflections before gently collapsing into a spectacular explosion of droplets illuminated by the sunrise. Hollywood title sequence, ultra-realistic fluid simulation, volumetric lighting, IMAX quality. In one continuous 15-second camera move, establish the dawn lake first, raise and spiral the droplets in the middle, hold the completed readable word before its final liquid collapse.\n\noverall_soundscape: Quiet mountain wind, small lake ripples and faint distant dawn birds establish the lakeside. Delicate water movement grows into the rushing liquid-letter collapse and a crisp shower of droplets.\n\nnon_diegetic_music: Slow, sustained string notes and sparse high piano tones rise gently with the water lettering and resolve as the droplets fall.", + "prompt_sha256": "7abb51ba6a7327cb9266b13da9a8fa218928aeb0d97f444c82dc95ea21c245c2", + "duration": 15, "seed": 44, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Ultra cinematic macro shot: A calm mountain lake at dawn", "zh": "Ultra cinematic macro shot: A calm mountain lake at dawn" }, - "pipeline_s": 13.366133071947843, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.10105863492936, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/03.mp4", - "poster": "assets/lightx2v-8/03.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/03.mp4", + "poster": "assets/v2-15s/lightx2v-8/03.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3068508, - "sha256": "0b770143038296df3c9d67bb8bcff777ba9d4af78efa15cb9751a6ed59d5c24e", + "duration_s": 15.166666, + "bytes": 6564820, + "sha256": "7d0669c435502ed2c7a25c2771be534d7846656231602cab6540bfb3c6d650b7", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/03.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/03.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -4934,6 +5771,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -4943,7 +5781,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -4961,34 +5799,44 @@ "name": "@HBCoop_", "url": "https://x.com/HBCoop_/status/2082827829172117943" }, + "source_prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", "source_duration": "15s", - "prompt": "Premium food commercial. Extreme macro shot of a single ripe strawberry falling toward a white ceramic bowl. Before it lands, the environment changes through four seamless seasonal match cuts: a sunlit spring greenhouse, a hot summer picnic, an autumn kitchen and a candlelit winter dining room. The strawberry stays in exactly the same screen position, size, orientation and downward motion through every transition. It lands in fresh cream during the winter scene, sending one graceful crown-shaped splash upward. Photorealistic food texture, high-speed product cinematography, clean commercial lighting. Each season has distinct environmental sound, but one continuous musical phrase connects the entire sequence. No hands, labels or additional fruit.", - "prompt_sha256": "0f456ed5bd7c35beb4dfc627e3b3082bc499140191221431d6a8fd6aa413e6ed", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Photorealistic premium food commercial, extreme macro high-speed product cinematography. A single ripe strawberry falls toward a white ceramic bowl in a sunlit spring greenhouse. Its red seeds and moist surface are sharply resolved against soft green glasshouse bokeh. Keep the strawberry's screen position, size, orientation and downward motion consistent through every seasonal match cut. No hands, labels or additional fruit.\n[Shot 2] At 00:03.000, the shot transitions to a hot summer picnic in a seamless seasonal match cut around the same falling strawberry and bowl. Warm sunlight replaces the greenhouse light; distant insects accompany its uninterrupted descent.\n[Shot 3] At 00:06.000, the shot transitions to an autumn kitchen with amber window light, preserving the strawberry and bowl alignment and the same downward motion. Soft indoor room tone replaces the outdoor insects.\n[Shot 4] At 00:09.000, the shot transitions to a candlelit winter dining room. The strawberry completes its fall into fresh cream in the white bowl, sending one graceful crown-shaped splash upward. Hold the detailed splash and its settling ripples through the final seconds. One continuous musical phrase connects all four seasons.\n\noverall_soundscape: Spring birds and a gentle greenhouse breeze change to summer insects, quiet autumn kitchen ambience, then the still winter dining room. The final cream splash and falling droplets remain clearly audible.\n\nnon_diegetic_music: One uninterrupted light piano-and-plucked-string phrase runs at a moderate tempo across all four seasons, resolving with the strawberry splash.", + "prompt_sha256": "0c4614d6b1c88a86974b1393eab7a574817abc11c37bf72e3cb53e2d9bbf718f", + "duration": 15, "seed": 45, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Strawberry seasonal match-cut food commercial", "zh": "草莓四季匹配剪辑食品广告" }, - "pipeline_s": 9.84859362617135, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.147748040035367, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/04.mp4", - "poster": "assets/lightx2v-8/04.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/04.mp4", + "poster": "assets/v2-15s/lightx2v-8/04.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1365673, - "sha256": "053d303545a25d2a456f53569d5bf6367139299901ac6e3ce35c24559c430c34", + "duration_s": 15.166666, + "bytes": 1963781, + "sha256": "4e377794bc98d94006d5de362057d2b6e7edd605444af045c3fab0fea5f7ec20", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/04.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/04.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5003,6 +5851,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5012,7 +5861,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5030,34 +5879,44 @@ "name": "@AIandDesign", "url": "https://x.com/AIandDesign/status/2082522979339288619" }, + "source_prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", "source_duration": "15s", - "prompt": "A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown.", - "prompt_sha256": "745f72b9452c89f647147db06593c1202b51e0af1a2f2b769f48d1bbdc3e0dfb", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A stunning action packed scene following a little robot running away from soldiers, drones and other things that are trying to get him. Chase cam, dystopian futuristic cyberpunk environment, cinematic lighting. The camera follows the robot from behind, chase cam style. Every now and then the robot looks behind him and the entities chasing it are shown. Maintain one continuous 15-second rear chase shot with brief arcs that reveal the pursuers when the robot looks back; preserve the robot’s identity and forward motion.\n\noverall_soundscape: Rapid metallic footfalls and small robot servo movements run through the city ambience. Heavy pursuing footsteps, drone rotors and passing machinery track the chase without intelligible dialogue.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be75f61112b0d84692fb5ce91ced32ae50c3f1a3a1ad6b115fba2089944d4073", + "duration": 15, "seed": 46, - "adaptations": [], - "source_title": { + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], + "source_title": { "en": "Little Robot Cyberpunk Escape", "zh": "小机器人赛博朋克逃亡" }, - "pipeline_s": 7.962908416055143, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.32196999480948, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/05.mp4", - "poster": "assets/lightx2v-8/05.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/05.mp4", + "poster": "assets/v2-15s/lightx2v-8/05.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2967742, - "sha256": "5d16379750e51ee0c782eda42244ff81e7c8175444b4df31c1da9096e18ed834", + "duration_s": 15.166666, + "bytes": 9521493, + "sha256": "2c52140f346b2b052a8e2dec2436cab24e57517d08abdde6671d2146192f8442", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/05.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/05.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5072,6 +5931,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5081,7 +5941,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5099,34 +5959,44 @@ "name": "@TaliaAariz", "url": "https://x.com/TaliaAariz/status/2095857062987120666" }, + "source_prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", "source_duration": "15s", - "prompt": "A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field.", - "prompt_sha256": "99600269200a60ed5566c134a5fdb59a19a7543be9c1e1e8bb9d68661a1d8320", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A cinematic medium shot of a gentle East Asian young woman with dark hair styled in a bun adorned with a blue-and-white checkered bow, standing on a sunny pebbled lakefront. She wears a sleeveless light blue polka-dot tulle dress with long, translucent white ribbons flowing from her shoulder that flutter gracefully in a gentle breeze. Holding a lush bouquet of pastel pink, baby blue, and white flowers, she smiles warmly at the camera, softly touches the flower petals, and tilts her head back slightly to close her eyes and feel the breeze. In the background, glistening blue lake water sparkles under the sun, bordered by tall green reeds along the shore and serene blue mountains under a bright, clear sky, captured in soft natural daylight with a shallow depth of field. Keep one continuous 15-second medium shot, with a very slow small push-in as she touches the flowers, smiles and finally closes her eyes to feel the breeze.\n\noverall_soundscape: Gentle shore wind moves the ribbons and reeds while small waves wash over the pebbles. Quiet bouquet and fabric rustles accompany her hand movements.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7e05bce3aa444d99e753af9bcc4865bd2934e2c9102f5a1ce63f952302516d15", + "duration": 15, "seed": 47, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3 Max", "zh": "Created with MiniMax H3 Max" }, - "pipeline_s": 10.940343246329576, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.228797840885818, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/06.mp4", - "poster": "assets/lightx2v-8/06.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/06.mp4", + "poster": "assets/v2-15s/lightx2v-8/06.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1381291, - "sha256": "7149acbd144af016572d0ce735f39fc4b8653413c42dee2af284d9a27b9e0fea", + "duration_s": 15.166666, + "bytes": 4388971, + "sha256": "dc2fefcde62a6c513dcf62f2bdcc75c4247d2288885cbf453dfe81db5d1a99e4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/06.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/06.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5141,6 +6011,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5150,7 +6021,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5168,34 +6039,45 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087408671101989121" }, + "source_prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", "source_duration": "15s", - "prompt": "A young Western artist sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience.", - "prompt_sha256": "022127583b887243e5d890afedd26196c9e3a7cf6665e714e0b8259b867c8c07", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western artist with a relaxed conversational English voice (S1) sets up an easel in a peaceful city park and paints the skyline overlooking a lake. As the artwork comes to life, she compares the finished painting with the real view, holding the canvas in front of the skyline and smiling proudly. Ultra-photorealistic visuals, cinematic documentary style, realistic brush strokes, authentic hand movements, natural English dialogue, handheld camera movement, detailed cityscape, premium lighting, seamless continuity, and immersive outdoor ambience. Her brief natural English remarks follow her painting actions. Keep the artist, easel and skyline consistent over the 15-second take. Begin with her positioning the easel, follow several visible finishing brush strokes, then give the final canvas-to-skyline comparison time to read.\n\noverall_soundscape: Soft city-park wind, distant traffic and lakeside birds surround the easel. Light wood movements, brush-on-canvas friction and cloth handling follow the artist’s actions.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "f39be188be70b922d6337aa1436e276c98592d5adf01475408b43bbd227e6cbc", + "duration": 15, "seed": 48, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "Created with MiniMax H3 on .", "zh": "Created with MiniMax H3 on ." }, - "pipeline_s": 7.701512912288308, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.21258383290842, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/07.mp4", - "poster": "assets/lightx2v-8/07.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/07.mp4", + "poster": "assets/v2-15s/lightx2v-8/07.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1762879, - "sha256": "0473e75b593999b6d62b18eabfae5990294b63727b9d9b1a99ea4edf60c76fd4", + "duration_s": 15.166666, + "bytes": 3948104, + "sha256": "9ed77d7a90bbe1912931c8d6781454641992bc26214a6831ab92cf9a61a6de53", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/07.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/07.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5210,6 +6092,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5219,7 +6102,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5237,34 +6120,44 @@ "name": "@aiwithaly", "url": "https://x.com/aiwithaly/status/2087102541146522089" }, + "source_prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", "source_duration": "15s", - "prompt": "A young Western female street photographer walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says “Look at that,” then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism.", - "prompt_sha256": "1c1f768a47fd7db3bf3402c3e52f828f3de297d6a4ef32f449c7f75cb0cbb651", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A young Western female street photographer with a warm conversational English voice (S1) walks through a lively downtown street and notices an elderly man sitting outside a café with his small dog. She carefully composes the candid moment through her camera, captures the photo, then turns the camera toward the viewer to proudly show the shot she just took. She smiles, says: [English] Look at that, then continues walking through the city. Ultra-photorealistic visuals, natural handheld documentary movement, realistic camera interaction, authentic facial expressions, accurate hand movements, realistic dog behavior, natural daylight, cinematic depth of field, continuous character consistency, immersive city ambience, premium documentary realism. Over one continuous 15-second handheld shot, follow her noticing and photographing the café scene, then turn with her as she shows the camera display, speaks and continues walking.\n\noverall_soundscape: Layered downtown footsteps, distant traffic and soft café room spill continue beneath a camera shutter click. The dog shifts and lightly pants beside the seated man.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "be91933cc24647e7105ff883517384c8c83c2ccef42801f72555f509c467e319", + "duration": 15, "seed": 49, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with MiniMax H3.", "zh": "Created with MiniMax H3." }, - "pipeline_s": 7.5979786603711545, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.277623313944787, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/08.mp4", - "poster": "assets/lightx2v-8/08.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/08.mp4", + "poster": "assets/v2-15s/lightx2v-8/08.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2658215, - "sha256": "fcbc0120d3de8fae218a87831a7ada667866246d9138625e3c31e039c50628b8", + "duration_s": 15.166666, + "bytes": 9975230, + "sha256": "4a663ea277af887bb157f498af70c7c735f2ea4afa77fdbf1736f89cc6e513c4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/08.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/08.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5279,6 +6172,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5288,7 +6182,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5306,34 +6200,45 @@ "name": "@Kiber_Alla", "url": "https://x.com/Kiber_Alla/status/2083583963512062019" }, + "source_prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", "source_duration": "15s", - "prompt": "Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. She sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look.", - "prompt_sha256": "dd3b9a92fa180a0f865cf0f4e1869330f0aedf3b94bfa2a4df24ff87cd65ba1a", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra realistic cinematic video, shot on ARRI Alexa 35 with spherical lens, natural daylight, photorealistic, looks like real footage filmed on location, high-end fashion film aesthetic, extreme detail, realistic skin texture, real fabric physics, subtle film grain. Continuous low angle tracking shot, camera smoothly moving backward as a beautiful young woman with long dark wind-blown hair walks and performs down a narrow sun-drenched urban alley between old textured stone buildings under deep blue sky. She wears a flowing milky white asymmetrical off-shoulder top and high-waist wide-leg technical cotton culottes with a thin rope belt. The wide pants and fabric move naturally and dramatically with the wind and her body. The woman, with a light melodic singing voice (S1), sings with genuine joy, smiling and softly laughing, maintaining strong eye contact with the camera, while performing free, light dance elements — fluid arm gestures, body swaying, and elegant full turns around herself as she continues walking forward. Her movement is energetic yet effortless, full of lightness and positive energy. Cinematic natural lighting with strong sunlight and deep shadows, realistic motion blur, high dynamic range, masterpiece, best quality, no artificial look. She performs a wordless vocal melody. Keep the low-angle backward tracking shot continuous for the full 15 seconds. Leave enough space for a full turn and return to eye contact without cutting.\n\noverall_soundscape: Light footsteps on stone, wind through the alley and natural fabric movement accompany the performance. Her soft laughter and breaths are synchronized to the visible movement.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "7c2058cc5dc84a848b29169d61ecc607158e3ce4e68dd07d43622c10bfd34d91", + "duration": 15, "seed": 50, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词或歌词；未编造逐字台词，歌唱使用无词旋律。" + ], "source_title": { "en": "Low-Angle Fashion Tracking Film", "zh": "低机位时尚跟拍短片" }, - "pipeline_s": 7.36335717421025, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.00326671404764, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/09.mp4", - "poster": "assets/lightx2v-8/09.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/09.mp4", + "poster": "assets/v2-15s/lightx2v-8/09.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2695056, - "sha256": "543f6285320f2c32c5e51f9219c3f30afdcb450ee1cff13991ebfbae182baa85", + "duration_s": 15.166666, + "bytes": 9065393, + "sha256": "93292f48af79012c7f029b1610c960b4b626e8b189af59879a9b1d2b8d283659", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/09.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/09.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5348,6 +6253,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5357,7 +6263,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5375,34 +6281,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085448660612182507" }, + "source_prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", "source_duration": "14s", - "prompt": "A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose.", - "prompt_sha256": "f778e6b31c8426ea25d455ac33f5b0767d095059e1a69fbc2d2b97f0969b7eb0", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] A casual handheld smartphone video filmed by the dog's owner in an ordinary suburban backyard. A cute, stocky French bulldog excitedly runs through a small homemade agility course, awkwardly weaving between plastic poles, hopping over two low hurdles, scrambling through a short fabric tunnel, and finally bumping a little bell with its nose. Follow the complete obstacle sequence in one 15-second handheld take, with the bell contact visible in the final seconds.\n\noverall_soundscape: Backyard birds and distant neighborhood ambience surround quick paw taps and excited dog panting. Plastic poles rattle, the fabric tunnel scrapes, and a small bell rings once at the finish.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "b441f7d70cc0c7a54efccb096492e34839955225bc31781700cf560af438bc8d", + "duration": 15, "seed": 51, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Generated a french bulldog with MiniMax H3", "zh": "Generated a french bulldog with MiniMax H3" }, - "pipeline_s": 7.844183468725532, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.325014712288976, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/10.mp4", - "poster": "assets/lightx2v-8/10.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/10.mp4", + "poster": "assets/v2-15s/lightx2v-8/10.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 12600782, - "sha256": "d56ea7b764b85771567980471f8f779b7ca7202bb4e8dc8e45471bb9c7f655c5", + "duration_s": 15.166666, + "bytes": 29972293, + "sha256": "31f8a5f8e6ac2376f2cdfb2dc2719f5bae332cb55349b527d040c9b48ea6e052", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/10.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/10.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5417,6 +6334,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5426,7 +6344,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5444,34 +6362,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2090740724803182619" }, + "source_prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", "source_duration": "15s", - "prompt": "Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX.", - "prompt_sha256": "12a8a0c253466cbb68c3f4896f05c48ac47a2a4612fde6d00519fd3cf7813e0e", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Ultra cinematic title sequence: A colossal glacier fractures across an endless frozen ocean. Deep blue cracks spread for miles beneath translucent ice before the entire shelf collapses. Mountains of ice are launched into the sky, suspended in breathtaking slow motion. Every crystal catches the sunlight while frozen debris rotates around the camera. The crystalline storm assembles the word \"EXTINCTION\", carved from pure glacial ice with intricate frozen textures. Hairline fractures slowly spread through every letter until the title violently shatters into millions of glittering ice fragments. Photorealistic ice simulation, IMAX-scale environmental VFX. Over 15 seconds, establish the glacier, follow the expanding fractures and airborne ice, hold the completed readable ice title, then finish on its violent shatter.\n\noverall_soundscape: Deep frozen-ocean wind surrounds spreading ice creaks and resonant fractures. Heavy ice collapse becomes a crystalline debris storm, followed by the sharp final title shatter.\n\nnon_diegetic_music: Sustained low strings swell with the ice fracture, joined by deep percussion at the shelf collapse and final title shatter.", + "prompt_sha256": "4450f406537c05f4eaf4263bdbb5e1e4d60f9b8e253dbe76aac8d0be5e2b24bd", + "duration": 15, "seed": 52, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Created with Hailuo H3.", "zh": "Created with Hailuo H3." }, - "pipeline_s": 9.434969489928335, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.329117204993963, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/11.mp4", - "poster": "assets/lightx2v-8/11.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/11.mp4", + "poster": "assets/v2-15s/lightx2v-8/11.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3436016, - "sha256": "d3260a5b4c71433c1b1355c57474f839c1ba4f428c761611d71acb5387cb7f95", + "duration_s": 15.166666, + "bytes": 8513657, + "sha256": "7efb6225b1c2ed0f5271ab8bb8b8a5de528e9fa7aff2bf092dc8c23b7d38e538", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/11.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/11.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5486,6 +6414,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5495,7 +6424,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5513,34 +6442,44 @@ "name": "@CharaspowerAI", "url": "https://x.com/CharaspowerAI/status/2096554023809695817" }, + "source_prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", "source_duration": "15s", - "prompt": "Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX.", - "prompt_sha256": "1622e4ef07a926746ba75db53791e1ee48aaa6d1da9cbdf512e40f27daf6cd0b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Premium action-film title sequence beginning with a gigantic transparent glass monolith floating against complete blackness. A tiny projectile impacts the exact center and an intricate network of cracks instantly races across the entire surface. The camera pushes directly into one expanding fracture, traveling through the microscopic crystalline structure as tension builds everywhere. Suddenly the monolith violently detonates into millions of glass fragments. The camera performs a fast 360-degree orbital move through the suspended crystal storm as every shard reflects sharp beams of cinematic light. An invisible force begins pulling selected fragments together, constructing the enormous word \"SHATTER\" from overlapping razor-thin glass shards. The completed transparent typography refracts the entire environment like a giant prism before another pressure wave passes through it, producing thousands of secondary fractures. The letters explode outward directly toward the lens in spectacular ultra slow motion. Extremely detailed glass physics, realistic refraction and caustics, elegant brutality, premium Hollywood VFX. Keep the camera move continuous over 15 seconds: establish the floating monolith, enter the spreading fracture, orbit the suspended fragments, hold the completed title briefly, then finish on fragments expanding toward the lens.\n\noverall_soundscape: A projectile impact triggers fine traveling glass cracks and a swelling structural groan. Dense crystalline shattering and passing shards follow the orbit, with a second pressure-wave crack and outward burst at the end.\n\nnon_diegetic_music: A sparse low electronic pulse builds beneath bowed-metal tones, with two deep percussion hits at the main and final glass explosions.", + "prompt_sha256": "b198e1ddb0d1d5c79cab984ce06e7944ecd4efbca41f105960c7312c540dd572", + "duration": 15, "seed": 53, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "SHATTER.", "zh": "SHATTER." }, - "pipeline_s": 10.233981292694807, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.027931007090956, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/12.mp4", - "poster": "assets/lightx2v-8/12.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/12.mp4", + "poster": "assets/v2-15s/lightx2v-8/12.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2856314, - "sha256": "33c4051394b2d5014d0a555680606a1c0b1bcc78777dcb56c995e23eef955a0b", + "duration_s": 15.166666, + "bytes": 8906760, + "sha256": "55e4aa88d12443b97ec5c6eb07d3ff9129352b281785f346b15da0e475bc0242", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/12.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/12.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5555,6 +6494,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5564,7 +6504,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5582,34 +6522,44 @@ "name": "@umesh_ai", "url": "https://x.com/umesh_ai/status/2082700637444452380" }, + "source_prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", "source_duration": "15s", - "prompt": "The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below.", - "prompt_sha256": "543d21958bff63c8152dd8eafe797abae9972759bb8d0fe35a73fa0d456de3f4", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] The shot opens tight on a beautiful woman face, wind tearing at her coat as she stands on a jagged cliff. The camera eases over her shoulder and tilts down, following her line of sight. We descend past the crumbling lip into the void, the frame widening to reveal cliffs dropping away. The ocean roars into view, violent surf detonating against rock. The move continues into a high overhead, locking into an aerial top-down that frames the lone figure at the precipice as waves explode below. Perform the described camera path as one continuous 15-second move, ending on the high overhead view of the lone figure and exploding surf.\n\noverall_soundscape: Strong coastal wind drives the coat fabric. Deep ocean roar becomes clearer as the camera descends, with violent waves striking rock beneath the final overhead view.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "8a9fb882646cdb521b55ea777eea72c255e9b3863f257602d34d5bda4da3b879", + "duration": 15, "seed": 54, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Cliffside Descent to the Ocean", "zh": "悬崖下行至海边" }, - "pipeline_s": 13.589119987096637, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 21.941673386842012, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/13.mp4", - "poster": "assets/lightx2v-8/13.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/13.mp4", + "poster": "assets/v2-15s/lightx2v-8/13.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3268512, - "sha256": "1eb2d749ff64c609752e2d1ef37b691f695cc1cc2d33d15a20bf8c92272e78b6", + "duration_s": 15.166666, + "bytes": 9885251, + "sha256": "b97f41616ef6c9c604d21015046bbfcb6ca537dfc7891e6d12a55c5e0ca64311", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/13.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/13.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5624,6 +6574,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5633,7 +6584,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5651,36 +6602,45 @@ "name": "@ZaraIrahh", "url": "https://x.com/ZaraIrahh/status/2083011066800242986" }, + "source_prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, vertical 9:16, social media ad quality, highly realistic natural skin texture, not overly retouched.", "source_duration": "15s", - "prompt": "UGC-style skincare video featuring a realistic young woman speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched.", - "prompt_sha256": "992264f4313ef249573cc318eb959dd1e4f84f8c9ef13d0a6dc4fda46e877d40", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] UGC-style skincare video featuring a realistic young woman with a clear conversational English voice (S1) speaking directly to the camera in front of a bathroom mirror while holding a premium facial serum bottle. She smiles naturally, talks about her skincare routine, and demonstrates applying the serum to her face. Authentic, conversational presentation with genuine facial expressions and hand gestures, natural morning light, casual home bathroom setting, clean skincare aesthetic, healthy glowing skin, minimal makeup, smartphone-shot look, candid lifestyle beauty content, soft depth of field, luxury beauty brand feel, landscape 16:9, social media ad quality, highly realistic natural skin texture, not overly retouched. Her short natural English presentation stays about her skincare routine. Use a continuous 15-second landscape smartphone shot framed to keep both her face and the bottle clearly visible. She presents the serum, applies it to her face, and returns her attention to the camera.\n\noverall_soundscape: Quiet bathroom room tone, small glass-bottle handling sounds and soft hand-on-skin movements accompany the presentation. Keep breaths and clothing movement natural and close.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "61f30fc7218593117f92f6c7e46d55e03de10b31642a4122a69a9eba43d40b4c", + "duration": 15, "seed": 55, "adaptations": [ - "Normalized vertical 9:16 to landscape 16:9 for the shared 1344x768 canvas." + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" ], "source_title": { "en": "Bathroom Mirror Skincare UGC Ad", "zh": "浴室镜前护肤 UGC 广告" }, - "pipeline_s": 5.502857238985598, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 21.861816126853228, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/14.mp4", - "poster": "assets/lightx2v-8/14.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/14.mp4", + "poster": "assets/v2-15s/lightx2v-8/14.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1093887, - "sha256": "c4f228ac8e934cdc56e1a02f4b7aed6723afe25cbd89d3279fa60fa87ba8ba87", + "duration_s": 15.166666, + "bytes": 3979216, + "sha256": "4d77b24dadb5f8d7184438f0f6f764ed31c8c0067197fe7835c7f91dd5d3e727", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/14.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/14.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5695,6 +6655,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5704,7 +6665,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5722,34 +6683,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2084842607411335284" }, + "source_prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", "source_duration": "14s", - "prompt": "\"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A distorted voice says, \"Finally. I’ve been trying to switch places with you.\" The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\"", - "prompt_sha256": "11220e80abd1c920147a97e2b8ccc2436a5c0b8fecedb0b63751a2297c07f3af", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"Hyper-realistic found-footage horror video, filmed by a cheap action camera mounted inside a small fishing boat on a quiet lake. A lone fisherman violently reels in something heavy. He pulls up a muddy smartphone tangled in weeds. He answers, breathing hard. A low distorted English voice from the phone speaker (S1) says: [English] Finally. I’ve been trying to switch places with you. The phone flashes blinding white and the footage glitches. When the image returns, the phone shows the terrified fisherman trapped inside its screen, silently pounding on the glass. Behind him, an identical soaking-wet fisherman slowly climbs out of the lake into the boat, notices the camera, and reaches toward the lens. Abrupt cut to black. Raw handheld realism, imperfect autofocus, clipped audio, no music, no stylization.\" Keep the cheap boat-mounted camera viewpoint continuous for the 15-second scene, with only the specified white-flash glitch interrupting the image. Establish the fishing action, allow the phone line to finish, show the trapped image and then the wet double before the abrupt final black frame.\n\noverall_soundscape: Low lake water lapping, boat creaks and fishing-reel strain accompany the fisherman’s heavy breathing. Wet weeds scrape the hull, the phone glitches with clipped electronic noise, then water splashes as the soaked double enters the boat.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "778e3b3e9a6ca16d59312279338db624dd6281b72e5ade48402f63ed4b25f617", + "duration": 15, "seed": 56, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "Horror Film Study 335284", "zh": "恐怖短片案例 335284" }, - "pipeline_s": 5.73663988802582, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.16846158914268, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/15.mp4", - "poster": "assets/lightx2v-8/15.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/15.mp4", + "poster": "assets/v2-15s/lightx2v-8/15.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2517197, - "sha256": "f2b298bfe9f4ecebe41b060ac85aa056a19dd67cfd8c70b5551b4b5f54f2f150", + "duration_s": 15.166666, + "bytes": 6226808, + "sha256": "384609c46099c2f747abbe61a659f1dd061a6229b6cb60bc151dc890c005ffcc", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/15.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/15.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5764,6 +6736,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5773,7 +6746,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5791,34 +6764,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2085498501384024495" }, + "source_prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", "source_duration": "14s", - "prompt": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic uniforms follows a mysterious distress signal and boards a dark, abandoned spacecraft drifting in space. Flashlights cut through darkness as they move cautiously through narrow metallic corridors. Warning lights flicker, metal creaks, and their boots echo on the floor. One crew member says, \"Signal's coming from deeper inside.\" They enter a large chamber filled with hundreds of empty cryogenic pods, all open and abandoned. Another crew member says nervously, \"Where did everyone go?\" They approach a final closed door at the far end. The captain slowly opens it, and instead of another ship compartment, it reveals a completely normal present-day elementary school classroom in bright daylight. Children sit at desks, backpacks on the floor, colorful posters on the walls, and a teacher at the front turns to the crew and scolds them: \"You're late again. Take your seats.\" The astronauts stand frozen in confusion at the doorway. Realistic tone, smooth camera movement, strong suspense buildup, with the final reveal clear, sudden, and absurd. Include realistic audio: radio static, low ship hum, footsteps, pod machinery ambience, door hiss, then ordinary classroom sounds and the teacher's voice.", - "prompt_sha256": "03fa5bde85bff179c3e1b7bfcb9f378a3e9a1a931a80faff11dbf7ad57e1fc7d", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic science-fiction suspense, realistic spaceship crew in full futuristic uniforms boarding a dark abandoned spacecraft in response to a distress signal. A smooth tracking shot follows them through a narrow metallic corridor; flashlight beams cross flickering warning lights. One crew member with a restrained mid-pitched English voice (S1) says: [English] Signal's coming from deeper inside. Radio static, a low ship hum, creaking metal and hard boot echoes accompany their advance.\n[Shot 2] At 00:04.500, the camera cuts to a wide view of a large chamber with hundreds of open, empty cryogenic pods. A second crew member with an uneasy, slightly higher English voice (S2) looks from the abandoned pods toward a final closed door and says nervously: [English] Where did everyone go? The captain approaches that door as the camera pushes forward with the crew. Pod machinery resonates beneath their footsteps.\n[Shot 3] At 00:09.000, the camera cuts behind the captain as the door hisses open. Bright daylight reveals an ordinary present-day elementary school classroom, children at desks, backpacks on the floor and colorful posters on the walls. A teacher at the front with a clear, matter-of-fact English voice (S3) turns toward the astronauts and scolds them: [English] You're late again. Take your seats. The astronauts remain frozen in confusion at the doorway. Hold the absurd reveal until the end while ordinary classroom room tone replaces the ship ambience.\n\noverall_soundscape: Radio static, a low ship hum, creaking metal, boot echoes and empty pod machinery fill the spacecraft. A door hiss gives way abruptly to ordinary classroom chair movement and quiet paper rustling.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "515b6f9c02a9aff5ebdd48c593434b27848f0a398afc1c0651103dfdd475c0da", + "duration": 15, "seed": 57, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic", "zh": "cinematic sci-fi suspense video. A realistic spaceship crew in full futuristic" }, - "pipeline_s": 5.749303802847862, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.159618015866727, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/16.mp4", - "poster": "assets/lightx2v-8/16.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/16.mp4", + "poster": "assets/v2-15s/lightx2v-8/16.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 2592139, - "sha256": "0ba4b53452b69cf343118bc0be6d3226fc0d8afa177a3d696351a2f3b49bcb0a", + "duration_s": 15.166666, + "bytes": 5828206, + "sha256": "051fa4ea1d9615dfbc7276abb34bd1e7b28cc4647f2b1a0d762be56bd9e75796", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/16.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/16.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5833,6 +6817,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5842,7 +6827,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5860,34 +6845,44 @@ "name": "@Strength04_X", "url": "https://x.com/Strength04_X/status/2082712692159344891" }, + "source_prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", "source_duration": "15s", - "prompt": "Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds.", - "prompt_sha256": "bc2776b295ff7d9b6bb9f534b11b75dd1a94c4d7a5666701b6abc440e047f69b", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Skyship flight across a floating kingdom (single continuous shot) From above a sea of endless clouds, the camera dives toward a majestic skyship soaring between gigantic floating islands. Lock-on: the airship accelerating into a violent thunderstorm suspended in the sky. The camera slingshots beneath the hull, whips across the towering sails, then drops tight behind the glowing engines: lightning tears across the clouds, rain lashes the deck, shattered rock fragments tumble through the air. A collapsing floating island sends massive boulders crashing into the storm; the captain steers beneath a breaking stone arch before threading between waterfalls pouring into the sky, hanging bridges and drifting islands in one uninterrupted line. The camera mirrors every impossible maneuver before bursting through the storm into brilliant sunlight, revealing an endless kingdom of floating continents stretching beyond the horizon above a sea of clouds. Complete the described path in one uninterrupted 15-second flight, establishing the skyship first and reserving the final seconds for the sunlight reveal beyond the storm.\n\noverall_soundscape: High-altitude wind and engine thrust deepen as the skyship enters thunder, lashing rain and cracking rock. Waterfalls rush past the hull before the storm sound falls behind the ship in open sunlight.\n\nnon_diegetic_music: Broad sustained strings and steady low orchestral percussion build through the storm, then open into a long brass-and-string chord in sunlight.", + "prompt_sha256": "15ed03a4ff9ad4d170616c8220d15f8c7ddadb16a059a522c567a1cede541838", + "duration": 15, "seed": 58, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Skyship Through a Floating Kingdom", "zh": "飞空艇穿越浮空王国" }, - "pipeline_s": 6.388358146417886, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 21.763935210183263, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/17.mp4", - "poster": "assets/lightx2v-8/17.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/17.mp4", + "poster": "assets/v2-15s/lightx2v-8/17.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3613818, - "sha256": "6c29cbf5b4af34b6cf7e01ad9a5904ccf6f4916b710818e555497bbc260ed594", + "duration_s": 15.166666, + "bytes": 9926134, + "sha256": "794cbc5060de760017731b8e163c12e1d49304ce3ee560ad9898f2a2d3fc3bd0", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/17.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/17.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5902,6 +6897,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5911,7 +6907,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5929,34 +6925,45 @@ "name": "@cocktailpeanut", "url": "https://x.com/cocktailpeanut/status/2086879654116495564" }, + "source_prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", "source_duration": "14s", - "prompt": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\"", - "prompt_sha256": "7fd67989d4363071c6d2e8384ecf994ee181b236e5a2f03f81d32aaefc7d931c", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] \"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright normal afternoon. A person casually waters plants near a fence while birds chirp, distant traffic hums, and everything feels completely ordinary and mundane. The camera drifts naturally like a real home video, capturing the lawn, patio furniture, and blue sky with soft white clouds. After several seconds of normal peaceful activity, a deep cracking sound suddenly comes from above. Without warning, the entire sky begins dropping straight downward like a gigantic solid ceiling, with the blue sky and clouds moving as one physical surface descending toward the yard. The person looks up in shock just as the sky rapidly fills the frame, swallowing the scene in a terrifying instant. The camera jerks and falls to the ground at the last moment. No cinematic buildup, no surreal visual style, it should feel like a totally normal real-life recording interrupted by one sudden impossible and visceral event.\" Use a single 15-second home-video take. Preserve the first several seconds of ordinary watering, begin the overhead crack after that peaceful lead-in, and end with the sky descending and camera falling.\n\noverall_soundscape: Ordinary backyard birds, distant traffic and a watering stream continue for several seconds. A sudden deep overhead crack becomes a heavy descending rumble, followed by startled breathing and the phone hitting the ground.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "42f8695e30223aeca06566bf0476f7882751c30d3db90ad0136585d517c6ba26", + "duration": 15, "seed": 59, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源条目元数据为 14s，本次明确扩展为 15 秒。" + ], "source_title": { "en": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright", "zh": "\"A hyper-realistic handheld phone video of a quiet suburban backyard on a bright" }, - "pipeline_s": 10.516135799698532, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.57516465522349, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/18.mp4", - "poster": "assets/lightx2v-8/18.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/18.mp4", + "poster": "assets/v2-15s/lightx2v-8/18.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3044967, - "sha256": "150f4a84b0a92046a4d407665e9c1323c9f3b648104b13fdf65a127aa46aacd9", + "duration_s": 15.166666, + "bytes": 8492778, + "sha256": "832f6c0b87fb6457b6f4e651b42f388838230ef03ed0dd6993bafd9301ce87e4", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/18.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/18.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -5971,6 +6978,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -5980,7 +6988,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -5998,34 +7006,45 @@ "name": "@junwatu", "url": "https://x.com/junwatu/status/2084840152715911311" }, + "source_prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", "source_duration": "15s", - "prompt": "Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism.", - "prompt_sha256": "6545b609b933b4f6574b2aa774e0198a9e4380c5cff593db1e2cd3cfd429491f", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Traditional Javanese market in Yogyakarta, early morning, present day, ultra-realistic handheld one-take. A bakso vendor hurries through impossibly narrow, crowded market aisles carrying a large steaming bowl of freshly prepared bakso. Shoppers and vendors step aside to let him pass. He weaves between vegetable stalls overflowing with chilies, shallots, and leafy greens, ducks beneath low-hanging tarps, squeezes past batik merchants and baskets of traditional snacks, and carefully avoids children running through the market. The market feels dense, humid, noisy, and full of life. He passes porters carrying heavy sacks on their shoulders, women bargaining in Javanese, and motorcycles slowly pushing through the narrow lane. He finally reaches a tiny food stall at the back of the market and serves the steaming bowl just as the power briefly flickers, leaving the market dim for a moment before the bustle continues. Oppressive realism, natural morning light mixed with fluorescent bulbs, worn concrete floors, damp walls, wooden stalls, hanging plastic signs, Javanese conversations, vendors calling out prices, clattering bowls, footsteps echoing through the covered market, immersive cinematic pacing. No cyberpunk, no stylization, documentary-like realism. Nearby women bargaining in Javanese (S1) and vendors calling prices in Javanese (S2) remain overlapping background voices. Follow the vendor in one continuous 15-second handheld take, keeping the steaming bowl and his careful steps legible; finish with serving the bowl and the brief power flicker.\n\noverall_soundscape: Dense covered-market bustle includes footsteps, bowl clatter, steaming food and sacks brushing clothing. Slow motorcycle engines pass among the stalls, followed by a brief fluorescent power buzz and the uninterrupted market activity.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "644ee708e360b5247f1f86b9510bc54be46b1ede81be1dfd8e283227dfab466f", + "duration": 15, "seed": 60, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。", + "源文未提供固定台词；仅保留原文要求的语言与说话动作。" + ], "source_title": { "en": "also has an impressive understanding of different cultures", "zh": "also has an impressive understanding of different cultures" }, - "pipeline_s": 5.905952313914895, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.259512493852526, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/19.mp4", - "poster": "assets/lightx2v-8/19.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/19.mp4", + "poster": "assets/v2-15s/lightx2v-8/19.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 3383181, - "sha256": "7628f7b72318052975ee6c85e83dc3fcdf696fcfada503d6daf2775265a67367", + "duration_s": 15.166666, + "bytes": 8432796, + "sha256": "d5fa2b9596441bb344e6a14ca2021d85d14ac975a638bc857d920f7ac7677b01", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/19.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/19.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" }, { "adapter": "lightx2v-8", @@ -6040,6 +7059,7 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, @@ -6049,7 +7069,7 @@ "audio_shift": 3.0, "adapter_alpha": 8, "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "cold_load_s": 433.3525244933553, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -6067,34 +7087,44 @@ "name": "@AIwithAliya", "url": "https://x.com/AIwithAliya/status/2083132770650571041" }, + "source_prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", "source_duration": "15s", - "prompt": "A 16:9 cinematic wuxia mystery set in a bamboo forest at night. Use a low-saturation palette of cold blue, ink green, charcoal, and gray. Thin mist fills the forest and fine snow drifts through the air. The mood is austere, lethal, and controlled, with the tension of a martial-arts sect investigating a case and exchanging secret intelligence.\n\nDeep in the forest, dense vertical bamboo fills the background while cold white mist-light glows in the distance. Soft, out-of-focus leaves partially obscure the foreground, creating the sense of watching from within the grove. A cool, soft front-side key lights the actors’ faces; backlight keeps the foreground dark and the distance luminous. Use shallow depth of field so leaves, snow, and bamboo dissolve into soft bokeh.\nPrioritize facial close-ups and measured shot/reverse-shot coverage. Keep the rhythm restrained but tense. Photoreal period-drama production value, cinematic lighting, and no modern elements. No subtitles, on-screen text, watermarks, modern clothing or architecture, animation styling, over-smoothed skin, bright daylight, or comic performance.", - "prompt_sha256": "1bece371d5c458f6ba5e2f3cfffa5837017493101b52b6a3ecb5ee037df9b894", - "duration": 5, + "prompt": "integrated_multimodal_description: [Shot 1] Live-action cinematic wuxia mystery in a bamboo forest at night, in a restrained cold-blue, ink-green, charcoal and gray palette. Thin mist and fine drifting snow surround two period-clothed martial-arts investigators. Dense vertical bamboo fills the background; soft foreground leaves partially obscure a tight facial close-up of the first investigator. Cool, soft front-side light reveals natural skin texture while cold white mist-light glows in the distance. The investigator looks toward the other person with measured, wary attention. No modern objects, text, subtitles, watermarks or comic performance.\n[Shot 2] At 00:05.000, the camera cuts to the other investigator in matching close-up for a restrained reverse angle. Keep the same cold lighting, shallow depth of field and softly blurred falling snow. A small shift of the eyes toward the dark grove conveys the exchange of secret intelligence without adding invented dialogue. Clothing rustles quietly under the bamboo wind.\n[Shot 3] At 00:10.000, the camera cuts back to the first investigator seen through out-of-focus leaves. A slow, small push-in follows a controlled glance into the luminous mist. Hold the austere, tense expression until the end, with the distant forest remaining still except for snow and moving leaves. Photoreal period-drama lighting, no animation styling, bright daylight or over-smoothed skin.\n\noverall_soundscape: Thin wind passes through bamboo leaves with occasional dry stalk creaks. Soft period-clothing movement and faint footsteps remain close while falling snow and distant forest space stay quiet.\n\nnon_diegetic_music: N/A", + "prompt_sha256": "11628027721bd267f57680bb10ff37aff3623b6ea6bbfb064178144173e3a828", + "duration": 15, "seed": 61, - "adaptations": [], + "adaptations": [ + "保留社区原文及作者链接；本次按 H3 T2VA 三字段、镜头和语音规则整理。", + "按用户要求生成 15 秒（362 帧 / 24 FPS），画幅统一为 1344×768；新增环境音和镜头节奏均属于本次整理。" + ], "source_title": { "en": "Bamboo Forest Wuxia Mystery", "zh": "竹林武侠谜案" }, - "pipeline_s": 5.5389588521793485, + "frames": 362, + "prompt_version": "v2-h3-15s", + "prompt_origin": "H3-format adaptation of attributed community source; not claimed to be the original author’s structured prompt", + "pipeline_s": 22.103271580766886, "observed_sparse_calls": 336, - "video": "assets/lightx2v-8/20.mp4", - "poster": "assets/lightx2v-8/20.jpg", + "video": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media/resolve/2d72924012f74288d6ce319ebfe915fa21162237/videos/v2-15s/lightx2v-8/20.mp4", + "poster": "assets/v2-15s/lightx2v-8/20.jpg", "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 1668933, - "sha256": "b1508a2ab3bdf951997d435afe1af26ec86aa940c1fc4741ee91cf1e8d86085b", + "duration_s": 15.166666, + "bytes": 2766861, + "sha256": "e8c29c07fde216be0a76f0028a57a49ee24caea50661a5fdbf27bb84d357cad1", "verification": "ffprobe frame count and complete audio/video decode" - } + }, + "local_video": "assets/v2-15s/lightx2v-8/20.mp4", + "dataset_path": "videos/v2-15s/lightx2v-8/20.mp4", + "media_dataset_revision": "2d72924012f74288d6ce319ebfe915fa21162237" } ], "benchmarks": [ @@ -6882,17 +7912,17 @@ 10, 0 ], - "runtime_variant": "Sol-H3-int64", + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "dense", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 408.11689551733434, + "cold_load_s": 425.4020757121034, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -6901,13 +7931,13 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 71.77961420686916, + "warmup_s": 65.33505730889738, "repeats_s": [ - 71.96028976980597, - 72.14366453094408, - 71.87395159387961 + 65.48566196393222, + 65.6046371506527, + 65.33273641113192 ], - "median_s": 71.96028976980597, + "median_s": 65.48566196393222, "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", @@ -6919,8 +7949,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 1635071, - "sha256": "32b29a029110098314abfe342730a5ad3c2501ad6b8f7d8d1c05be15351fa6ab", + "bytes": 1572026, + "sha256": "be9021f12dc83658342a1574ade7038e90c277c680c798519f7e319843b7735a", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -6937,17 +7967,17 @@ 10, 0 ], - "runtime_variant": "Sol-H3-int64", + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "dense", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 408.11689551733434, + "cold_load_s": 425.4020757121034, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -6956,13 +7986,13 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 137.42840397078544, + "warmup_s": 129.85651914030313, "repeats_s": [ - 137.85603942908347, - 137.42607980174944, - 136.93979512900114 + 129.51566505106166, + 129.90104290517047, + 129.23715650010854 ], - "median_s": 137.42607980174944, + "median_s": 129.51566505106166, "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", @@ -6974,8 +8004,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 1889995, - "sha256": "c8687863f2ed2460df95d6eb44c5534a5a57ae4de8cf3fdfbe0c36afb2784b4b", + "bytes": 1919343, + "sha256": "7d26055f0c76b99c1f09bc9365060170a8ad680c7164daa06a07f3727c5c67a5", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -6992,17 +8022,17 @@ 10, 0 ], - "runtime_variant": "Sol-H3-int64", + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "dense", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 408.11689551733434, + "cold_load_s": 425.4020757121034, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7011,13 +8041,13 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 40.22856629593298, + "warmup_s": 42.42403999203816, "repeats_s": [ - 25.559987880755216, - 25.68099537305534, - 25.446089582052082 + 22.414969386998564, + 22.482791357208043, + 22.4277695142664 ], - "median_s": 25.559987880755216, + "median_s": 22.4277695142664, "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", @@ -7029,8 +8059,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 932090, - "sha256": "88d411f81708016df471e2f0d46aa8a46968b9dec59c431649031cd3e12fef27", + "bytes": 833903, + "sha256": "efa2a6e0dc9de037cb13d7657b3ff8fc249ab5c691fd7f95d835e16772968e3b", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7039,7 +8069,7 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "hsg", - "gpus": 4, + "gpus": 1, "allocated_gpus_per_node": "4", "nodes": 1, "gpu": "NVIDIA GB200", @@ -7047,16 +8077,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "sol_bsa", + "attention_backend": "dense", "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 408.11689551733434, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7065,13 +8096,178 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 113.69129496766254, + "warmup_s": 71.77961420686916, "repeats_s": [ - 13.347808117046952, - 13.34682912286371, - 13.33125697216019 + 71.96028976980597, + 72.14366453094408, + 71.87395159387961 ], - "median_s": 13.34682912286371, + "median_s": 71.96028976980597, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 243, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 10.208344, + "bytes": 1635071, + "sha256": "32b29a029110098314abfe342730a5ad3c2501ad6b8f7d8d1c05be15351fa6ab", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, + "cluster": "hsg", + "gpus": 1, + "allocated_gpus_per_node": "4", + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3-int64", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "dense", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 408.11689551733434, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 15, + "frames": 362, + "seed": 20260903, + "warmup_s": 137.42840397078544, + "repeats_s": [ + 137.85603942908347, + 137.42607980174944, + 136.93979512900114 + ], + "median_s": 137.42607980174944, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 362, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 15.166666, + "bytes": 1889995, + "sha256": "c8687863f2ed2460df95d6eb44c5534a5a57ae4de8cf3fdfbe0c36afb2784b4b", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, + "cluster": "hsg", + "gpus": 1, + "allocated_gpus_per_node": "4", + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3-int64", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "dense", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 408.11689551733434, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 40.22856629593298, + "repeats_s": [ + 25.559987880755216, + 25.68099537305534, + 25.446089582052082 + ], + "median_s": 25.559987880755216, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 932090, + "sha256": "88d411f81708016df471e2f0d46aa8a46968b9dec59c431649031cd3e12fef27", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, + "cluster": "hsg", + "gpus": 4, + "allocated_gpus_per_node": "4", + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "mxfp8", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 479.468451640103, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 10, + "frames": 243, + "seed": 20260903, + "warmup_s": 18.84309404483065, + "repeats_s": [ + 11.544127263128757, + 11.55223888810724, + 11.559273229911923 + ], + "median_s": 11.55223888810724, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 2304, @@ -7107,10 +8303,10 @@ "route_density": { "blocks": 1147, "sink_blocks": 13, - "threshold_density": 0.20341, - "effective_density": 0.22407, - "route_count_min": 101, - "route_count_mean": 257.01, + "threshold_density": 0.20331, + "effective_density": 0.22398, + "route_count_min": 97, + "route_count_mean": 256.9, "route_count_max": 1147, "metadata_capacity": 1152 }, @@ -7130,8 +8326,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 1843870, - "sha256": "4aded58f1921b5870ea204051b6c73ef368c94b603f258ed0e6b2eee98628d0d", + "bytes": 1807659, + "sha256": "dae00a32244bf582679e80092cc171054e9ee4df1f8a235f892836c63f715979", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7148,16 +8344,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 479.468451640103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7166,13 +8363,13 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 235.1847071829252, + "warmup_s": 28.802728682756424, "repeats_s": [ - 24.262370044831187, - 24.287067759316415, - 24.273042456246912 + 21.485745965968817, + 21.545224134344608, + 21.510514890309423 ], - "median_s": 24.273042456246912, + "median_s": 21.510514890309423, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 3456, @@ -7208,10 +8405,10 @@ "route_density": { "blocks": 1705, "sink_blocks": 20, - "threshold_density": 0.20321, - "effective_density": 0.22349, + "threshold_density": 0.20332, + "effective_density": 0.22357, "route_count_min": 158, - "route_count_mean": 381.05, + "route_count_mean": 381.19, "route_count_max": 1705, "metadata_capacity": 1728 }, @@ -7231,8 +8428,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 1936290, - "sha256": "4d93e2878b88f9c775ab46f1df3233003f5fa5331742eac56a1c5c7df6926ea3", + "bytes": 1934424, + "sha256": "3ca76c8c6af7724a6bb79045521999ae61b82d38a7b6a40574838c334c1c3421", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7249,16 +8446,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 409.0616293284111, + "cold_load_s": 479.468451640103, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7267,13 +8465,13 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 30.525569978170097, + "warmup_s": 33.40463601704687, "repeats_s": [ - 5.545859571546316, - 5.51854232000187, - 5.509740656707436 + 4.606777192093432, + 4.593384526204318, + 4.615335638169199 ], - "median_s": 5.51854232000187, + "median_s": 4.606777192093432, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 1152, @@ -7309,10 +8507,10 @@ "route_density": { "blocks": 590, "sink_blocks": 7, - "threshold_density": 0.2084, - "effective_density": 0.23181, - "route_count_min": 52, - "route_count_mean": 136.77, + "threshold_density": 0.20832, + "effective_density": 0.23173, + "route_count_min": 51, + "route_count_mean": 136.72, "route_count_max": 590, "metadata_capacity": 640 }, @@ -7332,8 +8530,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 926954, - "sha256": "6522c05d8ea33aebbb4be2cdea7ee5e122061509a6de1d5ab3e72fe12249e360", + "bytes": 901555, + "sha256": "2ef7388b3b04cf6d9e61c0ab38192ac7a43c69651e9a8a4b5c8f7a5284525bb6", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7342,25 +8540,24 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "hsg", - "gpus": 8, + "gpus": 4, "allocated_gpus_per_node": "4", - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, 0 ], - "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "mxfp8", + "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 439.58678179606795, + "cold_load_s": 409.0616293284111, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7369,13 +8566,13 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 14.64438444795087, + "warmup_s": 113.69129496766254, "repeats_s": [ - 6.551806676667184, - 6.531784829683602, - 6.554025718010962 + 13.347808117046952, + 13.34682912286371, + 13.33125697216019 ], - "median_s": 6.551806676667184, + "median_s": 13.34682912286371, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 2304, @@ -7411,10 +8608,10 @@ "route_density": { "blocks": 1147, "sink_blocks": 13, - "threshold_density": 0.18735, - "effective_density": 0.20817, - "route_count_min": 144, - "route_count_mean": 238.77, + "threshold_density": 0.20341, + "effective_density": 0.22407, + "route_count_min": 101, + "route_count_mean": 257.01, "route_count_max": 1147, "metadata_capacity": 1152 }, @@ -7434,8 +8631,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 1825597, - "sha256": "04260482aa6b6d7ec1550321d6dc7e0f04d7fa8cce53714711313f84ac47fc84", + "bytes": 1843870, + "sha256": "4aded58f1921b5870ea204051b6c73ef368c94b603f258ed0e6b2eee98628d0d", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7444,25 +8641,24 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "hsg", - "gpus": 8, + "gpus": 4, "allocated_gpus_per_node": "4", - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, 0 ], - "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "mxfp8", + "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 439.58678179606795, + "cold_load_s": 409.0616293284111, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7471,13 +8667,13 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 19.447643855120987, + "warmup_s": 235.1847071829252, "repeats_s": [ - 12.063323335722089, - 12.064927336759865, - 12.09263978432864 - ], - "median_s": 12.064927336759865, + 24.262370044831187, + 24.287067759316415, + 24.273042456246912 + ], + "median_s": 24.273042456246912, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 3456, @@ -7513,10 +8709,10 @@ "route_density": { "blocks": 1705, "sink_blocks": 20, - "threshold_density": 0.18567, - "effective_density": 0.20671, - "route_count_min": 220, - "route_count_mean": 352.45, + "threshold_density": 0.20321, + "effective_density": 0.22349, + "route_count_min": 158, + "route_count_mean": 381.05, "route_count_max": 1705, "metadata_capacity": 1728 }, @@ -7536,8 +8732,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 1936513, - "sha256": "9fae8f6915cf26a28dbafda31f976c0cd4456c8b14f0a63ccc8b4010d00ac44b", + "bytes": 1936290, + "sha256": "4d93e2878b88f9c775ab46f1df3233003f5fa5331742eac56a1c5c7df6926ea3", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7546,25 +8742,24 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "hsg", - "gpus": 8, + "gpus": 4, "allocated_gpus_per_node": "4", - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, 0 ], - "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "mxfp8", + "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 439.58678179606795, + "cold_load_s": 409.0616293284111, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7573,13 +8768,13 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 35.399448071140796, + "warmup_s": 30.525569978170097, "repeats_s": [ - 2.8105423622764647, - 2.755053977947682, - 2.758813951164484 + 5.545859571546316, + 5.51854232000187, + 5.509740656707436 ], - "median_s": 2.758813951164484, + "median_s": 5.51854232000187, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 1152, @@ -7615,10 +8810,10 @@ "route_density": { "blocks": 590, "sink_blocks": 7, - "threshold_density": 0.18678, - "effective_density": 0.20962, - "route_count_min": 63, - "route_count_mean": 123.67, + "threshold_density": 0.2084, + "effective_density": 0.23181, + "route_count_min": 52, + "route_count_mean": 136.77, "route_count_max": 590, "metadata_capacity": 640 }, @@ -7638,8 +8833,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 915450, - "sha256": "13610b4257d37ebe6d83e9726d0194712b588c044c49bce154a17e68b6a7ddc8", + "bytes": 926954, + "sha256": "6522c05d8ea33aebbb4be2cdea7ee5e122061509a6de1d5ab3e72fe12249e360", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7656,16 +8851,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 405.47777480073273, + "cold_load_s": 439.58678179606795, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7674,13 +8870,13 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 47.894321171566844, + "warmup_s": 14.64438444795087, "repeats_s": [ - 7.4758722125552595, - 7.466093507129699, - 7.483112619724125 + 6.551806676667184, + 6.531784829683602, + 6.554025718010962 ], - "median_s": 7.4758722125552595, + "median_s": 6.551806676667184, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 2304, @@ -7716,10 +8912,10 @@ "route_density": { "blocks": 1147, "sink_blocks": 13, - "threshold_density": 0.18758, - "effective_density": 0.2084, - "route_count_min": 141, - "route_count_mean": 239.03, + "threshold_density": 0.18735, + "effective_density": 0.20817, + "route_count_min": 144, + "route_count_mean": 238.77, "route_count_max": 1147, "metadata_capacity": 1152 }, @@ -7739,8 +8935,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 1786747, - "sha256": "b02db4baaf08434ef61616481171b10bb916eeab1db2291b763309091e2be215", + "bytes": 1825597, + "sha256": "04260482aa6b6d7ec1550321d6dc7e0f04d7fa8cce53714711313f84ac47fc84", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7757,16 +8953,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 405.47777480073273, + "cold_load_s": 439.58678179606795, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7775,13 +8972,13 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 266.72677598381415, + "warmup_s": 19.447643855120987, "repeats_s": [ - 13.31669664895162, - 13.355944743845612, - 13.337900619022548 + 12.063323335722089, + 12.064927336759865, + 12.09263978432864 ], - "median_s": 13.337900619022548, + "median_s": 12.064927336759865, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 3456, @@ -7817,10 +9014,10 @@ "route_density": { "blocks": 1705, "sink_blocks": 20, - "threshold_density": 0.18555, - "effective_density": 0.20659, - "route_count_min": 218, - "route_count_mean": 352.24, + "threshold_density": 0.18567, + "effective_density": 0.20671, + "route_count_min": 220, + "route_count_mean": 352.45, "route_count_max": 1705, "metadata_capacity": 1728 }, @@ -7840,8 +9037,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 1909484, - "sha256": "9df91e67eb180ae2d0ba30865c89b5b44fc11df78cf0c83ffe207570f1e5c4e6", + "bytes": 1936513, + "sha256": "9fae8f6915cf26a28dbafda31f976c0cd4456c8b14f0a63ccc8b4010d00ac44b", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -7858,16 +9055,17 @@ 10, 0 ], + "runtime_variant": "Sol-H3-int64", "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, "attention_backend": "sol_bsa", - "compute_quant": "none", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 405.47777480073273, + "cold_load_s": 439.58678179606795, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -7876,13 +9074,13 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 143.39529226301238, + "warmup_s": 35.399448071140796, "repeats_s": [ - 3.241004516836256, - 3.2079915329813957, - 3.2244293242692947 + 2.8105423622764647, + 2.755053977947682, + 2.758813951164484 ], - "median_s": 3.2244293242692947, + "median_s": 2.758813951164484, "observed_sparse_calls": 288, "sparse_stats": { "sparse_calls": 1152, @@ -7918,10 +9116,10 @@ "route_density": { "blocks": 590, "sink_blocks": 7, - "threshold_density": 0.18679, - "effective_density": 0.20964, - "route_count_min": 61, - "route_count_mean": 123.69, + "threshold_density": 0.18678, + "effective_density": 0.20962, + "route_count_min": 63, + "route_count_mean": 123.67, "route_count_max": 590, "metadata_capacity": 640 }, @@ -7941,19 +9139,19 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 896271, - "sha256": "d152b689622c1ad2239157c2dbecafb11819034e5ce227a9715e229a2ce44233", + "bytes": 915450, + "sha256": "13610b4257d37ebe6d83e9726d0194712b588c044c49bce154a17e68b6a7ddc8", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "lightx2v-4", - "adapter_sha256": "c3d4a2cf618efea71b9e21a4baaa12d412f1eb6c2b6f86efacaf0ebb6814b689", - "nfe": 4, + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, "cluster": "hsg", - "gpus": 4, + "gpus": 8, "allocated_gpus_per_node": "4", - "nodes": 1, + "nodes": 2, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -7964,97 +9162,97 @@ "cudnn": 91501, "attention_backend": "sol_bsa", "compute_quant": "none", - "video_shift": 6.0, + "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 8, - "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", - "cold_load_s": 382.99306421587244, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 405.47777480073273, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, "fps": 24, "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", - "duration": 5, - "frames": 124, + "duration": 10, + "frames": 243, "seed": 20260903, - "warmup_s": 27.033580739051104, + "warmup_s": 47.894321171566844, "repeats_s": [ - 2.964362938888371, - 2.9692238410934806, - 2.9639265229925513 + 7.4758722125552595, + 7.466093507129699, + 7.483112619724125 ], - "median_s": 2.964362938888371, - "observed_sparse_calls": 144, + "median_s": 7.4758722125552595, + "observed_sparse_calls": 288, "sparse_stats": { - "sparse_calls": 576, - "dense_calls": 224, - "sparse_calls_measured_request": 144, + "sparse_calls": 2304, + "dense_calls": 896, + "sparse_calls_measured_request": 288, "sparse_fraction": 0.72, - "requests": 4, - "last_step": 3, - "video_start": 427, + "requests": 8, + "last_step": 7, + "video_start": 823, "sink_mode": "prefix", "sink_start": 0, - "sink_tokens": 427, + "sink_tokens": 823, "sink_blocks": null, - "sequence_length": 37723, + "sequence_length": 73399, "tau": 1.0, "thresh_type": "diag", "backend": "sol_bsa", "qkv_wire_dtype": "int8_qkv", "qkv_wire_policy": "int8_qkv_all", - "qkv_int8_calls_measured_request": 200, + "qkv_int8_calls_measured_request": 400, "qkv_bf16_calls_measured_request": 0, "qkv_wire_ratio": 0.520833, "output_wire_dtype": "fp8", "output_int8_calls_measured_request": 0, - "output_fp8_calls_measured_request": 200, + "output_fp8_calls_measured_request": 400, "output_bf16_calls_measured_request": 0, "output_wire_ratio": 0.5, "attention_comm_ratio": 0.515625, "packed_input": true, "compile_bucket_size": 4096, - "dense_steps": 1, + "dense_steps": 2, "dense_layers": 2, "route_density": { - "blocks": 590, - "sink_blocks": 7, - "threshold_density": 0.20726, - "effective_density": 0.23063, - "route_count_min": 54, - "route_count_mean": 136.07, - "route_count_max": 590, - "metadata_capacity": 640 + "blocks": 1147, + "sink_blocks": 13, + "threshold_density": 0.18758, + "effective_density": 0.2084, + "route_count_min": 141, + "route_count_mean": 239.03, + "route_count_max": 1147, + "metadata_capacity": 1152 }, "gate": null, "declined": { - "warmup_step": 200, - "dense_layer": 24 + "warmup_step": 800, + "dense_layer": 96 } }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 243, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 655912, - "sha256": "419a740c383f13a3d0b9ef617ab32f710a59492f4b787779833c9634b260d612", + "duration_s": 10.208344, + "bytes": 1786747, + "sha256": "b02db4baaf08434ef61616481171b10bb916eeab1db2291b763309091e2be215", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "lightx2v-8", - "adapter_sha256": "9b0efe3613b43a84e30febaa43af27432ea9d0711eac7bba904b2556b175f6d4", + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "hsg", - "gpus": 4, + "gpus": 8, "allocated_gpus_per_node": "4", - "nodes": 1, + "nodes": 2, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8065,35 +9263,338 @@ "cudnn": 91501, "attention_backend": "sol_bsa", "compute_quant": "none", - "video_shift": 6.0, + "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 8, - "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", - "cold_load_s": 391.5484301429242, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 405.47777480073273, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, "fps": 24, "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", - "duration": 5, - "frames": 124, + "duration": 15, + "frames": 362, "seed": 20260903, - "warmup_s": 30.07582237990573, + "warmup_s": 266.72677598381415, "repeats_s": [ - 5.652423293795437, - 5.53552816901356, - 5.481880785897374 + 13.31669664895162, + 13.355944743845612, + 13.337900619022548 ], - "median_s": 5.53552816901356, - "observed_sparse_calls": 336, + "median_s": 13.337900619022548, + "observed_sparse_calls": 288, "sparse_stats": { - "sparse_calls": 1344, - "dense_calls": 256, - "sparse_calls_measured_request": 336, - "sparse_fraction": 0.84, - "requests": 4, + "sparse_calls": 3456, + "dense_calls": 1344, + "sparse_calls_measured_request": 288, + "sparse_fraction": 0.72, + "requests": 12, "last_step": 7, - "video_start": 427, + "video_start": 1219, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 1219, + "sink_blocks": null, + "sequence_length": 109075, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 400, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 400, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 2, + "dense_layers": 2, + "route_density": { + "blocks": 1705, + "sink_blocks": 20, + "threshold_density": 0.18555, + "effective_density": 0.20659, + "route_count_min": 218, + "route_count_mean": 352.24, + "route_count_max": 1705, + "metadata_capacity": 1728 + }, + "gate": null, + "declined": { + "warmup_step": 1200, + "dense_layer": 144 + } + }, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 362, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 15.166666, + "bytes": 1909484, + "sha256": "9df91e67eb180ae2d0ba30865c89b5b44fc11df78cf0c83ffe207570f1e5c4e6", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, + "cluster": "hsg", + "gpus": 8, + "allocated_gpus_per_node": "4", + "nodes": 2, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 405.47777480073273, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 143.39529226301238, + "repeats_s": [ + 3.241004516836256, + 3.2079915329813957, + 3.2244293242692947 + ], + "median_s": 3.2244293242692947, + "observed_sparse_calls": 288, + "sparse_stats": { + "sparse_calls": 1152, + "dense_calls": 448, + "sparse_calls_measured_request": 288, + "sparse_fraction": 0.72, + "requests": 4, + "last_step": 7, + "video_start": 427, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 427, + "sink_blocks": null, + "sequence_length": 37723, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 400, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 400, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 2, + "dense_layers": 2, + "route_density": { + "blocks": 590, + "sink_blocks": 7, + "threshold_density": 0.18679, + "effective_density": 0.20964, + "route_count_min": 61, + "route_count_mean": 123.69, + "route_count_max": 590, + "metadata_capacity": 640 + }, + "gate": null, + "declined": { + "warmup_step": 400, + "dense_layer": 48 + } + }, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 896271, + "sha256": "d152b689622c1ad2239157c2dbecafb11819034e5ce227a9715e229a2ce44233", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "lightx2v-4", + "adapter_sha256": "c3d4a2cf618efea71b9e21a4baaa12d412f1eb6c2b6f86efacaf0ebb6814b689", + "nfe": 4, + "cluster": "hsg", + "gpus": 4, + "allocated_gpus_per_node": "4", + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 6.0, + "audio_shift": 3.0, + "adapter_alpha": 8, + "adapter_filename": "minimax_h3_fl2v_turbo_4step_v1.2_768p_bf16.safetensors", + "cold_load_s": 382.99306421587244, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 27.033580739051104, + "repeats_s": [ + 2.964362938888371, + 2.9692238410934806, + 2.9639265229925513 + ], + "median_s": 2.964362938888371, + "observed_sparse_calls": 144, + "sparse_stats": { + "sparse_calls": 576, + "dense_calls": 224, + "sparse_calls_measured_request": 144, + "sparse_fraction": 0.72, + "requests": 4, + "last_step": 3, + "video_start": 427, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 427, + "sink_blocks": null, + "sequence_length": 37723, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 200, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 200, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 1, + "dense_layers": 2, + "route_density": { + "blocks": 590, + "sink_blocks": 7, + "threshold_density": 0.20726, + "effective_density": 0.23063, + "route_count_min": 54, + "route_count_mean": 136.07, + "route_count_max": 590, + "metadata_capacity": 640 + }, + "gate": null, + "declined": { + "warmup_step": 200, + "dense_layer": 24 + } + }, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 655912, + "sha256": "419a740c383f13a3d0b9ef617ab32f710a59492f4b787779833c9634b260d612", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "lightx2v-8", + "adapter_sha256": "9b0efe3613b43a84e30febaa43af27432ea9d0711eac7bba904b2556b175f6d4", + "nfe": 8, + "cluster": "hsg", + "gpus": 4, + "allocated_gpus_per_node": "4", + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 6.0, + "audio_shift": 3.0, + "adapter_alpha": 8, + "adapter_filename": "minimax_h3_fl2v_turbo_8step_v1.0_768p_bf16.safetensors", + "cold_load_s": 391.5484301429242, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 30.07582237990573, + "repeats_s": [ + 5.652423293795437, + 5.53552816901356, + 5.481880785897374 + ], + "median_s": 5.53552816901356, + "observed_sparse_calls": 336, + "sparse_stats": { + "sparse_calls": 1344, + "dense_calls": 256, + "sparse_calls_measured_request": 336, + "sparse_fraction": 0.84, + "requests": 4, + "last_step": 7, + "video_start": 427, "sink_mode": "prefix", "sink_start": 0, "sink_tokens": 427, @@ -8136,15 +9637,70 @@ "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 124, + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 772102, + "sha256": "39eceb078f98bfc967f27262f224b2b60e83b0cdb2539f48463debf049e11dda", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "fasth3-4", + "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", + "nfe": 4, + "cluster": "poly", + "gpus": 1, + "allocated_gpus_per_node": null, + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "dense", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 64, + "adapter_filename": "adapter_model.safetensors", + "cold_load_s": 286.39243665800313, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 10, + "frames": 243, + "seed": 20260903, + "warmup_s": 37.879758728988236, + "repeats_s": [ + 37.825742068002, + 37.85681245397427, + 37.924753985978896 + ], + "median_s": 37.85681245397427, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 243, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 5.258344, - "bytes": 772102, - "sha256": "39eceb078f98bfc967f27262f224b2b60e83b0cdb2539f48463debf049e11dda", + "duration_s": 10.208344, + "bytes": 3532418, + "sha256": "fb524c8a5ffc3a4d63485ff7b9e58b5c9ce5fd52d39f449be410df377d7fec9a", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8177,29 +9733,29 @@ "height": 768, "fps": 24, "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", - "duration": 10, - "frames": 243, + "duration": 15, + "frames": 362, "seed": 20260903, - "warmup_s": 37.879758728988236, + "warmup_s": 71.84661786601646, "repeats_s": [ - 37.825742068002, - 37.85681245397427, - 37.924753985978896 + 71.74971473100595, + 71.8363306170213, + 71.73549409999396 ], - "median_s": 37.85681245397427, + "median_s": 71.74971473100595, "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 243, + "frames": 362, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 10.208344, - "bytes": 3532418, - "sha256": "fb524c8a5ffc3a4d63485ff7b9e58b5c9ce5fd52d39f449be410df377d7fec9a", + "duration_s": 15.166666, + "bytes": 5503515, + "sha256": "a6a09f60e54a1029746dbb329c0a2598e69891ce1e36f56de59b099f9c1be00b", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8232,17 +9788,221 @@ "height": 768, "fps": 24, "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 37.64529103299719, + "repeats_s": [ + 13.841052805015352, + 13.813988408015575, + 13.851194803020917 + ], + "median_s": 13.841052805015352, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 1439546, + "sha256": "0fd2c6240936b378cd40976fb5c34d09e2bdb1653324b45a87d57e5315127bc9", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "fasth3-4", + "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", + "nfe": 4, + "cluster": "poly", + "gpus": 4, + "allocated_gpus_per_node": null, + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 64, + "adapter_filename": "adapter_model.safetensors", + "cold_load_s": 294.7777604120056, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 10, + "frames": 243, + "seed": 20260903, + "warmup_s": 14.31319376299507, + "repeats_s": [ + 7.229829151008744, + 7.21938319200126, + 7.224561576003907 + ], + "median_s": 7.224561576003907, + "observed_sparse_calls": 144, + "sparse_stats": { + "sparse_calls": 1152, + "dense_calls": 448, + "sparse_calls_measured_request": 144, + "sparse_fraction": 0.72, + "requests": 8, + "last_step": 3, + "video_start": 823, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 823, + "sink_blocks": null, + "sequence_length": 73399, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 200, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 200, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 1, + "dense_layers": 2, + "route_density": { + "blocks": 1147, + "sink_blocks": 13, + "threshold_density": 0.20161, + "effective_density": 0.22213, + "route_count_min": 115, + "route_count_mean": 254.78, + "route_count_max": 1147, + "metadata_capacity": 1152 + }, + "gate": null, + "declined": { + "warmup_step": 400, + "dense_layer": 48 + } + }, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 243, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 10.208344, + "bytes": 3665198, + "sha256": "cc8777567182eac7e0962fce854141f29518479974f0b71ecd541e55989f8985", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "fasth3-4", + "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", + "nfe": 4, + "cluster": "poly", + "gpus": 4, + "allocated_gpus_per_node": null, + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 64, + "adapter_filename": "adapter_model.safetensors", + "cold_load_s": 294.7777604120056, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 71.84661786601646, + "warmup_s": 55.12173893299769, "repeats_s": [ - 71.74971473100595, - 71.8363306170213, - 71.73549409999396 + 12.985925676999614, + 12.991896356994403, + 12.980969693002407 ], - "median_s": 71.74971473100595, - "observed_sparse_calls": 0, + "median_s": 12.985925676999614, + "observed_sparse_calls": 144, + "sparse_stats": { + "sparse_calls": 1728, + "dense_calls": 672, + "sparse_calls_measured_request": 144, + "sparse_fraction": 0.72, + "requests": 12, + "last_step": 3, + "video_start": 1219, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 1219, + "sink_blocks": null, + "sequence_length": 109075, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 200, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 200, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 1, + "dense_layers": 2, + "route_density": { + "blocks": 1705, + "sink_blocks": 20, + "threshold_density": 0.20161, + "effective_density": 0.22174, + "route_count_min": 160, + "route_count_mean": 378.07, + "route_count_max": 1705, + "metadata_capacity": 1728 + }, + "gate": null, + "declined": { + "warmup_step": 600, + "dense_layer": 72 + } + }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8253,8 +10013,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 5503515, - "sha256": "a6a09f60e54a1029746dbb329c0a2598e69891ce1e36f56de59b099f9c1be00b", + "bytes": 5528454, + "sha256": "61dad109e536f05e0d29c26266e12391d76ca29a3d5c02b2749caf0bdf74de32", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8263,7 +10023,7 @@ "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", "nfe": 4, "cluster": "poly", - "gpus": 1, + "gpus": 4, "allocated_gpus_per_node": null, "nodes": 1, "gpu": "NVIDIA GB200", @@ -8275,13 +10035,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "dense", + "attention_backend": "sol_bsa", "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 286.39243665800313, + "cold_load_s": 294.7777604120056, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8290,14 +10050,61 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 37.64529103299719, + "warmup_s": 29.798163390005357, "repeats_s": [ - 13.841052805015352, - 13.813988408015575, - 13.851194803020917 + 3.0818069269880652, + 3.0768804979888955, + 3.075204674998531 ], - "median_s": 13.841052805015352, - "observed_sparse_calls": 0, + "median_s": 3.0768804979888955, + "observed_sparse_calls": 144, + "sparse_stats": { + "sparse_calls": 576, + "dense_calls": 224, + "sparse_calls_measured_request": 144, + "sparse_fraction": 0.72, + "requests": 4, + "last_step": 3, + "video_start": 427, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 427, + "sink_blocks": null, + "sequence_length": 37723, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 200, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 200, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 1, + "dense_layers": 2, + "route_density": { + "blocks": 590, + "sink_blocks": 7, + "threshold_density": 0.20582, + "effective_density": 0.22908, + "route_count_min": 60, + "route_count_mean": 135.16, + "route_count_max": 590, + "metadata_capacity": 640 + }, + "gate": null, + "declined": { + "warmup_step": 200, + "dense_layer": 24 + } + }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8308,8 +10115,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 1439546, - "sha256": "0fd2c6240936b378cd40976fb5c34d09e2bdb1653324b45a87d57e5315127bc9", + "bytes": 1413144, + "sha256": "1417f5af3c2c2b283d6361b2b58db2d3f8ccb7ccb6bf36f6856d832d898867d3", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8318,9 +10125,9 @@ "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", "nfe": 4, "cluster": "poly", - "gpus": 4, + "gpus": 8, "allocated_gpus_per_node": null, - "nodes": 1, + "nodes": 2, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8336,7 +10143,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 294.7777604120056, + "cold_load_s": 285.7827172730031, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8345,13 +10152,13 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 14.31319376299507, + "warmup_s": 10.177076838008361, "repeats_s": [ - 7.229829151008744, - 7.21938319200126, - 7.224561576003907 + 3.789100617010263, + 3.7869828450056957, + 3.783537929004524 ], - "median_s": 7.224561576003907, + "median_s": 3.7869828450056957, "observed_sparse_calls": 144, "sparse_stats": { "sparse_calls": 1152, @@ -8387,10 +10194,10 @@ "route_density": { "blocks": 1147, "sink_blocks": 13, - "threshold_density": 0.20161, - "effective_density": 0.22213, - "route_count_min": 115, - "route_count_mean": 254.78, + "threshold_density": 0.18575, + "effective_density": 0.20664, + "route_count_min": 143, + "route_count_mean": 237.01, "route_count_max": 1147, "metadata_capacity": 1152 }, @@ -8410,8 +10217,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 3665198, - "sha256": "cc8777567182eac7e0962fce854141f29518479974f0b71ecd541e55989f8985", + "bytes": 3510338, + "sha256": "94d23d2ff334d76b883502f2ee5190f86a470301b7e3de530d67fab0c0d59b8d", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8420,9 +10227,9 @@ "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", "nfe": 4, "cluster": "poly", - "gpus": 4, + "gpus": 8, "allocated_gpus_per_node": null, - "nodes": 1, + "nodes": 2, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8438,7 +10245,7 @@ "audio_shift": 3.0, "adapter_alpha": 64, "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 294.7777604120056, + "cold_load_s": 285.7827172730031, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8447,13 +10254,13 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 55.12173893299769, + "warmup_s": 12.688988487992901, "repeats_s": [ - 12.985925676999614, - 12.991896356994403, - 12.980969693002407 + 6.722846845994354, + 6.70660141800181, + 6.693554066994693 ], - "median_s": 12.985925676999614, + "median_s": 6.70660141800181, "observed_sparse_calls": 144, "sparse_stats": { "sparse_calls": 1728, @@ -8489,10 +10296,10 @@ "route_density": { "blocks": 1705, "sink_blocks": 20, - "threshold_density": 0.20161, - "effective_density": 0.22174, - "route_count_min": 160, - "route_count_mean": 378.07, + "threshold_density": 0.18448, + "effective_density": 0.20552, + "route_count_min": 220, + "route_count_mean": 350.42, "route_count_max": 1705, "metadata_capacity": 1728 }, @@ -8505,24 +10312,181 @@ "media_validation": { "video_codec": "h264", "audio_codec": "aac", - "frames": 362, + "frames": 362, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 15.166666, + "bytes": 5590962, + "sha256": "88c646c9d73ccae4f10b8dc0421df69b99d7c56825ec6249d5455485e85c2d02", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "fasth3-4", + "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", + "nfe": 4, + "cluster": "poly", + "gpus": 8, + "allocated_gpus_per_node": null, + "nodes": 2, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "sol_bsa", + "compute_quant": "none", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 64, + "adapter_filename": "adapter_model.safetensors", + "cold_load_s": 285.7827172730031, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 5, + "frames": 124, + "seed": 20260903, + "warmup_s": 32.965851381988614, + "repeats_s": [ + 1.632835114010959, + 1.6313898380030878, + 1.6286788399884244 + ], + "median_s": 1.6313898380030878, + "observed_sparse_calls": 144, + "sparse_stats": { + "sparse_calls": 576, + "dense_calls": 224, + "sparse_calls_measured_request": 144, + "sparse_fraction": 0.72, + "requests": 4, + "last_step": 3, + "video_start": 427, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 427, + "sink_blocks": null, + "sequence_length": 37723, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 200, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 200, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 1, + "dense_layers": 2, + "route_density": { + "blocks": 590, + "sink_blocks": 7, + "threshold_density": 0.18663, + "effective_density": 0.20946, + "route_count_min": 66, + "route_count_mean": 123.58, + "route_count_max": 590, + "metadata_capacity": 640 + }, + "gate": null, + "declined": { + "warmup_step": 200, + "dense_layer": 24 + } + }, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 124, + "width": 1344, + "height": 768, + "fps": 24, + "sample_rate": 32000, + "channels": 2, + "duration_s": 5.258344, + "bytes": 1514911, + "sha256": "9d72b5fc5ee6b831d9c7a47827c19dd5b59d3ca554c7b325698f079ea1877393", + "verification": "ffprobe frame count and complete audio/video decode" + } + }, + { + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, + "cluster": "poly", + "gpus": 1, + "allocated_gpus_per_node": null, + "nodes": 1, + "gpu": "NVIDIA GB200", + "capability": [ + 10, + 0 + ], + "runtime_variant": "Sol-H3", + "torch": "2.10.0+cu130", + "cuda": "13.0", + "cudnn": 91501, + "attention_backend": "dense", + "compute_quant": "mxfp8", + "video_shift": 12.0, + "audio_shift": 3.0, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 294.1646487209946, + "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", + "width": 1344, + "height": 768, + "fps": 24, + "timing": "warm pipeline: text encoding + DiT + video/audio decoding; excludes load, warmup and MP4 encoding", + "duration": 10, + "frames": 243, + "seed": 20260903, + "warmup_s": 63.84408454998629, + "repeats_s": [ + 63.75714229498408, + 63.8483764519915, + 63.88921740002115 + ], + "median_s": 63.8483764519915, + "observed_sparse_calls": 0, + "media_validation": { + "video_codec": "h264", + "audio_codec": "aac", + "frames": 243, "width": 1344, "height": 768, "fps": 24, "sample_rate": 32000, "channels": 2, - "duration_s": 15.166666, - "bytes": 5528454, - "sha256": "61dad109e536f05e0d29c26266e12391d76ca29a3d5c02b2749caf0bdf74de32", + "duration_s": 10.208344, + "bytes": 1574774, + "sha256": "da06db16d82e047ffb3fe684bb0ba0afee4be6b4691beac518e121e4f8fa413e", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "fasth3-4", - "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", - "nfe": 4, + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, "cluster": "poly", - "gpus": 4, + "gpus": 1, "allocated_gpus_per_node": null, "nodes": 1, "gpu": "NVIDIA GB200", @@ -8534,13 +10498,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "sol_bsa", - "compute_quant": "none", + "attention_backend": "dense", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 64, - "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 294.7777604120056, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 294.1646487209946, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8549,61 +10513,14 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 29.798163390005357, + "warmup_s": 30.864501964009833, "repeats_s": [ - 3.0818069269880652, - 3.0768804979888955, - 3.075204674998531 + 21.856848203984555, + 21.847355916979723, + 21.877207930985605 ], - "median_s": 3.0768804979888955, - "observed_sparse_calls": 144, - "sparse_stats": { - "sparse_calls": 576, - "dense_calls": 224, - "sparse_calls_measured_request": 144, - "sparse_fraction": 0.72, - "requests": 4, - "last_step": 3, - "video_start": 427, - "sink_mode": "prefix", - "sink_start": 0, - "sink_tokens": 427, - "sink_blocks": null, - "sequence_length": 37723, - "tau": 1.0, - "thresh_type": "diag", - "backend": "sol_bsa", - "qkv_wire_dtype": "int8_qkv", - "qkv_wire_policy": "int8_qkv_all", - "qkv_int8_calls_measured_request": 200, - "qkv_bf16_calls_measured_request": 0, - "qkv_wire_ratio": 0.520833, - "output_wire_dtype": "fp8", - "output_int8_calls_measured_request": 0, - "output_fp8_calls_measured_request": 200, - "output_bf16_calls_measured_request": 0, - "output_wire_ratio": 0.5, - "attention_comm_ratio": 0.515625, - "packed_input": true, - "compile_bucket_size": 4096, - "dense_steps": 1, - "dense_layers": 2, - "route_density": { - "blocks": 590, - "sink_blocks": 7, - "threshold_density": 0.20582, - "effective_density": 0.22908, - "route_count_min": 60, - "route_count_mean": 135.16, - "route_count_max": 590, - "metadata_capacity": 640 - }, - "gate": null, - "declined": { - "warmup_step": 200, - "dense_layer": 24 - } - }, + "median_s": 21.856848203984555, + "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8614,19 +10531,19 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 1413144, - "sha256": "1417f5af3c2c2b283d6361b2b58db2d3f8ccb7ccb6bf36f6856d832d898867d3", + "bytes": 832822, + "sha256": "c4f648d03ccbdab9a6b31b4086c3cb02ae5d6c5fc7ebb9b123549aba8728647b", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "fasth3-4", - "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", - "nfe": 4, + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, "cluster": "poly", - "gpus": 8, + "gpus": 1, "allocated_gpus_per_node": null, - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8636,13 +10553,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "sol_bsa", + "attention_backend": "dense", "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 64, - "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 285.7827172730031, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 290.6031925349962, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8651,61 +10568,14 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 10.177076838008361, + "warmup_s": 70.50726453700918, "repeats_s": [ - 3.789100617010263, - 3.7869828450056957, - 3.783537929004524 + 70.32355213700794, + 70.31223342899466, + 70.11613318399759 ], - "median_s": 3.7869828450056957, - "observed_sparse_calls": 144, - "sparse_stats": { - "sparse_calls": 1152, - "dense_calls": 448, - "sparse_calls_measured_request": 144, - "sparse_fraction": 0.72, - "requests": 8, - "last_step": 3, - "video_start": 823, - "sink_mode": "prefix", - "sink_start": 0, - "sink_tokens": 823, - "sink_blocks": null, - "sequence_length": 73399, - "tau": 1.0, - "thresh_type": "diag", - "backend": "sol_bsa", - "qkv_wire_dtype": "int8_qkv", - "qkv_wire_policy": "int8_qkv_all", - "qkv_int8_calls_measured_request": 200, - "qkv_bf16_calls_measured_request": 0, - "qkv_wire_ratio": 0.520833, - "output_wire_dtype": "fp8", - "output_int8_calls_measured_request": 0, - "output_fp8_calls_measured_request": 200, - "output_bf16_calls_measured_request": 0, - "output_wire_ratio": 0.5, - "attention_comm_ratio": 0.515625, - "packed_input": true, - "compile_bucket_size": 4096, - "dense_steps": 1, - "dense_layers": 2, - "route_density": { - "blocks": 1147, - "sink_blocks": 13, - "threshold_density": 0.18575, - "effective_density": 0.20664, - "route_count_min": 143, - "route_count_mean": 237.01, - "route_count_max": 1147, - "metadata_capacity": 1152 - }, - "gate": null, - "declined": { - "warmup_step": 400, - "dense_layer": 48 - } - }, + "median_s": 70.31223342899466, + "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8716,19 +10586,19 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 3510338, - "sha256": "94d23d2ff334d76b883502f2ee5190f86a470301b7e3de530d67fab0c0d59b8d", + "bytes": 1635071, + "sha256": "32b29a029110098314abfe342730a5ad3c2501ad6b8f7d8d1c05be15351fa6ab", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "fasth3-4", - "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", - "nfe": 4, + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, "cluster": "poly", - "gpus": 8, + "gpus": 1, "allocated_gpus_per_node": null, - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8738,13 +10608,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "sol_bsa", + "attention_backend": "dense", "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 64, - "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 285.7827172730031, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 290.6031925349962, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8753,61 +10623,14 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 12.688988487992901, + "warmup_s": 135.63105801402708, "repeats_s": [ - 6.722846845994354, - 6.70660141800181, - 6.693554066994693 + 135.65047668898478, + 135.54529769800138, + 135.5381122569961 ], - "median_s": 6.70660141800181, - "observed_sparse_calls": 144, - "sparse_stats": { - "sparse_calls": 1728, - "dense_calls": 672, - "sparse_calls_measured_request": 144, - "sparse_fraction": 0.72, - "requests": 12, - "last_step": 3, - "video_start": 1219, - "sink_mode": "prefix", - "sink_start": 0, - "sink_tokens": 1219, - "sink_blocks": null, - "sequence_length": 109075, - "tau": 1.0, - "thresh_type": "diag", - "backend": "sol_bsa", - "qkv_wire_dtype": "int8_qkv", - "qkv_wire_policy": "int8_qkv_all", - "qkv_int8_calls_measured_request": 200, - "qkv_bf16_calls_measured_request": 0, - "qkv_wire_ratio": 0.520833, - "output_wire_dtype": "fp8", - "output_int8_calls_measured_request": 0, - "output_fp8_calls_measured_request": 200, - "output_bf16_calls_measured_request": 0, - "output_wire_ratio": 0.5, - "attention_comm_ratio": 0.515625, - "packed_input": true, - "compile_bucket_size": 4096, - "dense_steps": 1, - "dense_layers": 2, - "route_density": { - "blocks": 1705, - "sink_blocks": 20, - "threshold_density": 0.18448, - "effective_density": 0.20552, - "route_count_min": 220, - "route_count_mean": 350.42, - "route_count_max": 1705, - "metadata_capacity": 1728 - }, - "gate": null, - "declined": { - "warmup_step": 600, - "dense_layer": 72 - } - }, + "median_s": 135.54529769800138, + "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8818,19 +10641,19 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 5590962, - "sha256": "88c646c9d73ccae4f10b8dc0421df69b99d7c56825ec6249d5455485e85c2d02", + "bytes": 1889995, + "sha256": "c8687863f2ed2460df95d6eb44c5534a5a57ae4de8cf3fdfbe0c36afb2784b4b", "verification": "ffprobe frame count and complete audio/video decode" } }, { - "adapter": "fasth3-4", - "adapter_sha256": "4ce198c83132251b7fd0de2503823aa49c53983f068318f66cb19eaefb7fcc12", - "nfe": 4, + "adapter": "hyperflow-8", + "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", + "nfe": 8, "cluster": "poly", - "gpus": 8, + "gpus": 1, "allocated_gpus_per_node": null, - "nodes": 2, + "nodes": 1, "gpu": "NVIDIA GB200", "capability": [ 10, @@ -8840,13 +10663,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "sol_bsa", + "attention_backend": "dense", "compute_quant": "none", "video_shift": 12.0, "audio_shift": 3.0, - "adapter_alpha": 64, - "adapter_filename": "adapter_model.safetensors", - "cold_load_s": 285.7827172730031, + "adapter_alpha": 256.0, + "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", + "cold_load_s": 290.6031925349962, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8855,61 +10678,14 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 32.965851381988614, + "warmup_s": 71.83553511201171, "repeats_s": [ - 1.632835114010959, - 1.6313898380030878, - 1.6286788399884244 + 24.93977749699843, + 24.9366850979859, + 24.955815777997486 ], - "median_s": 1.6313898380030878, - "observed_sparse_calls": 144, - "sparse_stats": { - "sparse_calls": 576, - "dense_calls": 224, - "sparse_calls_measured_request": 144, - "sparse_fraction": 0.72, - "requests": 4, - "last_step": 3, - "video_start": 427, - "sink_mode": "prefix", - "sink_start": 0, - "sink_tokens": 427, - "sink_blocks": null, - "sequence_length": 37723, - "tau": 1.0, - "thresh_type": "diag", - "backend": "sol_bsa", - "qkv_wire_dtype": "int8_qkv", - "qkv_wire_policy": "int8_qkv_all", - "qkv_int8_calls_measured_request": 200, - "qkv_bf16_calls_measured_request": 0, - "qkv_wire_ratio": 0.520833, - "output_wire_dtype": "fp8", - "output_int8_calls_measured_request": 0, - "output_fp8_calls_measured_request": 200, - "output_bf16_calls_measured_request": 0, - "output_wire_ratio": 0.5, - "attention_comm_ratio": 0.515625, - "packed_input": true, - "compile_bucket_size": 4096, - "dense_steps": 1, - "dense_layers": 2, - "route_density": { - "blocks": 590, - "sink_blocks": 7, - "threshold_density": 0.18663, - "effective_density": 0.20946, - "route_count_min": 66, - "route_count_mean": 123.58, - "route_count_max": 590, - "metadata_capacity": 640 - }, - "gate": null, - "declined": { - "warmup_step": 200, - "dense_layer": 24 - } - }, + "median_s": 24.93977749699843, + "observed_sparse_calls": 0, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8920,8 +10696,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 1514911, - "sha256": "9d72b5fc5ee6b831d9c7a47827c19dd5b59d3ca554c7b325698f079ea1877393", + "bytes": 932090, + "sha256": "88d411f81708016df471e2f0d46aa8a46968b9dec59c431649031cd3e12fef27", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8930,7 +10706,7 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "poly", - "gpus": 1, + "gpus": 4, "allocated_gpus_per_node": null, "nodes": 1, "gpu": "NVIDIA GB200", @@ -8942,13 +10718,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "dense", - "compute_quant": "none", + "attention_backend": "sol_bsa", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 290.6031925349962, + "cold_load_s": 288.39288462500554, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -8957,14 +10733,61 @@ "duration": 10, "frames": 243, "seed": 20260903, - "warmup_s": 70.50726453700918, + "warmup_s": 18.310874960006913, "repeats_s": [ - 70.32355213700794, - 70.31223342899466, - 70.11613318399759 + 11.83686811098596, + 11.855843441007892, + 11.825516913988395 ], - "median_s": 70.31223342899466, - "observed_sparse_calls": 0, + "median_s": 11.83686811098596, + "observed_sparse_calls": 288, + "sparse_stats": { + "sparse_calls": 2304, + "dense_calls": 896, + "sparse_calls_measured_request": 288, + "sparse_fraction": 0.72, + "requests": 8, + "last_step": 7, + "video_start": 823, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 823, + "sink_blocks": null, + "sequence_length": 73399, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 400, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 400, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 2, + "dense_layers": 2, + "route_density": { + "blocks": 1147, + "sink_blocks": 13, + "threshold_density": 0.20331, + "effective_density": 0.22398, + "route_count_min": 97, + "route_count_mean": 256.9, + "route_count_max": 1147, + "metadata_capacity": 1152 + }, + "gate": null, + "declined": { + "warmup_step": 800, + "dense_layer": 96 + } + }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -8975,8 +10798,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 10.208344, - "bytes": 1635071, - "sha256": "32b29a029110098314abfe342730a5ad3c2501ad6b8f7d8d1c05be15351fa6ab", + "bytes": 1807659, + "sha256": "dae00a32244bf582679e80092cc171054e9ee4df1f8a235f892836c63f715979", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -8985,7 +10808,7 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "poly", - "gpus": 1, + "gpus": 4, "allocated_gpus_per_node": null, "nodes": 1, "gpu": "NVIDIA GB200", @@ -8997,13 +10820,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "dense", - "compute_quant": "none", + "attention_backend": "sol_bsa", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 290.6031925349962, + "cold_load_s": 288.39288462500554, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -9012,14 +10835,61 @@ "duration": 15, "frames": 362, "seed": 20260903, - "warmup_s": 135.63105801402708, + "warmup_s": 28.251016705005895, "repeats_s": [ - 135.65047668898478, - 135.54529769800138, - 135.5381122569961 + 22.0627290699922, + 22.035492566996254, + 22.032053962000646 ], - "median_s": 135.54529769800138, - "observed_sparse_calls": 0, + "median_s": 22.035492566996254, + "observed_sparse_calls": 288, + "sparse_stats": { + "sparse_calls": 3456, + "dense_calls": 1344, + "sparse_calls_measured_request": 288, + "sparse_fraction": 0.72, + "requests": 12, + "last_step": 7, + "video_start": 1219, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 1219, + "sink_blocks": null, + "sequence_length": 109075, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 400, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 400, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 2, + "dense_layers": 2, + "route_density": { + "blocks": 1705, + "sink_blocks": 20, + "threshold_density": 0.20332, + "effective_density": 0.22357, + "route_count_min": 158, + "route_count_mean": 381.19, + "route_count_max": 1705, + "metadata_capacity": 1728 + }, + "gate": null, + "declined": { + "warmup_step": 1200, + "dense_layer": 144 + } + }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -9030,8 +10900,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 15.166666, - "bytes": 1889995, - "sha256": "c8687863f2ed2460df95d6eb44c5534a5a57ae4de8cf3fdfbe0c36afb2784b4b", + "bytes": 1934424, + "sha256": "3ca76c8c6af7724a6bb79045521999ae61b82d38a7b6a40574838c334c1c3421", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -9040,7 +10910,7 @@ "adapter_sha256": "4d7dec1363ebcb9fd63117621b65f8bd19fecacf7ba41f38dd098be363d3972d", "nfe": 8, "cluster": "poly", - "gpus": 1, + "gpus": 4, "allocated_gpus_per_node": null, "nodes": 1, "gpu": "NVIDIA GB200", @@ -9052,13 +10922,13 @@ "torch": "2.10.0+cu130", "cuda": "13.0", "cudnn": 91501, - "attention_backend": "dense", - "compute_quant": "none", + "attention_backend": "sol_bsa", + "compute_quant": "mxfp8", "video_shift": 12.0, "audio_shift": 3.0, "adapter_alpha": 256.0, "adapter_filename": "minimax_h3_hyperflow_8step_v1.0.safetensors", - "cold_load_s": 290.6031925349962, + "cold_load_s": 288.39288462500554, "base_model_revision": "83db0c0efe6ef9824e0e194be110346c0a9542ed", "width": 1344, "height": 768, @@ -9067,14 +10937,61 @@ "duration": 5, "frames": 124, "seed": 20260903, - "warmup_s": 71.83553511201171, + "warmup_s": 21.500160736002726, "repeats_s": [ - 24.93977749699843, - 24.9366850979859, - 24.955815777997486 + 4.695578521001153, + 4.63270866998937, + 4.648682238010224 ], - "median_s": 24.93977749699843, - "observed_sparse_calls": 0, + "median_s": 4.648682238010224, + "observed_sparse_calls": 288, + "sparse_stats": { + "sparse_calls": 1152, + "dense_calls": 448, + "sparse_calls_measured_request": 288, + "sparse_fraction": 0.72, + "requests": 4, + "last_step": 7, + "video_start": 427, + "sink_mode": "prefix", + "sink_start": 0, + "sink_tokens": 427, + "sink_blocks": null, + "sequence_length": 37723, + "tau": 1.0, + "thresh_type": "diag", + "backend": "sol_bsa", + "qkv_wire_dtype": "int8_qkv", + "qkv_wire_policy": "int8_qkv_all", + "qkv_int8_calls_measured_request": 400, + "qkv_bf16_calls_measured_request": 0, + "qkv_wire_ratio": 0.520833, + "output_wire_dtype": "fp8", + "output_int8_calls_measured_request": 0, + "output_fp8_calls_measured_request": 400, + "output_bf16_calls_measured_request": 0, + "output_wire_ratio": 0.5, + "attention_comm_ratio": 0.515625, + "packed_input": true, + "compile_bucket_size": 4096, + "dense_steps": 2, + "dense_layers": 2, + "route_density": { + "blocks": 590, + "sink_blocks": 7, + "threshold_density": 0.20832, + "effective_density": 0.23173, + "route_count_min": 51, + "route_count_mean": 136.72, + "route_count_max": 590, + "metadata_capacity": 640 + }, + "gate": null, + "declined": { + "warmup_step": 400, + "dense_layer": 48 + } + }, "media_validation": { "video_codec": "h264", "audio_codec": "aac", @@ -9085,8 +11002,8 @@ "sample_rate": 32000, "channels": 2, "duration_s": 5.258344, - "bytes": 932090, - "sha256": "88d411f81708016df471e2f0d46aa8a46968b9dec59c431649031cd3e12fef27", + "bytes": 901555, + "sha256": "2ef7388b3b04cf6d9e61c0ab38192ac7a43c69651e9a8a4b5c8f7a5284525bb6", "verification": "ffprobe frame count and complete audio/video decode" } }, @@ -10214,7 +12131,7 @@ } ], "source_revision": "03a0d24259341413597b3c6ff0ce0c0164007e8b", - "updated_at": "2026-09-13T07:34:59.563582+00:00", + "updated_at": "2026-09-13T09:37:36.623351+00:00", "gallery_profile": { "cluster": "hsg", "active_gpus": 4, @@ -10228,9 +12145,15 @@ "comparison_note": "HSG FastH3: eight of nine configurations are 0.19–2.90% slower than the published B300 references; one-GPU 15s is 36.86% slower. Hardware and benchmark prompt differ, so this is not exact same-hardware reproduction.", "poly_comparison_note": "Poly FastH3 除单卡 15 秒外，相对原 B300 耗时差异为 -1.28% 至 +5.43%。单卡 15 秒耗时 71.750 秒，比原参照慢 37.29%；两个 GB200 集群都复现了这一差距。", "gallery_revision": { - "version": "v1-5s-raw", - "status": "superseded_pending_15s_regeneration", - "prompt_validation": "not H3 structured; all 20 actual inputs lack H3 section fields", - "replacement_duration_s": 15 + "version": "v2-h3-15s", + "status": "complete", + "duration_s": 15, + "frames": 362, + "prompt_validation": "H3-format rewrite of attributed source; full original text retained" + }, + "media_repository": { + "repo_id": "Lawrence-cj/sol-h3-hyperflow-20260913-media", + "latest_commit": "2d72924012f74288d6ce319ebfe915fa21162237", + "url": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media" } } diff --git a/index.html b/index.html index 76ff7cecc2abd9132dac4dfc2ad3c32b76db60ee..a2be46b2c02b52dbeb20d2dafe274f97f1e3261b 100644 --- a/index.html +++ b/index.html @@ -9,19 +9,26 @@
Sol—H3CHECKPOINT STUDY · SEPTEMBER 2026Sol-H3 代码 ↗
-
MINIMAX H3 / MATCHED PROMPTS

同一个场景，
看看不同 checkpoint。

HyperFlow 8-step 与公开 H3 少步数模型。固定 prompt、seed、分辨率和帧数，保留各 checkpoint 的采样方式。

20 个场景1344 × 768 分辨率124 帧 · 24 FPS读取结果中…
- +
MINIMAX H3 / MATCHED PROMPTS

同一个场景，
看看不同 checkpoint。

HyperFlow 8-step 与公开 H3 少步数模型。固定 prompt、seed、分辨率和帧数，保留各 checkpoint 的采样方式。

20 个场景1344 × 768 分辨率362 帧 · 24 FPS读取结果中…
+ - - + + +
+ diff --git a/ref-data.json b/ref-data.json new file mode 100644 index 0000000000000000000000000000000000000000..213d74e8eb73b135b86b61e72a2c4b538750f8d8 --- /dev/null +++ b/ref-data.json @@ -0,0 +1,202 @@ +{ + "status": "running", + "models": [ + { + "id": "hyperflow-8", + "name": "HyperFlow", + "steps": 8, + "label": "FL2VA 训练 · 作者支持 Ref2VA", + "video_shift": 12, + "audio_shift": 3 + }, + { + "id": "fasth3-4", + "name": "FastH3", + "steps": 4, + "label": "T2VA 权重迁移测试", + "video_shift": 12, + "audio_shift": 3 + }, + { + "id": "lightx2v-ref-8", + "name": "LightX2V Ref2VA", + "steps": 8, + "label": "Ref2VA 专用 v1.0", + "video_shift": 6, + "audio_shift": 3 + }, + { + "id": "lightx2v-ref-4", + "name": "LightX2V Ref2VA", + "steps": 4, + "label": "Ref2VA 专用 v0.1", + "video_shift": 6, + "audio_shift": 3 + } + ], + "samples": [ + { + "id": 1, + "source_case": 1, + "slug": "reference-case-1", + "title": { + "zh": "雨夜双角色交锋 · 人物与环境参考", + "en": "Public Ref2VA case 1" + }, + "source": { + "name": "LightX2V public Ref2VA testset", + "url": "https://github.com/ModelTC/Minimax-H3-Turbo/blob/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/prompts_ref2va_test.json" + }, + "source_prompt": "subject_definitions:\n is the long-haired East Asian female hunter in , preserving her face, long wet black hair, black leather combat suit, tattered black cloak, black gloves, and single silver short blade.\n is the East Asian female tech hunter in , preserving her face, short black hair, white fox mask with red markings, black trench coat with cyan circuit lines, black boots, and single blue energy dagger.\n is the rainy cyberpunk intersection in , preserving its skyscrapers, giant blue-purple holographic billboard, neon signs, wet reflective asphalt, rain, and drifting fog.\n\nsummary:\n[reference generation] In one continuous 5.1667-second cinematic shot, and begin in a tense standoff inside , charge across the wet street, exchange one fast blade combination, and end locked blade-to-blade in a burst of blue energy.\n\nretention_analysis:\n (appears throughout): fully_preserved - retain the same face, long hair, black combat suit, cloak, body proportions, and one silver blade.\n (appears throughout): fully_preserved - retain the same face, short hair, fox mask, cyan-lit trench coat, body proportions, and one energy dagger.\n (appears throughout): fully_preserved - retain the wet intersection, billboard, neon palette, rain, fog, and reflective pavement.\n\ndetailed_description:\nPhotorealistic live-action cyberpunk action cinema, 16:9, high contrast purple-and-cyan night lighting, heavy rain, physically coherent motion, and stable identities. One continuous shot with no cuts, no duplicate people, no duplicate weapons, no costume changes, no gore, and no readable text.\n\n[Shot 1, 00:00.000-00:01.200] A low wide camera frames both fighters five meters apart on the rain-soaked street. crouches on frame left with her silver blade held low; stands on frame right with the blue energy dagger raised. Rain streaks through the holographic light and fog crosses the reflective road.\n\n[Shot 1, 00:01.200-00:03.600] Both women burst forward. The camera tracks sideways at waist height as 's cloak opens behind her. She makes one upward diagonal slash; sidesteps and blocks with the energy dagger. A compact blue-white spark and a ring of displaced droplets mark the impact. answers with one horizontal counter, which deflects with the flat of her blade.\n\n[Shot 1, 00:03.600-00:05.166] They rotate once around their joined weapons and stop in a close blade lock, faces and costumes still distinct. Cyan electricity crawls briefly across the crossed blades while rain runs from the fox mask and black cloak. The camera settles into a medium two-shot as the billboard flickers behind them and both hold the unresolved confrontation.\n\noverall_soundscape:\nHeavy rain, wet footfalls, two sharp blade impacts, controlled fabric movement, electrical crackle, distant traffic, and a low thunder roll. No dialogue.\n\nnon_diegetic_music:\nA dark electronic pulse rises through the charge and stops on the final blade lock.\n", + "source_duration": 5.1666666667, + "prompt": "subject_definitions:\n is the long-haired East Asian female hunter in , preserving her face, long wet black hair, black leather combat suit, tattered black cloak, black gloves, and single silver short blade.\n is the East Asian female tech hunter in , preserving her face, short black hair, white fox mask with red markings, black trench coat with cyan circuit lines, black boots, and single blue energy dagger.\n is the rainy cyberpunk intersection in , preserving its skyscrapers, giant blue-purple holographic billboard, neon signs, wet reflective asphalt, rain, and drifting fog.\n\nsummary:\n[reference generation] In one continuous 15-second cinematic shot, and begin in a tense standoff inside , charge across the wet street, exchange one fast blade combination, and end locked blade-to-blade in a burst of blue energy.\n\nretention_analysis:\n (appears throughout): fully_preserved - retain the same face, long hair, black combat suit, cloak, body proportions, and one silver blade.\n (appears throughout): fully_preserved - retain the same face, short hair, fox mask, cyan-lit trench coat, body proportions, and one energy dagger.\n (appears throughout): fully_preserved - retain the wet intersection, billboard, neon palette, rain, fog, and reflective pavement.\n\ndetailed_description:\nPhotorealistic live-action cyberpunk action cinema, 16:9, high contrast purple-and-cyan night lighting, heavy rain, physically coherent motion, and stable identities. One continuous shot with no cuts, no duplicate people, no duplicate weapons, no costume changes, no gore, and no readable text.\n\n[Shot 1] A low wide camera frames both fighters five meters apart on the rain-soaked street. crouches on frame left with her silver blade held low; stands on frame right with the blue energy dagger raised. Rain streaks through the holographic light and fog crosses the reflective road.\n\nDuring the middle of the same continuous shot, Both women burst forward. The camera tracks sideways at waist height as 's cloak opens behind her. She makes one upward diagonal slash; sidesteps and blocks with the energy dagger. A compact blue-white spark and a ring of displaced droplets mark the impact. answers with one horizontal counter, which deflects with the flat of her blade.\n\nDuring the final phase of the same continuous shot, They rotate once around their joined weapons and stop in a close blade lock, faces and costumes still distinct. Cyan electricity crawls briefly across the crossed blades while rain runs from the fox mask and black cloak. The camera settles into a medium two-shot as the billboard flickers behind them and both hold the unresolved confrontation.\n\nThe character sheets provide identity and clothing, not a collage to reproduce. Show one physical version of each fighter in the environment, with their separate silhouettes legible against the billboard. The silver blade belongs only to ; the blue dagger belongs only to . Keep their hands connected naturally to the weapon handles and show planted feet disturbing shallow puddles. During the opening four seconds, let the rain establish depth while both women watch each other without changing sides. Across the middle six seconds, preserve the contact order of approach, slash, block and counter; the sideways camera travels only far enough to keep both bodies in view. Reserve the last five seconds for the rotation, final lock and held reaction. The final two-shot keeps both referenced faces readable, with rain, breathing and small cloak movements continuing around their otherwise stable pose.\n\noverall_soundscape:\nHeavy rain, wet footfalls, two sharp blade impacts, controlled fabric movement, electrical crackle, distant traffic, and a low thunder roll. No dialogue.\n\nnon_diegetic_music:\nA dark electronic pulse rises through the charge and stops on the final blade lock.\n", + "prompt_sha256": "5d2a08b4360b9080603ddd8fb85c73c3386a5dda88600dd557e1626eaaabedb9", + "references": [ + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_1_1.jpg", + "label": "Picture 1", + "sha256": "1665c52d9e20ace593cbc30d1e17b5cf996ad25a59c7f1890d636711da93dbd3", + "bytes": 344933, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_1_1.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_1_1.jpg" + }, + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_1_2.jpg", + "label": "Picture 2", + "sha256": "df77cc201252e90cb828132863d4126c4d62d512f79299ec2651bcc9be8c5b1f", + "bytes": 396459, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_1_2.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_1_2.jpg" + }, + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_1_3.jpg", + "label": "Picture 3", + "sha256": "82cdb7a20d6ec0d7c6c396cc8e99d3f5fe49a2f67f7c3e5e75d45102f4a91062", + "bytes": 390923, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_1_3.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_1_3.jpg" + } + ], + "duration": 15, + "frames": 362, + "seed": 7301, + "adaptations": [ + "保留上游六字段提示词与原始参考图片；时间安排由约 5.17 秒扩展为 15 秒。", + "将上游重复的 Shot 1 时间段整理为一个连续镜头；明确对白说话人；原文台词未改。", + "新增动作衔接与参考保持说明属于本次整理；本组仅测试图片参考，声音由模型生成。" + ] + }, + { + "id": 2, + "source_case": 2, + "slug": "reference-case-2", + "title": { + "zh": "维多利亚书房对白 · 双角色身份", + "en": "Public Ref2VA case 2" + }, + "source": { + "name": "LightX2V public Ref2VA testset", + "url": "https://github.com/ModelTC/Minimax-H3-Turbo/blob/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/prompts_ref2va_test.json" + }, + "source_prompt": "subject_definitions:\n is the fictional Victorian detective in , preserving his angular clean-shaven face, combed-back dark hair, long dark overcoat, brown waistcoat, cream cravat, watch chain, and briar pipe.\n is the fictional Victorian doctor in , preserving his sturdy build, cropped brown hair, heavy moustache, brown tweed suit, dark tie, and wooden cane.\n is the Baker Street study in , preserving the wood paneling, lit fireplace, bookshelves, rain-streaked windows, leather armchair, desk, chandelier, rug, and warm-firelight-versus-cool-window-light composition.\n\nsummary:\n[reference generation] In one continuous 5.1667-second locked-off Victorian dialogue shot, asks whether could have become a criminal; answers with dry confidence while both remain naturally alive in .\n\nretention_analysis:\n (appears throughout): fully_preserved - retain his face, hair, clothing, pipe, slim proportions, and fireplace-side placement.\n (appears throughout): fully_preserved - retain his face, moustache, tweed suit, cane, sturdy proportions, and armchair placement.\n (appears throughout): fully_preserved - retain the room layout, fireplace, shelves, windows, desk, furniture, rain, and mixed warm-cool lighting.\n\ndetailed_description:\nPhotorealistic live-action Victorian feature-film style, restrained 35 mm texture, natural skin, subtle performance, and consistent eyelines. One continuous medium-wide two-shot from a fully static locked-off camera. No cuts, camera movement, identity swapping, wardrobe changes, extra people, readable text, or exaggerated gestures.\n\n[Shot 1, 00:00.000-00:01.300] stands beside the fireplace on frame left, his pipe held near his chest, while sits in the leather armchair on frame right with one hand resting on his cane. Wind-driven rain traces the cool windows behind them. A coal shifts in the grate, briefly sharpening warm firelight across 's angular profile and 's moustached face.\n\n[Shot 1, 00:01.300-00:03.200] leans forward just enough to make the leather chair creak, studies the detective, and asks with familiar curiosity, [English] What if you'd chosen crime? His fingers tighten once around the cane handle. remains still for a beat, draws quietly from the pipe, and turns only his eyes toward the doctor as smoke curls through the warm light.\n\n[Shot 1, 00:03.200-00:05.166] meets his eyeline and replies in a low, dry voice, [English] I'd have been remarkably successful. 's amused expression tightens into thoughtful concern. allows the faintest private smile as the fire gives one crisp crackle and a distant lightning reflection passes across the rain-streaked glass. Both settle into the charged silence without changing position.\n\noverall_soundscape:\nSteady rain on glass, close coal-fire crackle, faint room tone, one subtle leather creak, and distant carriage wheels. Dialogue stays clear and intimate.\n\nnon_diegetic_music:\nN/A\n", + "source_duration": 5.1666666667, + "prompt": "subject_definitions:\n is the fictional Victorian detective in , preserving his angular clean-shaven face, combed-back dark hair, long dark overcoat, brown waistcoat, cream cravat, watch chain, and briar pipe.\n is the fictional Victorian doctor in , preserving his sturdy build, cropped brown hair, heavy moustache, brown tweed suit, dark tie, and wooden cane.\n is the Baker Street study in , preserving the wood paneling, lit fireplace, bookshelves, rain-streaked windows, leather armchair, desk, chandelier, rug, and warm-firelight-versus-cool-window-light composition.\n\nsummary:\n[reference generation] In one continuous 15-second locked-off Victorian dialogue shot, asks whether could have become a criminal; answers with dry confidence while both remain naturally alive in .\n\nretention_analysis:\n (appears throughout): fully_preserved - retain his face, hair, clothing, pipe, slim proportions, and fireplace-side placement.\n (appears throughout): fully_preserved - retain his face, moustache, tweed suit, cane, sturdy proportions, and armchair placement.\n (appears throughout): fully_preserved - retain the room layout, fireplace, shelves, windows, desk, furniture, rain, and mixed warm-cool lighting.\n\ndetailed_description:\nPhotorealistic live-action Victorian feature-film style, restrained 35 mm texture, natural skin, subtle performance, and consistent eyelines. One continuous medium-wide two-shot from a fully static locked-off camera. No cuts, camera movement, identity swapping, wardrobe changes, extra people, readable text, or exaggerated gestures.\n\n[Shot 1] stands beside the fireplace on frame left, his pipe held near his chest, while sits in the leather armchair on frame right with one hand resting on his cane. Wind-driven rain traces the cool windows behind them. A coal shifts in the grate, briefly sharpening warm firelight across 's angular profile and 's moustached face.\n\nDuring the middle of the same continuous shot, (S1) leans forward just enough to make the leather chair creak, studies the detective, and asks in a warm, mid-pitched English voice with familiar curiosity, [English] What if you'd chosen crime? His fingers tighten once around the cane handle. remains still for a beat, draws quietly from the pipe, and turns only his eyes toward the doctor as smoke curls through the warm light.\n\nDuring the final phase of the same continuous shot, (S2) meets his eyeline and replies in a low, dry English voice, [English] I'd have been remarkably successful. 's amused expression tightens into thoughtful concern. allows the faintest private smile as the fire gives one crisp crackle and a distant lightning reflection passes across the rain-streaked glass. Both settle into the charged silence without changing position.\n\nUse the portrait and turnaround sheets to preserve each man's identity, not to reproduce a reference-board layout. The seated doctor's sturdy shoulders, moustache and cane distinguish him from the standing detective's narrow profile and pipe. Keep their hands and props separate and their eyelines directed toward each other. The first four seconds establish their positions and the room's depth: chair legs sit firmly on the rug, the desk remains behind them, and warm reflections move subtly across the wood paneling. Allow roughly five seconds for the doctor's complete question and the detective's silent consideration, then leave the remaining six seconds for the complete reply and both reactions. Do not add another line. The doctor listens with his lips closed during the reply; the detective keeps his lips closed during the question. Small breaths, a natural blink and a restrained finger adjustment prevent a frozen tableau. Preserve the fixed viewpoint through the final pause, keeping the fire visible and the cool rainy windows distinct behind the two warm-lit faces.\n\noverall_soundscape:\nSteady rain on glass, close coal-fire crackle, faint room tone, one subtle leather creak, and distant carriage wheels. Dialogue stays clear and intimate.\n\nnon_diegetic_music:\nN/A\n", + "prompt_sha256": "593dde5ba7be9ded40395af7de00ed01fd5cc7e1ac778276bac5c61c09f7f278", + "references": [ + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_2_1.jpg", + "label": "Picture 1", + "sha256": "3808f06a4b21910dcc04a4f62f6fe95714d0c9773cd9f81fd67695d4eb924373", + "bytes": 383601, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_2_1.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_2_1.jpg" + }, + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_2_2.jpg", + "label": "Picture 2", + "sha256": "b62cf2f93c256975c80fd58869216a819749a8f3aab4bc03b396b6147f0848eb", + "bytes": 278692, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_2_2.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_2_2.jpg" + }, + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_2_3.jpg", + "label": "Picture 3", + "sha256": "d1aac91407587e4457ad191e3867e5f24dd878870ca4bbafc40ca24c605c86df", + "bytes": 194821, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_2_3.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_2_3.jpg" + } + ], + "duration": 15, + "frames": 362, + "seed": 7302, + "adaptations": [ + "保留上游六字段提示词与原始参考图片；时间安排由约 5.17 秒扩展为 15 秒。", + "将上游重复的 Shot 1 时间段整理为一个连续镜头；明确对白说话人；原文台词未改。", + "新增动作衔接与参考保持说明属于本次整理；本组仅测试图片参考，声音由模型生成。" + ] + }, + { + "id": 3, + "source_case": 4, + "slug": "reference-case-4", + "title": { + "zh": "沙漠观测者 · 单图人物与场景", + "en": "Public Ref2VA case 4" + }, + "source": { + "name": "LightX2V public Ref2VA testset", + "url": "https://github.com/ModelTC/Minimax-H3-Turbo/blob/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/prompts_ref2va_test.json" + }, + "source_prompt": "subject_definitions:\n is a tan-skinned adult man with close-cropped dark hair, light stubble, a sand-colored field jacket, and a burgundy scarf, as shown in .\n is a vintage brass telescope on a dark tripod, as shown in .\n is a small red-glowing metal lantern on a weathered table, as shown in .\n is a quiet desert observatory at dusk with soft mountain silhouettes, as shown in .\n\nsummary:\n[reference generation] In one continuous 5.1667-second shot inside , notices a distant flash, turns toward , adjusts its focus ring once, and reaches the eyepiece as pulses red.\n\nretention_analysis:\n (appears in [Shot 1]): fully_preserved - Preserve the same adult identity, facial structure, hair, stubble, jacket, scarf, and body proportions.\n (appears in [Shot 1]): fully_preserved - Preserve the same brass tube, black fittings, tripod geometry, scale, and position.\n (appears in [Shot 1]): fully_preserved - Preserve the same metal housing and red light.\n (appears in [Shot 1]): fully_preserved - Preserve the dusk palette, mountain horizon, and sparse desert setting.\n\ndetailed_description:\nVisual style: cinematic naturalism with warm lantern light against a cool violet dusk. One locked medium shot preserves the composition from . The camera stays fixed with no pan, tilt, zoom, or dolly.\n\n[Shot 1, 00:00.000-00:01.300] stands beside as the final dusk light outlines the brass tube and distant mountain ridge. A faint cold flash crosses his eyes from outside frame. He stops breathing for a beat and turns his gaze sharply toward the telescope while holds a steady red glow on the table.\n\n[Shot 1, 00:01.300-00:03.600] He reaches across the brass tube, grips the focus ring, and makes exactly one deliberate adjustment. The mechanism gives a precise metal click. Reflected violet sky slides across the brass surface, his scarf lifts in the desert wind, and flickers once brighter without moving.\n\n[Shot 1, 00:03.600-00:05.166] leans quickly but naturally to the eyepiece, steadies the telescope with his free hand, and peers through it with sudden concentration. The lantern throws a brief red edge across his cheek and telescope fittings. He holds the tense viewing pose as the distant flash fades behind the mountains.\n\noverall_soundscape:\nA faint desert breeze, a soft metal adjustment click, fabric rustle, and a quiet lantern flame.\n\nnon_diegetic_music:\nN/A\n", + "source_duration": 5.1666666667, + "prompt": "subject_definitions:\n is a tan-skinned adult man with close-cropped dark hair, light stubble, a sand-colored field jacket, and a burgundy scarf, as shown in .\n is a vintage brass telescope on a dark tripod, as shown in .\n is a small red-glowing metal lantern on a weathered table, as shown in .\n is a quiet desert observatory at dusk with soft mountain silhouettes, as shown in .\n\nsummary:\n[reference generation] In one continuous 15-second shot inside , notices a distant flash, turns toward , adjusts its focus ring once, and reaches the eyepiece as pulses red.\n\nretention_analysis:\n (appears in [Shot 1]): fully_preserved - Preserve the same adult identity, facial structure, hair, stubble, jacket, scarf, and body proportions.\n (appears in [Shot 1]): fully_preserved - Preserve the same brass tube, black fittings, tripod geometry, scale, and position.\n (appears in [Shot 1]): fully_preserved - Preserve the same metal housing and red light.\n (appears in [Shot 1]): fully_preserved - Preserve the dusk palette, mountain horizon, and sparse desert setting.\n\ndetailed_description:\nVisual style: cinematic naturalism with warm lantern light against a cool violet dusk. One locked medium shot keeps the referenced man, telescope, lantern and desert setting spatially coherent. The camera stays fixed with no pan, tilt, zoom, or dolly.\n\n[Shot 1] stands beside as the final dusk light outlines the brass tube and distant mountain ridge. A faint cold flash crosses his eyes from outside frame. He stops breathing for a beat and turns his gaze sharply toward the telescope while holds a steady red glow on the table.\n\nDuring the middle of the same continuous shot, He reaches across the brass tube, grips the focus ring, and makes exactly one deliberate adjustment. The mechanism gives a precise metal click. Reflected violet sky slides across the brass surface, his scarf lifts in the desert wind, and flickers once brighter without moving.\n\nDuring the final phase of the same continuous shot, leans quickly but naturally to the eyepiece, steadies the telescope with his free hand, and peers through it with sudden concentration. The lantern throws a brief red edge across his cheek and telescope fittings. He holds the tense viewing pose as the distant flash fades behind the mountains.\n\nTreat as a source for the man, telescope, lantern and desert setting rather than as a required first frame. Keep a single adult man in the scene with the same cropped hair, stubble and burgundy scarf; preserve the jacket seams, brass tube proportions and dark tripod supports. His hands remain visibly attached to his arms as one hand finds the focus ring and the other steadies the instrument. The opening four seconds allow the distant light to pass across his eyes while the lantern remains steady. Over the next six seconds, his gaze follows the flash, his body turns toward the instrument, and his fingers make one deliberate focusing adjustment. During the final five seconds, he leans to the eyepiece and settles into concentrated observation. Let his scarf and jacket respond to the same gentle wind throughout. Keep the table and lantern stationary, with the red pulse changing only the light on nearby surfaces. The camera holds its viewpoint, maintaining a clear gap between the man, telescope tube and mountain skyline; no new people, additional telescopes, text or sudden camera cuts appear.\n\noverall_soundscape:\nA faint desert breeze, a soft metal adjustment click, fabric rustle, and a quiet lantern flame.\n\nnon_diegetic_music:\nN/A\n", + "prompt_sha256": "e427f08a215a42180ea4e363c45c306819885e80729580173bb9a4f59f321da2", + "references": [ + { + "type": "image", + "path": "ref-inputs/ref2va_testset/ref2va_test_4_1.jpg", + "label": "Picture 1", + "sha256": "3b54a682135c5e60c3584df15754c8d54177738686a0778ca89f0a326fe50bbd", + "bytes": 290315, + "source_url": "https://raw.githubusercontent.com/ModelTC/Minimax-H3-Turbo/02e26d591f7a04d5d1a074c9566d5dd4f22f6225/examples/ref2va_testset/ref2va_test_4_1.jpg", + "public_path": "assets/ref2va-inputs/ref2va_test_4_1.jpg" + } + ], + "duration": 15, + "frames": 362, + "seed": 7303, + "adaptations": [ + "保留上游六字段提示词与原始参考图片；时间安排由约 5.17 秒扩展为 15 秒。", + "将上游重复的 Shot 1 时间段整理为一个连续镜头；明确对白说话人；原文台词未改。", + "新增动作衔接与参考保持说明属于本次整理；本组仅测试图片参考，声音由模型生成。" + ] + } + ], + "outputs": [], + "source_revision": "02e26d591f7a04d5d1a074c9566d5dd4f22f6225", + "updated_at": "2026-09-13T09:40:38.581622+00:00", + "profile": { + "task": "ref2va", + "duration_s": 15, + "frames": 362, + "gpus": 4, + "cluster": "hsg", + "compute_quant": "none", + "reference_types_tested": [ + "image" + ] + }, + "media_repository": { + "repo_id": "Lawrence-cj/sol-h3-hyperflow-20260913-media", + "latest_commit": "2d72924012f74288d6ce319ebfe915fa21162237", + "url": "https://huggingface.co/datasets/Lawrence-cj/sol-h3-hyperflow-20260913-media" + } +} diff --git a/ref.js b/ref.js new file mode 100644 index 0000000000000000000000000000000000000000..55cc6268ea0e29ad031e287230fa44c69c9295bc --- /dev/null +++ b/ref.js @@ -0,0 +1,17 @@ +(() => { + const q=s=>document.querySelector(s), make=(tag,value,cls)=>{const e=document.createElement(tag);e.textContent=value;if(cls)e.className=cls;return e;}; + let study,index=0; + document.querySelectorAll('.tab').forEach(tab=>tab.addEventListener('click',()=>document.querySelectorAll('video').forEach(v=>v.pause()))); + function render(){ + const s=study.samples[index];q('#ref-title').textContent=s.title.zh;q('#ref-prompt').textContent=s.prompt;q('#ref-source-prompt').textContent=s.source_prompt;q('#ref-adaptations').textContent=s.adaptations.join(' '); + const source=make('a',s.source.name+' · 上游来源 ↗');source.href=s.source.url;source.target='_blank';source.rel='noreferrer';q('#ref-source').replaceChildren(source); + const inputs=q('#ref-inputs');inputs.replaceChildren();for(const ref of s.references){const figure=document.createElement('figure'),a=document.createElement('a'),img=document.createElement('img');a.href=ref.public_path;a.target='_blank';a.rel='noreferrer';img.src=ref.public_path;img.alt=ref.label;img.loading='lazy';a.append(img);figure.append(a,make('figcaption',`<${ref.label}> · 原始参考图片 · 点击查看原图`));inputs.append(figure);} + const grid=q('#ref-videos');grid.querySelectorAll('video').forEach(v=>v.pause());grid.replaceChildren(); + for(const m of study.models){const o=study.outputs.find(v=>v.id===s.id&&v.adapter===m.id),card=make('article','','card'+(m.id==='hyperflow-8'?' featured':'')),head=make('div','','card-head'),name=make('div',m.name,'model-name');name.append(make('span',m.steps+'-step','badge'));head.append(name,make('span',m.label,'card-label'));const frame=make('div','','video-frame'),foot=make('div','','card-foot'); + if(o){const v=document.createElement('video');v.src=o.video;v.poster=o.poster;v.controls=true;v.preload='metadata';v.muted=true;v.playsInline=true;frame.append(v);foot.append(make('span',`HSG · 4 GPU · 单次 ${o.pipeline_s.toFixed(2)} s`));const sound=make('button','听这段音频');sound.onclick=()=>{document.querySelectorAll('video').forEach(x=>x.muted=x!==v);v.play().catch(()=>{});};foot.append(sound);}else{frame.append(make('div','此配置正在运行，尚无已验证结果','empty'));} + card.append(head,frame,foot,make('div',`Ref2VA · 图片参考 · shift ${m.video_shift}/${m.audio_shift}`,'recipe'));grid.append(card); + } + } + q('#ref-select').addEventListener('change',e=>{index=Number(e.target.value);render();});q('#ref-play-all').onclick=()=>{const vs=[...q('#ref-videos').querySelectorAll('video')];vs.forEach(v=>{v.currentTime=0;v.muted=true;v.play().catch(()=>{});});}; + fetch('ref-data.json').then(r=>{if(!r.ok)throw new Error(r.status);return r.json();}).then(d=>{study=d;d.samples.forEach((s,i)=>{const o=make('option',`${s.id} · ${s.title.zh}`);o.value=i;q('#ref-select').append(o);});q('#ref-completion').textContent=`${d.outputs.length} / ${d.samples.length*d.models.length} 段已验证结果 · ${d.status==='complete'?'全部完成':'其余作业继续运行'}`;render();if(location.hash==='#ref2va')q('[data-panel="ref2va"]').click();}).catch(e=>{q('#ref-completion').textContent='参考结果读取失败：'+e.message;}); +})(); diff --git a/style.css b/style.css index 4fba6a0f23dc6d7177180e3bee58b22911298b4c..7b4c26c6ec40190e13d7ddc66042fd811bf716ec 100644 --- a/style.css +++ b/style.css @@ -1,3 +1,5 @@ :root{color-scheme:light;--ink:#162624;--muted:#677674;--line:#dce4df;--paper:#f7f8f4;--green:#17634e;--mint:#d8f2ce}*{box-sizing:border-box}body{margin:0;background:var(--paper);font-family:Inter,-apple-system,BlinkMacSystemFont,"PingFang SC","Microsoft YaHei",sans-serif;color:var(--ink);font-size:14px}a{color:inherit;text-underline-offset:4px}header{height:88px;border-bottom:1px solid var(--line);display:flex;align-items:center;justify-content:space-between;margin:0 5%;gap:24px}.brand{text-decoration:none;font-weight:800;font-size:26px;letter-spacing:-1.5px}.brand span{color:var(--green)}.header-note{font-size:11px;letter-spacing:2px;color:var(--muted)}header>a:last-child{font-size:12px}main{max-width:1600px;margin:auto;padding:0 5%}.intro{padding:60px 0 34px;position:relative}.eyebrow{font-size:11px;letter-spacing:1.6px;color:var(--green);font-weight:650;text-transform:uppercase}h1{font-size:clamp(35px,4vw,60px);letter-spacing:-2px;line-height:1.16;font-weight:650;margin:18px 0 24px}p{line-height:1.8;color:var(--muted)}.intro>p{max-width:660px;font-size:15px}.facts{display:flex;gap:12px;flex-wrap:wrap;margin-top:26px}.facts span{background:#edf0e8;padding:10px 15px;border-radius:6px;font-size:12px}.facts b{font-weight:600}.facts span:last-child{background:var(--mint);color:var(--green)}.tabs{display:flex;gap:32px;border-bottom:1px solid var(--line);margin:15px 0 30px}.tabs button{border:0;border-radius:0;padding:15px 0;background:transparent}.tabs .active{color:var(--green);border-bottom:2px solid var(--green);font-weight:650}.panel{min-height:400px}.hidden{display:none!important}button,select{font:inherit;color:inherit;border:1px solid var(--line);background:white;border-radius:7px;padding:10px 14px;cursor:pointer}button:hover{border-color:var(--green)}button:focus-visible,select:focus-visible,summary:focus-visible{outline:3px solid #8acdaa;outline-offset:3px}.toolbar{display:flex;gap:10px;align-items:center;flex-wrap:wrap}.toolbar>label{color:var(--muted);display:flex;align-items:center;gap:12px;flex:1}select{min-width:270px;max-width:430px;width:100%}.primary,button.active:not(.tab){background:var(--green);color:white;border-color:var(--green)}.step-filter{display:flex;gap:4px;margin-right:10px}.step-filter button{padding:9px 12px}.scene-heading{display:flex;justify-content:space-between;align-items:end;margin:32px 0 18px;gap:20px}h2{font-size:23px;letter-spacing:-.5px;margin:8px 0;font-weight:620}.scene-heading>span{font-family:monospace;color:var(--muted);font-size:12px}.video-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:22px}.card{border:1px solid var(--line);border-radius:10px;overflow:hidden;background:white}.card.featured{border-color:#a0c7a0}.card-head{padding:14px 17px;display:flex;justify-content:space-between;gap:12px;align-items:center}.model-name{font-size:16px;font-weight:650}.badge{background:#edf0e8;border-radius:4px;padding:4px 7px;margin-left:8px;font-size:11px;font-weight:500}.featured .badge{background:var(--mint);color:var(--green)}.card-label{font-size:11px;color:var(--muted)}.video-frame{aspect-ratio:1344/768;background:#14201e;display:flex;align-items:center;justify-content:center}.video-frame video{width:100%;height:100%;display:block}.empty{color:#a8bbb4;text-align:center;font-size:13px}.empty span{display:block;font-size:10px;letter-spacing:1.3px;margin-top:8px;color:#6f8980}.card-foot{padding:12px 17px;display:flex;justify-content:space-between;align-items:center;font-size:11px;color:var(--muted);gap:10px}.card-foot button{font-size:11px;padding:5px 9px}.prompt-box{border:1px solid var(--line);background:#f0f3ec;border-radius:8px;margin-top:25px;padding:17px 20px}summary{cursor:pointer;font-weight:550}pre{white-space:pre-wrap;overflow-wrap:anywhere;font-family:inherit;line-height:1.8;font-size:13px;color:#3d5049;max-height:300px;overflow-y:auto}#attribution,#adaptations{font-size:12px;line-height:1.8}.sample-nav{display:flex;justify-content:space-between;align-items:center;margin:26px 0 55px}.sample-nav>span{color:var(--muted);font-size:12px}.section-heading{max-width:770px;margin-bottom:25px}.table-scroll{overflow-x:auto}table{border-collapse:collapse;width:100%;text-align:left;background:white;font-size:12px}th,td{padding:16px;border-bottom:1px solid var(--line);white-space:nowrap}th{font-size:11px;color:var(--muted);font-weight:550;background:#edf0e8}td:nth-last-child(-n+3){font-variant-numeric:tabular-nums}.reference{max-width:800px;margin:36px 0 60px}.reference h3{font-size:17px}.reference p{font-size:12px}.method-grid{display:grid;grid-template-columns:1fr 1fr;gap:32px 60px;max-width:1120px;margin:40px 0 70px}.method-grid article{border-top:1px solid var(--line);padding-top:16px}.method-grid h2{font-size:19px}.method-grid p{font-size:13px}footer{border-top:1px solid var(--line);margin:0 5%;padding:25px 0 35px;display:flex;justify-content:space-between;gap:20px;color:var(--muted);font-size:11px}@media(min-width:1450px){.video-grid{grid-template-columns:repeat(4,minmax(0,1fr))}.video-grid.two{grid-template-columns:repeat(2,minmax(0,1fr))}}@media(max-width:760px){header{height:68px}.header-note{display:none}.intro{padding-top:36px}.toolbar>label{flex-basis:100%;flex-direction:column;align-items:start;gap:6px}select{max-width:none}.step-filter{margin-right:auto}.video-grid{grid-template-columns:1fr}.scene-heading{align-items:start}.scene-heading>span{white-space:nowrap;margin-top:8px}h2{font-size:19px}.facts{gap:6px}.facts span{padding:8px 10px}.method-grid{grid-template-columns:1fr;gap:20px}footer{flex-direction:column;gap:8px}.card-label{font-size:10px}.toolbar button{font-size:12px;padding:9px 11px}} .recipe{padding:0 17px 13px;font-size:10px;color:var(--muted)} + +.ref-input-grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(220px,1fr));gap:16px;margin:18px 0 28px}.ref-input-grid figure{margin:0;background:#12151d;padding:12px;border-radius:12px}.ref-input-grid img{width:100%;height:210px;object-fit:contain}.ref-input-grid figcaption{font-size:13px;margin-top:8px;color:#b4bdcd}