Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 82, in _split_generators
                  first_examples = list(islice(pipeline, self.NUM_EXAMPLES_FOR_FEATURES_INFERENCE))
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 32, in _get_pipeline_from_tar
                  for filename, f in tar_iterator:
                                     ^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/track.py", line 49, in __iter__
                  for x in self.generator(*self.args):
                           ~~~~~~~~~~~~~~^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/file_utils.py", line 1421, in _iter_from_urlpath
                  yield from cls._iter_zip(f)
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/file_utils.py", line 1391, in _iter_zip
                  zipf = zipfile.ZipFile(f)
                File "/usr/local/lib/python3.14/zipfile/__init__.py", line 1472, in __init__
                  self._RealGetContents()
                  ~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/zipfile/__init__.py", line 1535, in _RealGetContents
                  endrec = _EndRecData(fp)
                File "/usr/local/lib/python3.14/zipfile/__init__.py", line 352, in _EndRecData
                  fpin.seek(0, 2)
                  ~~~~~~~~~^^^^^^
                File "/usr/local/lib/python3.14/site-packages/huggingface_hub/hf_file_system.py", line 1644, in seek
                  raise ValueError("Cannot seek streaming HF file")
              ValueError: Cannot seek streaming HF file
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

synth-v-18k: 合成多视角游戏视频 (Veo 输入包, 2026-11-06)

每个场景 2-5 个人在同一个地方走动, 每个人一个视角 (第一人称 = 他的眼睛; 第三人称 = 跟在他身后的追尾镜头)。 每个视角的输入都已备好 (白模视频 + 带人首帧图 + 人物图 + prompt), 只差用 Veo 生成视频: 模型 gemini-omni-flash-preview, 每个视角一次请求, 所有视角互不依赖, 可以全部并行发。

  • 场景 6000 个, 视角 (= 要生成的视频) 18000 个 (计划 6000 场景 / 18000 视角)
  • 24 个 tar, 共 224 GB (每个约 10 GB, 一个场景不跨 tar), 校验见 SHA256SUMS, 每个 tar 含哪些场景见 manifest.json
  • 场景列表 + 属性 (风格 / 人数 / 人称 / 走位难度 / 时长): index.json (同时在 part 000 里)

数据分布 (按场景计 | 视角数)

人数 (2 / 3 / 4 / 5 人 = 4 : 3 : 2 : 1):

人数 场景 视角
2 人 2400 4800
3 人 1800 5400
4 人 1200 4800
5 人 600 3000

风格 (GTA / 原神 / 艾尔登 / 黑魂 / Apex / 赛博朋克 / 真实世界 = 1:1:1:1:1:1:4):

风格 场景 视角
GTA 600 1783
原神 600 1793
艾尔登法环 600 1787
黑暗之魂 600 1805
Apex 600 1774
赛博朋克 600 1820
真实世界 (6 类合计: 城市 400, 乡村 400, 雪地 400, 沙漠 400, 霓虹夜景 400, 雨天 400) 2400 7238

走位难度 (简单 / 中等 / 复杂 = 4 : 3 : 3):

难度 场景 视角
简单 (匀速, 路线近直) 2400 7209
中等 (一次停下或张望 + 变速) 1800 5384
复杂 (两次事件 + 变速 + 绕行) 1800 5407

时长 (5 / 6 / 7 / 8 / 9 s = 2 : 2 : 2 : 2 : 1; 同一场景的所有视角时长相同):

时长 场景 视角
5 s (120 帧) 1334 4033
6 s (144 帧) 1333 4027
7 s (168 帧) 1333 3972
8 s (192 帧) 1333 3998
9 s (216 帧) 667 1970

人称: 第一人称 3000 场景, 第三人称 3000 场景。

运动约束 (所有场景): 人转身 / 两种镜头转向都 <= 45 度/秒, 单次转向 <= 90 度, 步速 0.7-1.2 米/秒。

生成方式 (和 synth-v-1800 交接包里 omni4 并行版一致)

  • 白模 (灰色几何关卡 + 灰色人形) 由 Blender 渲染, 人形有鼻子 / 帽檐 / 胸板 / 长脚尖标出正面。
  • 首帧图: Qwen-Image 按白模第 0 帧生成写实场景图 (只换材质 / 光照, 从白模加噪起步, 轮廓贴白模; 同一场景的视角按顺序生成, 后面的视角参考前面的材质), 再把每个人按白模朝向 (连续角度) 生成全身像贴到对应位置; 背对 / 正对的人用人体关键点模型检查脸是否可见, 不对就重画。
  • prompt (omni4): "Edit Video1" + Image1 = 目标首帧 + 镜头时间线 (何时向哪边转多少度) + 每个灰色人形绑定一张人物图 (首帧位置 / 入画出画时间 / 画面里的运动轨迹, 由 3D 几何逐帧算出) + 只有 N 个人、不加路人。

下载 / 解包

sha256sum -c SHA256SUMS            # 或只校验下载了的: --ignore-missing
for t in synth_v_18k_part*.tar; do tar xf "$t"; done      # 全部解到 synth_v_18k/
synth_v_18k/
  run_veo_mv.sh                      跑法脚本 (断点续跑; 已有 output.mp4 的视角跳过)
  index.json                         场景列表
  <场景>/case.json                    风格 / 人数 / 人称 / 难度 / 帧数 / 视角列表, 首帧图里每个人的位置和朝向
  <场景>/p<i>/source_clay.mp4         本视角灰度白模视频 (1344x768, 24 fps, 5-9 s)
  <场景>/p<i>/prompt.txt              prompt
  <场景>/p<i>/request.template.json   请求体 (model = gemini-omni-flash-preview; 输入顺序: 源视频, 首帧图, 人物图, 文字;
                                     data 字段是文件名, 发送前替换成 base64, run_veo_mv.sh 自动做)
  <场景>/p<i>/images/                 第 1 张 = 带人首帧图 scene_people_p<i>.png, 其余 = 画面里会出现的每个人的全身像

场景名 v<风格 2 位><关卡 3 位>_<f|b><人数>_m<k>: f = 第一人称, b = 第三人称; 同一关卡 (同一套白模几何) 有 4 个场景 (m0-m3)。 风格编号: 00 GTA, 02 原神, 03 黑魂, 04 艾尔登, 05 Apex, 06-11 真实世界 (城市 / 乡村 / 雪地 / 沙漠 / 霓虹夜景 / 雨天), 12 赛博朋克。

怎么跑

pip install numpy pillow
export VEO_ENDPOINT='https://<endpoint>' VEO_API_KEY='<key>'
cd synth_v_18k && VEO_PARALLEL=64 bash run_veo_mv.sh            # 全部场景; 也可只给部分场景目录名

run_veo_mv.sh 里标了 "按实际接口改" 的三处 (鉴权 header / 任务 id 字段 / 查询 URL) 按你们的接口改; 每个视角成品写到 <场景>/p<i>/output.mp4, 进度在 _status.txt, 日志在 _logs/。

回传

请把 <场景>/p<i>/output.mp4 (和 response.json, 有的话) 按场景目录打包回传 (例如每 1,000 个场景一个 zip); 被拒的请求请保留返回的错误信息。

Downloads last month
70