Rename code
Browse files
README.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
---
|
| 2 |
license: other
|
| 3 |
license_name: apache-2.0-and-cc-by-nc-sa-4.0
|
| 4 |
-
license_link: https://github.com/statite-ai/
|
| 5 |
pipeline_tag: text-to-video
|
| 6 |
tags:
|
| 7 |
- video-generation
|
|
@@ -32,9 +32,9 @@ This eliminates online perceptual routing and its associated scoring overhead, w
|
|
| 32 |
|
| 33 |
StrideAR supports **text-to-video, image-to-video, and video-to-video generation** on Krea Realtime 14B, with a single implementation supporting NVIDIA Hopper and Blackwell GPUs.
|
| 34 |
|
| 35 |
-
**[Project Page](https://statite.ai/stridear/) · [Earlier preprint](https://statite.ai/
|
| 36 |
|
| 37 |
-
**.
|
| 267 |
|
| 268 |
-
[Project Page](https://statite.ai/stridear/) · [Earlier preprint](https://statite.ai/
|
|
|
|
| 1 |
---
|
| 2 |
license: other
|
| 3 |
license_name: apache-2.0-and-cc-by-nc-sa-4.0
|
| 4 |
+
license_link: https://github.com/statite-ai/StrideAR#license
|
| 5 |
pipeline_tag: text-to-video
|
| 6 |
tags:
|
| 7 |
- video-generation
|
|
|
|
| 32 |
|
| 33 |
StrideAR supports **text-to-video, image-to-video, and video-to-video generation** on Krea Realtime 14B, with a single implementation supporting NVIDIA Hopper and Blackwell GPUs.
|
| 34 |
|
| 35 |
+
**[Project Page](https://statite.ai/stridear/) · [Paper](https://statite.ai/stridear/stridear.pdf) · [Earlier preprint](https://statite.ai/stridear/specar.pdf) · [GitHub](https://github.com/statite-ai/StrideAR) · [ComfyUI](https://github.com/statite-ai/ComfyUI-StrideAR)**
|
| 36 |
|
| 37 |
+

|
| 38 |
|
| 39 |
*Generation comparison on Krea Realtime 14B using one NVIDIA H100. The methods generate the same video duration from the same prompt and seed. Reported latencies come from the paper's evaluation.*
|
| 40 |
|
|
|
|
| 106 |
### Krea Realtime 14B
|
| 107 |
|
| 108 |
```bash
|
| 109 |
+
git clone https://github.com/statite-ai/StrideAR
|
| 110 |
+
cd StrideAR
|
| 111 |
|
| 112 |
python3.11 -m venv .venv-krea
|
| 113 |
source .venv-krea/bin/activate
|
| 114 |
|
| 115 |
pip install ".[krea]"
|
| 116 |
+
stridear download --backbone krea
|
| 117 |
```
|
| 118 |
|
| 119 |
The Krea setup downloads approximately 50 GB of public model files, including the Krea Realtime 14B target, Self-Forcing 1.3B drafter, text encoder and VAE.
|
|
|
|
| 129 |
Generate video directly from a text prompt.
|
| 130 |
|
| 131 |
```bash
|
| 132 |
+
stridear generate --backbone krea \
|
| 133 |
--prompt "A papercraft coral reef full of colorful fish" \
|
| 134 |
--k 2 \
|
| 135 |
--out reef.mp4
|
|
|
|
| 142 |
Generate a video conditioned on an input image.
|
| 143 |
|
| 144 |
```bash
|
| 145 |
+
stridear generate --backbone krea \
|
| 146 |
--image photo.jpg \
|
| 147 |
--prompt "A cinematic mountain landscape at sunrise" \
|
| 148 |
--k 7 \
|
|
|
|
| 156 |
Generate a video conditioned on an existing clip.
|
| 157 |
|
| 158 |
```bash
|
| 159 |
+
stridear generate --backbone krea \
|
| 160 |
--video input.mp4 \
|
| 161 |
--prompt "A futuristic city at night" \
|
| 162 |
--strength 0.8 \
|
|
|
|
| 183 |
|
| 184 |
The default target budget is k=7. Lower values generally favor speed, while increasing the target budget changes the quality–latency trade-off.
|
| 185 |
|
| 186 |
+
For all options and sample outputs, see the [StrideAR GitHub repository](https://github.com/statite-ai/StrideAR).
|
| 187 |
|
| 188 |
## ComfyUI
|
| 189 |
|
| 190 |
StrideAR is also available as a ComfyUI custom-node integration for Krea Realtime 14B.
|
| 191 |
|
| 192 |
+
**[ComfyUI-StrideAR — Installation, Nodes and Workflows](https://github.com/statite-ai/ComfyUI-StrideAR)**
|
| 193 |
|
| 194 |
The ComfyUI integration uses the same StrideAR inference engine rather than a separate implementation of the decoding algorithm.
|
| 195 |
|
|
|
|
| 230 |
The repository provides benchmark commands and reference latent hashes for reproducing the original H100 experiments.
|
| 231 |
|
| 232 |
```bash
|
| 233 |
+
stridear-benchmark --backbone krea --weights weights
|
| 234 |
```
|
| 235 |
|
| 236 |
On the reference Hopper configuration with FlashAttention-3 and five denoising steps, the benchmark checks generated latents against the published SHA-256 hashes.
|
| 237 |
|
| 238 |
Different attention kernels and GPU architectures can produce numerically different outputs. On other supported configurations, the benchmark records execution times and output hashes without requiring bitwise identity with the H100 reference.
|
| 239 |
|
| 240 |
+
See the [repository documentation](https://github.com/statite-ai/StrideAR) for pinned dependencies, benchmark settings and Helios setup.
|
| 241 |
|
| 242 |
## Citation
|
| 243 |
|
|
|
|
| 259 |
|
| 260 |
Model weights are not redistributed. Each checkpoint remains governed by its original license.
|
| 261 |
|
| 262 |
+
See [LICENSE and file-level notices](https://github.com/statite-ai/StrideAR#license).
|
| 263 |
|
| 264 |
---
|
| 265 |
|
| 266 |
Developed by **[Statite](https://statite.ai)**.
|
| 267 |
|
| 268 |
+
[Project Page](https://statite.ai/stridear/) · [Paper](https://statite.ai/stridear/stridear.pdf) · [Earlier preprint](https://statite.ai/stridear/specar.pdf) · [Source Code](https://github.com/statite-ai/StrideAR) · [ComfyUI](https://github.com/statite-ai/ComfyUI-StrideAR)
|