File size: 1,476 Bytes
51aba6b 0722208 ffadf6c 0722208 91037c2 51aba6b 91037c2 51aba6b 91037c2 51aba6b 91037c2 51aba6b 0722208 51aba6b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 |
---
pipeline_tag: image-to-video
---
<p align="center">
<img src="./demos/demo-01.gif" width="70%" />
<img src="./demos/demo-02.gif" width="70%" />
</p>
<p align="center">Samples generated by AnimateLCM-SVD-xt</p>
## Introduction
Consistency Distilled [Stable Video Diffusion Image2Video-XT (SVD-xt)](https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt) following the strategy proposed in [AnimateLCM-paper](https://arxiv.org/abs/2402.00769).
AnimateLCM-SVD-xt can generate good quality image-conditioned videos with 25 frames in 2~8 steps with 576x1024 resolutions.
## Computation comparsion
AnimateLCM-SVD-xt can generally produces demos with good quality in 4 steps without requiring the classifier-free guidance, and therefore can save 25 x 2 / 4 = 12.5 times compuation resources compared with normal SVD models.
## Demos
| | | |
| :---: | :---: | :---: |
|  |  |  |
| 2 steps, cfg=1 | 4 steps, cfg=1 | 8 steps, cfg=1 |
|  |  |  |
| 2 steps, cfg=1 | 4 steps, cfg=1 | 8 steps, cfg=1 |
|  |  |  |
| 2 steps, cfg=1 | 4 steps, cfg=1 | 8 steps, cfg=1 |
Please contact Fu-Yun Wang ([email protected]) for the inference code and the scheduler design. |