metadata
pipeline_tag: image-to-video
Samples generated by AnimateLCM-SVD-xt
Introduction
Consistency Distilled Stable Video Diffusion Image2Video-XT (SVD-xt) following the strategy proposed in AnimateLCM-paper. 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 |
 |
 |
 |
| 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 (fywang@link.cuhk.edu.hk) for the inference code and the scheduler design. I might respond a bit later. Thank you!