Does MotionDirector work on Stable Video Diffusion?

[TruthSearcher]

[ruizhaocv]

This matter is on our to-do list, but we currently don't have a specific estimated time to finish it.
We welcome everyone to try adapting MotionDirector to SVD and collaborate in enhancing this project together. Thanks!

[ruizhaocv]

Thank you. I think it would be better to keep this issue open for others who have similar questions.

[dailingx]

I'm also interested in this

[olliacc]

We would also like to know if MotionDirector works motion modules like animatediff? @ruizhaocv

[ruizhaocv]

Hi olliacc. Thanks for your question. The answer is the same with the SVD. It is on our to-do list, but we don't have an estimated time to finish it. We welcome everyone to adapt MotionDirector to variant base models and collaborate in enhancing this project together.

[howardgriffin]

Any update?

[ruizhaocv]

Hi all. The MotionDirector training for AnimateDiff is available at https://github.com/ExponentialML/AnimateDiff-MotionDirector. Thank @ExponentialML very much for this awesome repo.

[eugenelyj]

@ruizhaocv Dear author, I am trying to implement the SVD version but I have two questions and maybe you have some ideas?

1. How to modify the spatial loss to adapt to the image-to-video pipeline of SVD?
2. How to modify the temporal loss to adapt to the v-prediction formula of SVD?