huggingface
[Core] introduce videoprocessor.
What does this PR do?
Introduces a VideoProcessor akin to VaeImageProcessor to encapsulate the logic of dealing with videos.
TODOs
- [x] Add tests
- [x] Docs
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.
@DN6 please review https://github.com/huggingface/diffusers/pull/7861 first.
@yiyixuxu @DN6 now that https://github.com/huggingface/diffusers/pull/7861 is in, WDYT about merging this? I have also added documentation.
The failing test is completely unexpected. Need to look deeper. @yiyixuxu WDYT about the preprocess_video refactor btw?
@yiyixuxu could you look into the failing test? It's likely coming from image processor. But if not, let me know.
could you look into the failing test? It's likely coming from image processor. But if not, let me know.
@sayakpaul can confirm that test failure is due to image processor, I will fix it!
@yiyixuxu I think I have addressed your comments. LMK if this is good to merge.
i left one comment https://github.com/huggingface/diffusers/pull/7776/files#r1597075529 we should accept same video input format across all our pipelines, I don't think that animatediff need to be different other than that it looks good to me!
we should accept same video input format across all our pipelines, I don't think that animatediff need to be different
Okay. I will delete the block that listifies a single image.