InvokeAI
InvokeAI copied to clipboard
invoke-ai
[enhancement]: Video generation (Wan 2.1)
Is there an existing issue for this?
- [x] I have searched the existing issues
Contact Details
No response
What should this feature add?
Ability to generate videos with the new Wan 2.1 model. (https://github.com/Wan-Video/Wan2.1)
The model requires only about 8-9 GB of video memory for the smallest version (1.3 billion parameters), which makes it accessible to many users with mid-range video cards. For example on a 4090 video card, a 5-second 480p video takes about 4 minutes without optimization. In a number of tests, Wan 2.1 s comparable to some closed video-gen models, which is impressive for an open solution.
Alternatives
ComfyUI integration of Wan 2.1 as a base example:
(https://comfyanonymous.github.io/ComfyUI_examples/wan/)
Additional Content
No response
I second this, however, I propose to make this on nodes level. I hardly see a place for this feature within linear interface as it's built around image processing in common, videos generation are not the main production pipeline of InvokeAI.
Still, InvokeAI can benefit from this addition in long-term.
It would be great if Invoke could also generate videos. I don't really want to install ComfyUI just for video generation ...
