After a number of days chatting w/ various LLMs, extracting/merging blocks, and generating over 50 33gb checkpoints w/ various combinations of minor alterations here's my best attempt at getting Vace 2.1 to work w/ Wan2.2. I'm not an expert at testing so no promises on quality or performance, but I figured after the time I played around w/ this I might as well toss it out for review and discussion rather than just sitting on it.
Example Outputs
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for CCP6/FakeVace2.2
Base model
Wan-AI/Wan2.1-VACE-14B