Kandinsky 5 step distill

Open noise2014 opened this issue 2 months ago • 4 comments

Hey team, thanks for all the amazing releases. Are you going to work on step distill model for Kandinsky 5? its getting native support later. thanks once again.

Nov 25 '25 14:11 noise2014

Soon it will get native ComfyUI support. The model is much better in motion dynamics and sematics than Wan2.2 and it's a single model unlike Wan so it will be much easier to train. There are two models, 19B and 2B.

https://huggingface.co/collections/kandinskylab/kandinsky-50-video-pro

Nov 25 '25 15:11 MeiYi-dev

it now has native support in ComfyUI.

Dec 07 '25 11:12 Pandaabear0

Yeah it's now natively supported in ComfyUI and diffusers. Can we get a distlled model please.

Dec 07 '25 14:12 MeiYi-dev

Agree, the model looks really good (t2v)! Can we get a distlled model/lora please (for pro and light versions)?

Dec 22 '25 12:12 Kvento