Could you apply the same technique for real-time video editing?
not sure if it's possible to just plug-and-play it or if we would need an extra LCM-LoRA for the motion module.
once we have these sort of models producing frames in milliseconds we should be able to do something similar to this demo but with videos.