https://github.com/logtd/ComfyUI-MochiEdit
Invert the input video to noise, then use this noise to resample the video with target cues