H3 Audio Retake Studio changes a line of dialogue in a finished clip and repairs the lip-sync, locally with MiniMax H3
H3 Audio Retake Studio is a free, open-source browser studio for ComfyUI that fixes a wrong line in a video without regenerating the shot. You select the words on a waveform, make a new take of the line, and MiniMax H3 with a Pika H3 Flex LoRA repaints the speaker's mouth to match.
What it does
Drag over a passage on the waveform (zoom up to 32x, set IN and OUT to the millisecond or snap to pauses), type the new words, add an optional voice reference and generate takes, with acting presets as prompt guidance. For lip-sync you pick the speaker on the first frame, track the face with FaceRefine, draw an oval over the mouth, then render and compare the original with the repair before exporting the full video. Other modes: Voice only and Off-screen voice keep the picture, Joint retake regenerates voice and mouth in one pass with up to eight dialogue turns per clip, Redo the SFX replaces a shot's sound, and Tracks splits the source into voices and everything else with Mel-Band RoFormer so you can redo one track.
Worth knowing
It is an early prototype tested on one local Windows/NVIDIA setup. You need a working MiniMax H3 setup in ComfyUI, the ComfyUI-H3-Motion-Context-MultiRef, ComfyUI-H3-FaceRefine and VideoHelperSuite node packs, FFmpeg, and a Pika H3 Flex Ref2VA LoRA; no models are included. Each lip repair is limited to about seven seconds, Joint retake needs about 5.2 seconds of source video, and the model can change the face, so check every result. Export re-encodes the MP4, so untouched pixels are not bit-identical. The app is MIT; the models keep their own licences.