ComfyUI Node
H3 Lip Sync Transfer (stable face + driven mouth)
A ComfyUI node in minimax/h3/audio with 9 inputs and 3 outputs.
H3 Lip Sync Transfer (stable face + driven mouth)
- source_images
- driving_images
- audio
- frames
- audio
- report
◄fps24.000►
◄animation_regionall►
◄driving_multiplier1.00►
◄audio_offset_frames0►
◄python_exeG:\ComfyUI-aki-v3.2\ComfyUI-aki-v3.2\python\python.exe►
◄liveportrait_dirG:\lipsync-tools\LivePortrait►
Categoryminimax/h3/audio
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| source_images | IMAGE | The stable render (H3AudioDriveLong with chain_frames=true). Supplies the face that survives the whole clip. | |
| driving_images | IMAGE | The audio-driven render of the same shot (chain_frames=false). Supplies the mouth and expression motion. | |
| audio | AUDIO | Returned for the mux, shifted by audio_offset_frames. | |
| fps | FLOAT | 24.0001–120 | — |
| animation_region | COMBO | all | Which motion regions LivePortrait transfers. 'all' is what was measured here; 'lip' only moves the mouth (safer for long takes). |
| driving_multiplier | FLOAT | 1.000.6–1.6 | Expression gain, matching LivePortrait's driving_multiplier. |
| audio_offset_framesopt | INT | 0-48–48 | Same meaning as in the driving nodes: shifts the returned audio against the frames so the lips land on the sound. |
| python_exeopt | STRING | G:\ComfyUI-aki-v3.2\ComfyUI-aki-v3.2\python\python.exe | Interpreter that can import LivePortrait's deps. |
| liveportrait_diropt | STRING | G:\lipsync-tools\LivePortrait | Local LivePortrait checkout (cwd for inference.py). |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| frames | IMAGE | — |
| audio | AUDIO | — |
| report | STRING | — |