ComfyUI Node
Halo-Lipsy v2.4.0
A ComfyUI node in Halo-Lipsy with 15 inputs and 2 outputs.
Halo-Lipsy v2.4.0
- images
- audio
- images
- audio
◄checkpointauto►
◄fps30.00►
◄modesequential►
◄trim_to_audiotrue►
◄face_detect_batch4►
◄face_detect_interval1►
◄inference_batch64►
◄face_padding5►
◄sync_offset0►
◄mel_step_multiplier1.00►
◄smooth_box_frames5►
◄temporal_smooth0.20►
◄force_cpufalse►
CategoryHalo-Lipsy
Inputs (15)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| audio | AUDIO | — | |
| checkpointopt | COMBO | auto | 1 options: auto |
| fpsopt | FLOAT | 30.001–120 | Video FPS - must match your input video for correct sync |
| modeopt | COMBO | sequential | 2 options: sequential, repetitive |
| trim_to_audioopt | BOOLEAN | true | ON = trim video to audio length, OFF = keep full video (no lip sync after audio ends) |
| face_detect_batchopt | INT | 41–32 | Batch size for face detection (CPU) |
| face_detect_intervalopt | INT | 11–10 | Detect face every Nth frame (interpolate between). Higher = faster, lower = more accurate |
| inference_batchopt | INT | 641–256 | Batch size for Wav2Lip inference (GPU) |
| face_paddingopt | INT | 50–50 | Padding around detected face in pixels (less = more mouth resolution) |
| sync_offsetopt | INT | 0-10–10 | Audio sync offset in frames (negative = audio earlier) |
| mel_step_multiplieropt | FLOAT | 1.000.5–2 | Lip sync timing (>1 = faster mouth) |
| smooth_box_framesopt | INT | 51–15 | Frames for smoothing face box movement |
| temporal_smoothopt | FLOAT | 0.200–0.5 | Temporal smoothing (0 = off, 0.2 = blend 20% of previous frame mouth) |
| force_cpuopt | BOOLEAN | false | Run all inference on CPU (no VRAM) |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| audio | AUDIO | — |