ComfyUI Node
LSDub Performance
A character image + a song -> a performance video as long as the song: LTX-2.5 image-to-video shots of ~5 s, placed one after another on the song timeline (open weights; local GPU or Comfy Cloud per the project). Dub it afterwards to make the mouth sing the words.
LSDub Performance
- project
- character
- song
- source
- video
◄promptclose-up, the character sings to the camera, mouth clearly visible►
◄negative_promptblurry, distorted face, extra fingers, text►
◄shot_seconds5.0►
◄max_seconds0►
◄chainsame_still►
CategoryLSDub/video
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| project | LSDUB_PROJECT | The LSDub project (fps, seed, backend). | |
| character | IMAGE | The character to animate (first image is used). | |
| song | LSDUB_SONG | The song: the video is made as long as it. | |
| prompt | STRING | close-up, the character sings to the camera, mouth clearly visible | The performance: what the character does, the camera, the light. |
| negative_prompt | STRING | blurry, distorted face, extra fingers, text | What to avoid. |
| shot_seconds | FLOAT | 5.02–10 | Length of each generated shot, s. |
| max_secondsopt | FLOAT | 00–600 | Stop after this many seconds of the song (0 = the whole song). |
| chainopt | COMBO | same_still | same_still: every shot starts from the character image (a new seed per shot) and keeps the identity best; last_frame: each shot continues from the previous one (smoother, but the face drifts over a long song). |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| source | LSDUB_SOURCE | The performance video (with the song). |
| video | VIDEO | The performance video. |