ComfyUI Node
TBDub Generate (video + new speech)
Redub an existing video with new speech (TBDub V1.1 Student). Returns the dubbed VIDEO (with the new speech as audio) and a JSON report.
TBDub Generate (video + new speech)
- runtime
- video
- audio
- video
- report
◄face_modefull_frame►
◄seed42►
◄start_frame0►
◄allow_pingpongfalse►
◄multiple_facesrefuse►
◄fps_policyresample_to_25►
CategoryTBDub
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| runtime | TBDUB_RUNTIME | — | |
| video | VIDEO | One person, face visible. Its own audio is not used. Non-25 fps video is converted to 25 fps (see fps_policy). | |
| audio | AUDIO | The new speech. It becomes the output's audio track, whole and unchanged (AAC). | |
| face_mode | COMBO | full_frame | full_frame: find the face with MediaPipe (CPU), dub the 512x512 crop and paste it back into the original frames. cropped: the video is already an aligned, square face crop (it is resized to 512x512). |
| seed | INT | 420–2147483647 | Official default 42; clip i uses seed + i. |
| start_frameopt | INT | 00–100000 | Skip this many source frames; the speech always starts at 0. |
| allow_pingpongopt | BOOLEAN | false | If the source is shorter than the speech, repeat it forward/backward (official behaviour). Off: refuse. |
| multiple_facesopt | COMBO | refuse | refuse: stop if any frame shows more than one face. highest_score: official behaviour (most confident face per frame). |
| fps_policyopt | COMBO | resample_to_25 | TBDub works at 25 fps. resample_to_25: convert other rates with ffmpeg's fps filter (reported). refuse: stop. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| report | STRING | — |