When lipsync is the right answer
| Situation | Best approach |
|---|---|
| Talking head, mouth clearly visible | Lipsync |
| Voiceover over B-roll | No sync needed — replace audio only |
| Mouth obscured or extreme profile | Re-cut or use voiceover |
| Multiple speakers overlapping | Usually not worth syncing; subtitle |
| Music or singing | Specialised handling; results vary widely |
Do not sync what does not need it: voiceover replacing audio over B-roll needs no lip sync at all. Syncing it wastes time and adds failure modes.
Write a dub-friendly script
This is the highest-leverage step and the one most often skipped. Translation that reads well on the page frequently does not fit the mouth.
- Match sentence length to the original timing. A translation that runs long forces rushed delivery.
- Prefer shorter sentences. They are easier to fit and easier to sync.
- Keep numbers, names and technical terms consistent — these are where dubs lose credibility.
- Read it aloud at the original pace before generating anything. If it does not fit spoken, it will not fit synced.
- Leave room for breath and pause. Natural speech is not continuous.
Voice selection and consent
- If cloning a specific person’s voice, get explicit written consent for that use — voice is usually a separate clearance from image.
- Lock and save the voice profile so it does not drift between episodes or campaigns.
- Check emotional range before committing. Some voices handle excitement or seriousness poorly.
- Test with your real script, including numbers and proper nouns, which expose pronunciation failures.
Consent for voice is separate: teams that clear a performer’s likeness routinely forget the voice. Cloning without specific consent creates avoidable liability.
The production workflow
-
Prepare the dub script with timing in mind
Read aloud and adjust until it fits the original pacing.
-
Generate or record clean dialogue audio
Do this first and approve it. Everything downstream syncs to this audio, so changes later are expensive.
-
Sync per shot
Process shot by shot rather than the whole timeline, so you can judge each result.
-
Judge on motion, not on a still frame
Play each synced shot. Still frames hide the timing errors that viewers notice.
-
Re-cut problem shots
Replace shots that will not sync rather than shipping a bad sync.
-
Mix and master last
Balance dubbed dialogue against music and effects after sync is approved.
Shots that cannot be synced
| Shot | Problem | Solution |
|---|---|---|
| Extreme profile | Mouth largely not visible | Re-cut to a different angle |
| Hand near face | Occlusion confuses sync | Re-cut or regenerate the shot |
| Fast head turns | Motion blur across the mouth | Shorten the shot |
| Eating or drinking | Mouth shape not speech-like | Replace the shot |
| Very wide shot | Mouth too small to read | Accept or subtitle |
Quality checks before delivery
- Watch every synced shot at full speed — do not judge from stills.
- Check the first and last frames of each shot, where drift is most visible.
- Listen for volume and tone consistency across the whole piece.
- Verify numbers, names and product terms are pronounced correctly.
- Confirm disclosure: several markets require labelling synthetic or dubbed performance.
Delivery formats
- Keep the original language as a separate audio track rather than destroying it.
- Ship subtitles as a separate track so viewers can choose.
- Preserve the un-dubbed master so you can revisit without re-generating.
- Label tracks clearly so downstream teams do not mix them up.