Start from footage of someone talking and swap in a different voice track, or start from one still photo and get a presenter who never existed. Either way the mouth matches the audio.
A clip to re-voice, or a photo to bring to life
Both routes take an audio track and hand back a talking take. Pick a video and the speaker on screen mouths the new words. Pick a single still portrait and the photo speaks, with no footage of that person needed.
Video + audio → the same take, re-voiced
One still photo + audio → a talking presenter, no filming required
Works with uploaded recordings or voiceovers generated on the Audio tab
Translate, fix a line, build a presenter
The same move covers jobs that used to mean a reshoot. Dub a take into another language, replace one flubbed sentence, or stand up a spokesperson from a single generated headshot.
Fix a single line without bringing anyone back on camera
Generate a headshot, hand it a script read, and you have a presenter
Need the words in another language too? The video translator does both at once
Questions
What does lip sync need from me?
Either a video with a clearly visible speaker or a single still portrait, plus an audio track. Both can come from your library: upload a recording, or generate the voice on the Audio tab.
Can I make a talking video from one photo?
Yes, that is the still-portrait route. Choose a photo and a voice track, and the model animates the face to match. Several models cover it at different price and quality points, including OmniHuman and Kling Avatar.
Does lip sync work on AI-generated characters?
Yes, and it is one step rather than two. Generate a headshot on the Image tab, then hand that still straight to a talking-head model with your voice track. There is no need to animate it into a clip first.
How long can the audio be?
Most models follow the whole track, so the output runs as long as the voiceover you give it. A few are fixed-length and say so in the picker, and the token price is quoted before you run.
How is lip sync priced?
Per second: of the source clip when you re-voice footage, or of the audio track when you animate a still. The token price is quoted before you confirm the run.
Where does the synced clip go?
Into your media library as a new clip. The original stays as it was, and the synced take drops onto the timeline like any other footage.