A model that redoes the lip movement in a finished clip to match a new audio track. You bring a video and a sound, and the person says exactly those words. The distinction is price: it stays the same up to 40 seconds of clip, so on short ad clips LatentSync comes out cheaper.
| What it does | Fits lips to new audio |
|---|---|
| Input | A clip and a ready track |
| Price up to 40 seconds | $0.26 for all |
| A 60-second clip | $0.39 |
| A 120-second clip | $0.78 |
| Strong side | Cheap on short clips |
A 5-second clip and a 40-second clip cost the same, $0.26. For short ad clips this beats billing by the second.
Past 40 seconds the price grows by about $0.0065 a second: a 60-second clip costs $0.39, a 120-second clip $0.78.
When the voice was swapped for another language, the lips still follow the old speech. This model fits them to the new track, and the clip stops looking dubbed.
If the face in the frame is large and you need care, look at VEED Lipsync: it holds the face more exactly, but it is billed by the second and costs more on long clips.
LatentSync is better on short clips up to 40 seconds: the price is fixed. VEED Lipsync holds the face more carefully, but it is billed by the second.
$0.26. A 40-second clip costs the same.
Make a voiceover in the “Voice only” mode, for example with ElevenLabs Turbo, or upload your own track: a voice-recorder file will do.
The result depends on the shot: sharp, frontal faces work best. If you need precision on a close-up, compare with the other option on a short fragment.