sync-3 lipsync modelsync-3 is an AI video model for lip sync and visual dubbing that was dropped in the beginning of April 2026. It changes how a person’s mouth moves so it matches new audio, sitting with with other models like lipsync-2 and react-1.
It comes from Synchronicity Labs Inc. also called Sync, a San Francisco startup from 2023. The company has backing from Y Combinator, Google Ventures, Rebel Fund and Creator Ventures.
The model focuses on dubbing and localization. It aims to handle harder shots better.That includes side angles, blocked faces, low light, close shots and scenes with more than one person. It also supports higher quality output like 4K at 60fps.
Sync says this version tries to understand the whole scene, not just the mouth. Earlier models worked in chunks, which could cause flicker or weird edges. This one processes the full shot at once, so results look more steady... more natural-ish.
The company angle is simple. They want video editing to feel like editing text.The team has research background. One co-founder worked on Wav2Lip, a known lip sync system. Another worked on human video generation research. So it’s not just a wrapper app, more like a research-driven product.
In use, it runs in a browser tool or API.People use it for dubbing, translating content, training videos, or custom video messages. It also supports batch work, so teams can process lots of videos at scale.
If you'd like to access this model, you can explore the following possibilities:
