One fun thing about training models is that you inevitably create a variety of…

April 24, 2025

One fun thing about training models is that you inevitably create a variety of interesting artifacts along the way. Really cool to see this lipsync model from the @heytavus team, and to be able to play with it via the @FAL platform.

Lipsync is one of those genai things that a) seemed like an impossible problem just a few years ago, b) now seems like something we *should* be able to do, and c) is still hard to get exactly right.

Tavus@tavus

We just dropped a new SoTA lipsync model on @FAL: Hummingbird-0

Available now as a research preview, it's the most accurate zero-shot lipsync model we’ve tested, open or closed source.