VEED launches a Lip Sync 2.0 API for video dubbing at scale
By Steven Van ·
The model handles occlusions and bad lighting, and launched the same day inside Magnific's Speak and on fal.
VEED turned its lip-sync technology into a platform play. The new VEED Lip Sync 2.0 API is built for producing endless versions of a single video, change the hook, fix a line, or dub into another language, all at scale and programmatically. It's built on VEED's most powerful model to date, and it launched with Magnific as a partner.
Built for the hard footage
What separates a production-grade lip-sync model from a demo is how it handles messy real-world video, and VEED called out exactly the cases that break lesser models: objects blocking the face, low lighting, fast camera movements, and multiple angles. A model that stays convincing through a hand passing over the mouth or a quick whip-pan is one you can point at real footage, not just clean studio shots. That robustness is the whole value proposition of a 2.0.
One video, endless versions
The framing, one video, endless versions, is the API mindset. For a creator or a product, the same source clip can spawn a dozen hook variations for A/B testing, corrected lines without a reshoot, and dubs across markets, each generated on demand. That's a different use case than a one-off edit in an app; it's video as a template you re-render, which is why it ships as an API for pipelines rather than only a button in an editor.
A model, distributed everywhere
VEED's launch strategy is instructive: rather than keep Lip Sync 2.0 locked in its own editor, VEED partnered out. The same model became the default in Magnific's Speak and went live on fal as a developer API, all the same day. That's how a specialized frontier model wins now, be the best at one thing (believable dubbing) and get embedded everywhere, and it puts VEED's engine in front of far more creators than its own app reaches.
Try it
The VEED Lip Sync 2.0 API is available now via VEED and its launch partners. If you produce video at volume, the test that matters is your hardest clip, one with movement, occlusion, or bad lighting, since that's what 2.0 was built to survive.