FLUX 3 Video is generally available — 1080p, 20 seconds, native audio and lip-sync in 14+ languages

Black Forest Labs made FLUX 3 Video generally available today, through its own API and select partners. When we covered FLUX 3 yesterday it was early access; this is the release.

What the announcement specifies:

Resolution is 720p and 1080p, with upscaling. Duration is up to 20 seconds in a single generation. Audio is native — generated alongside the video rather than added after.

Three input paths: image-to-video, keyframes, and video continuation, the last accepting up to four seconds of existing footage as a starting point. Continuation is the one that changes workflows, because it turns a 20-second ceiling into something you can extend deliberately rather than a hard stop.

Dialogue with lip-sync across English and its dialects, Chinese, Spanish, French, German, Japanese, Portuguese, Russian, Italian, Indonesian, Turkish, Hindi and Punjabi, with more listed. Precise lip-syncing is claimed, not merely audio in those languages.

And a draft mode: a fast, cheap preview of a prompt before committing to a full generation. Unglamorous and probably the feature that matters most to anyone paying per second.

Two things are not in the announcement, and we are flagging them because they are circulating as though they were. There is no pricing. And there is no 2K or 4K output — the page describes 720p and 1080p with upscaling. Posts announcing this release list 2K and 4K as shipped; the announcement does not.

FLUX 3 Dev, the open-weight variant, remains on the roadmap and unreleased. That matters for anyone who read "open-weight frontier video" and planned around it: today's release is an API, not weights.

Our board's FLUX 3 entry is updated from early access to generally available. We added Black Forest Labs to the registry yesterday after finding we had never tracked the lab at all — this is the first time that entry has earned its place.