Black Forest Labs drops FLUX 3 Video with native audio

Anyone who has tried to make an AI video recently knows the headache: you prompt a cool clip, export it, generate voice lines somewhere else, hunt down ambient audio, and spend an afternoon trying to line up the lips.

Black Forest Labs just rolled out FLUX 3 Video through its API and select partners, and it wants to handle that entire pipeline in one shot.

Sound, lips, and multi-shot directing

The marquee spec is 20-second generation right out of the box, natively rendering in 720p with an upscale option to 1080p.

Instead of treating audio as an afterthought, FLUX 3 creates native audio right alongside the frames. That includes ambient sounds, sound effects, and dialogue in over 14 languages, complete with matching lip-syncing across different dialects.

It also brings serious camera control. You can generate from text or images, set start and end keyframes, or string together multi-scene shots with shifting camera angles inside a single output.

There is also a video continuation feature: feed it up to four seconds of existing footage and sound, and it will keep the motion, camera, and dialogue flowing across the cut.

To save creators from burning credits on blind rolls, BFL added a Draft Mode. You can spin up quick, cheap low-res previews to lock down your motion and framing, then hit render on the full-quality version with the exact same subjects and movement.

According to BFL's internal human evaluation, FLUX 3 beats existing state-of-the-art models in text-to-video, and ties Seedance 2.0 while beating the rest of the field in image-to-video.

Open-source fans will also want to keep an eye out: BFL confirmed that an open-weight FLUX 3 Dev model and a FLUX 3 Image model are coming down the pipeline.

Why it matters

We think this is the clearest sign yet that the silent film era of generative video is officially over.

For creators and small production teams, bundling synchronized dialogue, audio, and camera direction into a single 20-second block fundamentally changes turnaround times. Draft Mode is a very smart addition, taking the costly guesswork out of video prompting before you commit API credits.

For the industry, BFL is shifting the goalposts. Competing purely on visual fidelity won't cut it anymore; unified audio-visual generation is quickly becoming table stakes.

The real catch will be fine-grained editability. While generating everything simultaneously sounds great, fixing a single weird background sound or slightly mismatched tone inside an integrated render can be trickier than tweaking modular layers. And until the open-weight FLUX 3 Dev actually ships, developers remain locked into API pricing.