FLUX 3 Video on VisionX.
Black Forest Labs moves to video, with sound in the same pass.
What FLUX 3 Video is on the roster
FLUX 3 is the first video model from the studio whose image work art directors already reach for. It renders a single continuous take of up to twenty seconds, and it generates the audio alongside the picture rather than leaving you to lay sound under it afterwards.
The same graphic sensibility that made FLUX.2 a design-team favourite carries over. Where most video engines are strongest on photographic motion, FLUX 3 holds composition, type and colour discipline through a moving frame, which is what a title sequence or a brand spot actually asks for.
Text-to-video runs from a prompt. Image-to-video pins up to ten keyframes, so a storyboard can brief the shot directly instead of being described back to the model in words.
Twenty seconds, one continuous take
Long enough to carry a complete beat without cutting around a seam, rendered as one shot rather than assembled from several.
Sound rendered with the picture
Audio is generated in the same pass and can be switched off per shot. There is no price difference either way.
Ten pinned keyframes
Hand it a storyboard rather than a description of one. The model works the motion out between the frames you pin.
Four rungs, 720p to 4K
Rendered at 720p and finished to 1080p, 2K or 4K by Black Forest Labs’ own video upsampler. Every rung is quoted per second before you run it.
What FLUX 3 Video actually does
Black Forest Labs’ first video model, and what it adds to a board.
Twenty seconds, one take
A single continuous shot from five to twenty seconds, long enough to carry a full beat without cutting around a seam.
Long takeTen pinned keyframes
Hand it up to ten stills in order and it works out the motion between them, so a storyboard briefs the shot directly.
KeyframesCarry a clip forward
Give it one finished clip of up to fifteen seconds and it continues the shot, up to fifteen seconds further.
ContinuationSound rendered with the picture
Audio is generated in the same pass and can be switched off per shot, with no difference in price either way.
Native audioRendered at 720p, finished to 4K
Every shot renders at 720p. The 1080p, 2K and 4K rungs are finished by Black Forest Labs’ own video upsampler, and each rung is quoted before you run it.
Upsampled rungsSix ratios, or auto
Choose 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16, or leave aspect on auto and let the model fit the frame to the content.
AspectEvery variant, exactly as the studio runs it
These rows are read live from the engine registry and the pricing engine. What you see here is what the composer quotes.
Self-serve rates run as low as $0.085 per VX at scale. The composer quotes the exact VX for your shot, references included, before you run it.
How to write for FLUX 3 Video
With keyframes attached, the frames carry the composition and the prompt carries the motion and the sound.
Pin the moments the shot has to hit
Keyframes are applied in the order you attach them. Pin the beats that matter and describe the movement between them.
Describe the sound
Audio renders with the picture, so ambience and effects belong in the prompt. Switch audio off when the cut will be scored.
Find the shot at 720p
The 720p rung is the model’s native render. Use it to find the shot, then re-run the keeper at a finishing rung.
Expect a preview model
Black Forest Labs calls FLUX 3 a preview, and its parameters are still moving. Lock a look you like down with keyframes rather than relying on it holding across updates.
What teams route to FLUX 3 Video
Where FLUX 3 earns the pick.
Title sequences and brand spots
Moving type, graphic composition and colour discipline, the territory FLUX’s image models are known for.
Storyboard to motion
Pin the approved boards as keyframes and let the model animate between them.
A take that ended too early
Continue the clip instead of re-rendering the whole shot to get the extra seconds.
Same Cast, same wallet, different look
Route a board line to FLUX 3 Video, then run the identical shot on another family and compare takes. Identity lives in your Cast, not in any one model, so switching engines never strands the campaign.