
Timeline concept marked at twenty seconds for a single AI video generation
FLUX 3 Video can generate up to 20 seconds with native audio in one pass. That number shows up in every capability list for a reason: it is the hard ceiling for a single generation. Longer stories are a chaining problem, not a slider problem.
Note: Figures follow Black Forest Labs' announcement and gated early access. TheFluxTrain support is coming soon.
Quick answer: FLUX 3 can create up to 20 seconds of video with audio in one generation. Multi-shot or minute-long pieces need connected clips, references, and continuation. There is no announced single-pass mode for full short films. Track TheFluxTrain on the FLUX 3 model page.
Related: video with audio · image-to-video · release guide
Write shots that fit inside one pass:
If a beat needs 35 seconds, split it before you prompt. Two coherent clips beat one overstretched generation.
When I ran a student animation company, we lived on shot lists for the same reason. You do not ask one take to carry a whole scene if the tool (or the schedule) cannot hold it. Generative video has the same constraint, just with credits instead of render farms.
BFL's early comparisons used 10-second clips. That is evaluation length, not the product maximum.
Announced tools for length beyond one pass:
Think like an editor: shot list first, generations second, timeline last.
| Shot | Target length | Mode |
|---|---|---|
| 1 | 6s | Image-to-video from product still |
| 2 | 8s | Text-to-video, same references |
| 3 | 7s | Continuation from shot 2 end |
| 4 | 5s | Keyframe title card transition |
| Assemble | ~26s | NLE or TheFluxTrain video editor later |
Stay honest about overlap and handles when you cut.
Shorter clips with clear references are cheaper to fix.
No. Generative continuation is not the same as cutting A-roll. You will still trim, rearrange, and mix. When FLUX 3 lands on TheFluxTrain, expect generation nodes plus editor assembly, not one button for a feature film. Status: model page.
Up to 20 seconds with audio in one pass.
Not according to the announced one-pass limit. Chain clips instead.
References are meant to help. Consistency is not guaranteed across long sequences.
Published early comparisons used 10-second, 720p text-to-video with audio.
BFL has not published a higher one-pass duration. Roadmap items focus more on access routes and editing APIs than on a new max length.
Platform limits will follow the model and product design. Watch the FLUX 3 model page.
Often yes. Many placements prefer short cuts. The limit matches that shape better than it matches a short film.