Flux 3 T2V is native audio to 20s
Flux 3 T2V is 34 credits for 5s, up to 136 at 20s, native audio. Black Forest Labs GA on 4 Aug 2026. One row, one job.
Flux 3 Text to Video is 34 credits for 5 seconds and 136 for 20, about 7 credits a second, with native audio. Black Forest Labs called FLUX.3 Video generally available on 4 August 2026. One row, one job: a text prompt in, a clip with sound out, up to twenty seconds.
The catalog description: FLUX.3 is Black Forest Labs’ frontier video model — video with native audio directly from a text prompt, at up to 1080p and 5–20 seconds across eight aspect ratios. Category is text-to-video. Audio is true. Credits start at 34 and run to 136, about 7 a second. Durations are every second from 5s through 20s. Resolution on the card is unset; the description’s ceiling is 1080p. Do not invent 4K for this slug.
BFL’s own early evals (23 July 2026, cited from BFL, not as Versely benchmarks) reported preference against Grok Imagine Video 69%, Kling v3 Pro 60%, Happy Horse 1.1 57%, Seedance 2.0 and Omni Flash 52%, Runway Gen-4.5 77%. Those are vendor evals with a date. They are not a Versely leaderboard and not a reason to skip reading the duration menu.
What the T2V row actually sells
A sentence becomes a take. Sound is born with the picture. Five seconds is the floor. Twenty is the ceiling. That combination is why this row shows up when Sora 2 T2V’s 12 / 16 / 20s ticks lose their API on 24 September 2026.
Native audio means you write the soundtrack in the prompt: room, line, foley, or an explicit no-speech instruction. A mute prompt still returns audio you did not choose. Optional in marketing copy is not optional in the file.
What this row is not:
- Not first-last. Flux 3 first-last is 34 credits, 5–20s, native audio, requires frames. Use that when two stills are the contract. Do not paste a SKU JPEG into a T2V prompt and call it locked.
- Not extend. Flux 3 extend is 82 credits, 5–20s, native audio. Use that when the last frame is already a keeper. T2V does not continue a file.
- Not Veo’s 4K dialogue ticket. Veo 3.1 is 64 credits, 4 / 6 / 8s, 4K. Flux 3 T2V wins the 9–20s band. Veo wins the short 4K talking plate.
- Not a silent draft. If you want audio false, Kling 3 Turbo is 22 credits, 3–15s, 1080p. Flux 3 T2V will make sound.
Thirty-four credits is the catalog sticker for a 5-second clip, and the row bills per second, so a 20-second take is 136. Black Forest Labs also publishes USD rates on its own site; those are BFL’s prices, and Versely bills credits.
The AI video generator is the run door. This page is the job description.
How to write a 5–20s native-audio prompt
Treat twenty seconds as a camera hold, not as five ads.
Lock the subject, the verb, the camera, the light, and the sound. If the prompt is a shot list with commas, the model will try to play all of it in one generate. That is how identity thins after 0:12.
Sound is a first-class paragraph:
- One speaker or none.
- The words, in quotes, if anyone talks.
- Room tone: street, kitchen, quiet studio.
- Music: none, or a specific bed you will replace anyway (if you will replace it, say no music).
Duration ticks are every second. Pick 8s if the action is 8s. Padding to 20s is how hands smear. Sora users who always clicked 20s should not always click 20s here.
Aspect: the description names eight ratios. Pick the delivery ratio before generate. A 9:16 prompt rendered 16:9 is a crop, not a feature.
Inspect the first keeper at 8–12s before you spend a 20s take on the same brief. Flux 3 bills per second, so a 20s take costs 136 credits against 34 for a 5s draft, and the time lesson is on top of that.
Do not prompt “in the style of Sora.” Prompt the shot. Sora 2 T2V was 32 credits, 720p, 4 / 8 / 12 / 16 / 20s. Flux 3 T2V is the 5–20s native-audio landing for that length habit, not a clone of the vendor look.
When not to use this row
The still is the product. T2V invents. Use first-last, Seedance R2V, or Veo R2V. If the JPEG is the brief, leave.
The talking close-up must be 4K and 8 seconds or less. Veo 3.1. Flux 3 T2V can talk; it is not the 4K Veo ticket.
You need 30 seconds in one pass. Seedance 2.5 is 34 credits (was 67), 4–30s, 720p, native audio. Grok Imagine Video is 24 credits to start (up to 168), 30s tick, 720p, native audio. Flux 3 T2V stops at 20.
You need 4 seconds. Flux 3 T2V’s floor is 5s. Veo has 4s. Seedance 2.0 / 2.5 have 4s. Kling 3 Turbo has 3s and 4s, silent.
You are drafting composition and will discard sound. Turbo, 22 credits, audio false. Paying for native audio you will mute is the Sora habit this row should not inherit.
You need a listed 4K native-audio draft. LTX 2.3 Fast starts at 20 credits, billed per second, 6–20s, up to 4K, native audio. Different family, different failure modes. Use LTX when listed 4K is the point; use Flux 3 T2V when this family’s 5–20s native-audio generate is the point.
Caption after you like the take. Native audio does not save you on a muted feed. Add captions to video.
After Sora, this is the 20-second native-audio T2V
The Sora 2 T2V card still lists 20s today. The API clock is 24 September 2026. Teams that lived on 16s and 20s talking-or-ambient takes need a living ceiling. Flux 3 T2V is that ceiling, from 34 credits, with native audio.
It is not the only ceiling. Seedance 2.5 and Grok Imagine Video go to 30s at 720p. LTX 2.3 Fast goes to 20s at up to 4K, from 20 credits. The reason this page is only Flux 3 T2V is so those other rows stay other pages.
If you must continue a Flux 3 keeper instead of rolling a new T2V, that is extend at 82 credits — a different slug. If you must travel between two approved frames, that is first-last at 34 credits — a different slug. Do not ask T2V to do either.
Run the 5–20s native-audio generate from Versely.
FAQ
Is Flux 3 T2V the same as Flux 3 extend or first-last?
No. T2V is 34 credits, 5–20s, native audio, text only. First-last is 34 credits, 5–20s, native audio, two frames required. Extend is 82 credits, 5–20s, native audio, continues an MP4. Three slugs, three jobs.
Did Versely measure Flux 3 against Kling and Seedance?
No. The preference percentages above are Black Forest Labs’ early evals dated 23 July 2026, not Versely benchmarks. Read duration, audio, and credits on the catalog card.
Can I use Flux 3 T2V as a drop-in Sora 2 4-second hook?
Not at 4 seconds. The floor is 5s. Use Veo 3.1’s 4s tick, Seedance’s 4s ticks, or Turbo’s 3–4s silent ticks. From 5s through 20s, Flux 3 T2V is in the map.