FLUX 3 Video — 20 Seconds With Native Audio, If the Claims Hold
Gated early access. No public API, no published pricing, no independent test.
FLUX 3 Video is the video model in Black Forest Labs' FLUX 3 family, announced July 23, 2026. BFL claims 20-second single-pass generation with native audio across five modes. This page lays out those specs alongside their verification status — and shows what covers the same jobs today, without an approval queue.
Official Video
Watch: FLUX 3, as Black Forest Labs Showed It
The announcement video Black Forest Labs published on July 23, 2026. Every clip in it was rendered by BFL — none of it was generated on ImagineToVideo, and none of it has been independently reproduced.
FLUX 3 Video Specs — What BFL Claims vs What Runs Today
20-Second Clips in One Generation (Claimed)
Black Forest Labs states FLUX 3 Video produces up to 20 seconds in a single pass. Seedance 2.5 already runs to 30 seconds in one pass on this site, so 20 seconds would not be a record. ⚠️ Vendor claim; the model is not publicly released, so no third party has measured it.
Native Audio, Multilingual Dialogue (Claimed)
Audio is generated with the picture rather than dubbed on afterwards, with multilingual dialogue and sound matched to on-screen events. Published evaluations were run at 720p. ⚠️ Vendor claim, not independently reproduced.
Five Generation Modes in One Model (Claimed)
Text-to-video, image-to-video (animation or reference), video-to-video, video-audio continuation and keyframe-to-video, plus agentic chaining for longer sequences — all from one backbone BFL calls Self Flow. ⚠️ Official capability list, not a verified one.
Audio-Native Video You Can Run Today
No application, no queue: Wan 3.0 (3–30s, 480p/720p/1080p), Seedance 2.5 (4–30s, 480p/720p, video-to-video included), Kling V3 and Kling V3 Omni (3–15s, 720p/1080p), VEO 3.1 and VEO 3.1 Fast (4/6/8s, 720p/1080p) and Seedance 2.0 (4–15s, 480p/720p/1080p). All with native audio, all live on ImagineToVideo.
What FLUX 3 Video Is Built For — And What Delivers It Now
Spoken-Dialogue Spots
The headline use for native audio: one prompt, picture and voice together. VEO 3.1 does this today at 4, 6 or 8 seconds — the standard length for a paid social cutdown, with no waiting list.
Long Single Takes
20 seconds in one generation is FLUX 3 Video's most quoted number (⚠️ untested). Seedance 2.5 already runs up to 30 seconds in a single pass at 480p/720p, and Kling V3 gives you 3–15 seconds per take at up to 1080p — so single-shot briefs are covered today without stitching.
Multi-Shot Sequences
BFL points to agentic chaining for anything longer than one generation. Kling V3 Omni already chains connected shots on ImagineToVideo while holding the same character across them.
Reworking Existing Footage
Video-to-video and video-audio continuation are both on FLUX 3 Video's official mode list. Seedance 2.5 runs video-to-video today — feed in footage, a reference image or a last frame, get up to 30 seconds back.
How to Work With FLUX 3 Video
Join the queue, and keep shipping while you wait
Apply for FLUX 3 Video Early Access
FLUX 3 Video sits behind an application-gated program at Black Forest Labs. Submit the request, then keep working — there is no endpoint to call until BFL approves you, and no published pricing to budget against.
Write the Audio Into the Prompt
Describe the shot, the spoken lines and the ambient sound in one prompt block. That is how every audio-native video model reads input — the ones running today, and the one FLUX 3 claims to be — so the prompt survives the switch.
Match the Model to the Job
Spoken dialogue in 4, 6 or 8 seconds: VEO 3.1. Camera movement and a 15-second single take: Kling V3. A single take up to 30 seconds, or reworking footage you already have: Seedance 2.5, which is also the model that runs video-to-video. All three sit in one panel at /video/text-to-video.
Draft at 720p, Finish at 1080p
Lock motion and framing on a cheap 720p pass, then re-render only the take that works. Chain shots in /video/multi-scene when you need more than one. When FLUX 3 Video opens up, re-run the same prompt and compare side by side.
FLUX 3 Video FAQ
What is FLUX 3 Video?
FLUX 3 Video is the video generation model in the FLUX 3 family that Black Forest Labs announced on July 23, 2026. It is trained jointly with image and audio on a shared backbone BFL calls Self Flow, and BFL states it generates audio natively alongside the picture rather than as a separate dubbing pass.
How long can a FLUX 3 video be, and what resolution?
Black Forest Labs states up to 20 seconds in a single generation, with published evaluations run at 720p. No other resolution has been disclosed, and neither has frame rate. ⚠️ Both figures are BFL's own — the model is not publicly available, so nobody outside the early access program has been able to verify either one.
Does FLUX 3 Video really generate sound?
That is the claim. BFL says audio is generated natively with the video, including multilingual dialogue and sound matched to physical events on screen, and lists video-audio continuation as a supported mode. ⚠️ Untested outside BFL. If native audio is what you need today, VEO 3.1 and Kling V3 already do it on ImagineToVideo with no waiting list.
What generation modes does FLUX 3 Video support?
The official list is text-to-video, image-to-video (as animation or as reference), video-to-video, video-audio continuation and keyframe-controlled transitions, with agentic chaining to build sequences longer than a single generation. ⚠️ This is the capability list BFL published, not a set of behaviours anyone has independently exercised.
How does FLUX 3 Video compare to Kling, VEO, Runway and Luma?
Black Forest Labs published preference win rates of 69% against Grok, 60% against Kling v3 Pro, 77% against Runway Gen-4.5 and 93% against Luma Ray 3.2. ⚠️ These are vendor self-reported results, not independent blind tests — no evaluation protocol, sample size, rater count or prompt set has been published. Until FLUX 3 Video is generally available, there is no way to run the comparison yourself.
Can I use FLUX 3 Video on ImagineToVideo?
Not today. FLUX 3 Video is in gated early access with no public API and no published pricing, so it is not in our model registry. We plan to evaluate integration once the official API and pricing are public, benchmark the real specifications against the claims, and then price it in credits — the timing depends on Black Forest Labs, so we are not committing to a date. Wan 3.0, Seedance 2.5, Kling V3 and VEO 3.1 are available right now with native audio.
Audio-Native Video, Without the Application Form
FLUX 3 Video is gated with no API and no pricing. Wan 3.0, Seedance 2.5, Kling V3 and VEO 3.1 generate video with native audio today — up to 30 seconds per clip on Seedance 2.5, and 720p and 1080p on Kling V3 and VEO 3.1.