Introducing Flux 3: One AI Model for Video, Image and Native Audio

Black Forest Labs announced Flux 3 on July 23, 2026, and it changes the shape of AI media generation. Instead of one tool for images, another for video, and a third for sound, Flux 3 is a single unified model that produces all three — natively in sync. On flux-ai.video you can start using Flux 3 for both flux image and flux video generation today.

One model, three modalities

Most creators still juggle separate generators. Styles drift between tools, characters change face from shot to shot, and audio never quite lands on the frame it should. Flux 3 is trained differently: image, video, and audio are learned together in one Self-Flow backbone. A still image, a 20-second clip, and its soundtrack all come out of the same model, so they actually match.

That unification is the headline. Flux 3 isn't chasing one more point of video quality — it's collapsing an entire multi-tool workflow into a single prompt.

What Flux 3 can do

• Text, image and keyframe to video. Turn a prompt, a still image, or a keyframe into flux video up to 20 seconds long in a single generation. Cameras, lighting, and motion are handled automatically.
• Native audio, in sync. Flux 3 generates dialogue, sound effects, music, and ambience timed to on-screen events — a closing door or a footstep lands exactly when it should, because sound and picture are trained jointly.
• Multilingual dialogue. Give a character lines in multiple languages and Flux 3 speaks them with matching motion.
• Photoreal images with sharp text. The same model renders flux image stills across a huge range of styles, with high-accuracy typography in multiple languages.
• Consistent characters. Reference images keep the same character and style steady as you chain clips into longer scenes.

Early access outputs render at 720p across a wide range of aspect ratios, with faster and higher-resolution options rolling out over time.

How it stacks up

In Black Forest Labs' early preference tests, Flux 3 was preferred over Luma Ray 3.2 in 93% of comparisons, Runway Gen-4.5 in 77%, and Kling v3 Pro in 60%, and ran close to Seedance 2.0 and Gemini Omni Flash. Those evaluations are early and evolving — but the direction is clear. Kling still leads on human motion and Runway on post-production control; Flux 3's edge is doing image, video, and physically-aligned audio from one model instead of stitching separate tools together.

Try Flux 3 free

You can start with Flux 3 on flux-ai.video for free — sign in to get free credits, describe your shot, and generate a flux image or a flux video with native audio. Refine the result, extend the clip, or continue it while keeping your character consistent, then download it.

Flux AI is an independent web interface for Flux 3 and is not affiliated with Black Forest Labs. For the official model announcement, see the Black Forest Labs blog.