Editorial illustration for Black Forest Labs Launches FLUX 3 for Images and Audio-Video
Black Forest Labs Launches FLUX 3 for Images and Audio-Video
Black Forest Labs is done treating images, video and audio as separate products stitched together. The Freiburg-based lab announced FLUX 3 today, a multimodal model that generates images or 20-second audio-video clips from a single prompt, and extends that same architecture into robotic vision and action. Unlike earlier approaches that bolt together specialized models behind a shared interface, BFL says FLUX 3 is trained jointly across all these modalities at once.
This marks the company's first public video generation model, arriving through four separate product lines: FLUX 3 Video, FLUX 3 Image, FLUX 3 Action, and an open source version called FLUX 3 Dev still to come. Video and Action are opening now through a gated Early Access program that anyone can apply for, though BFL controls who gets in. There's no API access yet, public or through partners, with Image expected to roll out over the coming weeks ahead of a wider release.
The staggered rollout puts BFL in step with how other frontier labs in the U.S. have been handling major model launches lately.
Black Forest Labs (BFL) is expanding its FLUX family beyond image generation with today's launch of FLUX 3, a multimodal frontier model trained to understand and generate images, or combined audio/video clips up to 20 seconds from a single prompt — and to extend the same underlying architecture to robotic vision and actions.
Why this matters
BFL is betting that a single backbone trained across images, video, audio and robotic vision beats stitching together specialist models, which is a real architectural claim worth watching, not just a marketing line. But the launch details matter more than the pitch. No downloadable weights, no open source license, just a promise that FLUX 3 Dev arrives "later this year" as open-weight access.
That's the same playbook BFL used with earlier FLUX releases: ship the closed flagship first, dangle the open version for developers who actually want to build on it. For founders evaluating whether to integrate FLUX 3 now, that gap matters, you're building on an API you don't control the roadmap for. Researchers curious about the joint-training approach will have to take BFL's technical blog at its word until weights are inspectable.
The robotics angle, extending the architecture to robotic vision and actions, is the detail to track over the next few months. If BFL follows through on open weights and shows real robotics use cases, this stops being another video generator and starts looking like infrastructure. Until then, it's a capable but locked box.
Related Reading
Open Source Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source SDOF Adds Two Defensive Layers via Intent Router and StateAwareDisp
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
From Archives Anthropic Beta Tests Claude Security Plugin for Terminal Vulnerability Scanning
From Archives Treasury threatens sanctions over alleged Anthropic IP theft
Further Reading
- Black Forest Labs launches FLUX 3, a multimodal model for images, video, audio, and action prediction - GlobeNewswire / Via Ritzau
- FLUX 3 | Early Access Showcase - Black Forest Labs
- FLUX, Open Research, and the Future of Visual AI — Stephen G. on unified image, video, and audio generation - Black Forest Labs
- Black Forest Labs introduces FLUX 3 as a unified architecture for visual intelligence - GlobeNewswire / Via Ritzau
- FLUX 3 community early-access discussion on native audio video generation - LinkedIn
