FLUX 3
Black Forest Labs' first unified model for images, video, and audio in one architecture
FLUX 3 is Black Forest Labs’ first model to train images, video, and audio together in a single architecture instead of stitching separate generators behind one interface. From a single prompt it can produce a still image or a combined video-and-audio clip up to 20 seconds long, and the same underlying approach, which the company calls Self-Flow, extends to robotic vision and action prediction. It launched July 23, 2026 in a limited release rather than a full public rollout, continuing FLUX’s role powering generative features inside tools like Adobe Photoshop and Picsart. Pricing and access remain narrower than the rest of the FLUX line at launch, with Black Forest Labs prioritizing existing API and platform partners before opening it more broadly.