FLUX 3

Black Forest Labs unveiled FLUX 3, a multimodal foundation model trained jointly on images, video, and audio. For filmmakers the piece that matters is FLUX 3 Video: text-, image-, and video-to-video up to 20 seconds at 720p with native synchronized audio and dialogue.

It went generally available on August 4, and a week later BFL's head-to-head claims got independent backing: Arena placed an updated build at #2 on its text-to-video leaderboard, 16 points behind Gemini Omni Flash. Two caveats: the ranked build isn't the one you can call yet, and the open-weight FLUX 3 Dev backbone, the part that matters most to anyone running locally, still has no date.