FLUX 3

FLUX 3 is a multimodal foundation model developed by Black Forest Labs for generating and editing images, videos, and audio within a unified architecture. It can produce videos with synchronized native audio up to 20 seconds long from text prompts, images, existing videos, or defined keyframes. Supported capabilities include text-to-video, image-to-video, video-to-video, video and audio continuation, multilingual dialogue, animated typography, multiple aspect ratios, and agent-controlled chaining of clips into longer sequences.

The model can also synthesize and edit images across different styles, resolutions, and aspect ratios, with support for complex prompts and multilingual text rendering. Beyond media generation, FLUX 3 includes action-prediction capabilities intended for robotics, simulation, computer interaction, and other physical AI applications. FLUX 3 is being introduced gradually through early-access APIs and private model weights, with an open-weight multimodal version named FLUX 3 Dev planned for a later release.