Black Forest Labs Launches Flux 3 Multimodal Model, Powers Audi Assembly Robots via Mimic Robotics
AI video model releases
Sign in to create alerts.
Black Forest Labs Launches Flux 3 Multimodal Model, Powers Audi Assembly Robots via Mimic Robotics
Black Forest Labs released an early version of Flux 3, a foundation model jointly trained on images, video, and audio from the start, unlike Flux 1 and Flux 2 which only generated images.
Flux 3 Video supports text-to-video, image-to-video, and video-to-video generation with clips reaching about 20 seconds and native synchronized audio in a single pass, currently in invite-only early access.
Partnering with Zurich robotics startup Mimic Robotics, Black Forest Labs built Flux-mimic, a video-action model now deployed on robots at Audi production lines for soft-body assembly tasks previously resistant to automation.
During training, adding robot action prediction caused human quality ratings on video tasks to drop up to 10%, but the model recovered full quality after 3,500 training steps while also predicting actions.
A separate Flux 3 Image component is planned for rollout in the following weeks, and an open-weight developer version called Flux 3 Dev is planned for later in 2026; all deployment and performance claims are self-reported by Black Forest Labs.
cho.sh