Mimic Robotics Begins Production of FLUX-mimic at Audi

(CONNECT) Mimic Robotics has introduced FLUX-mimic, a next-generation video action model. The Zurich-based company specializing in physical artificial intelligence (AI) developed it in collaboration with Black Forest Labs from Freiburg im Breisgau. It is already being tested and implemented at the German automaker Audi.

According to a press release, FLUX-mimic combines mimic's expertise in the fields of robot learning, precise manipulation, and production implementation with the visual intelligence architecture of Black Forest Labs' new FLUX 3 video model. With mimic-video, mimic had broken new ground and introduced innovative video-action models—that is, AI systems that link visual data, such as videos, to specific actions or movements.

This is the foundation of FLUX-mimic: It uses a generative video model that already understands dynamics and behavior through extensive pre-training on video data. Next, an action decoder is trained to predict robot actions directly from this visual prediction. According to mimic, this approach allows the model to be fine-tuned for a specific manipulation task using just 30 minutes of robot data, whereas earlier approaches required 30 hours or more. This is intended to shorten implementation cycles from months to weeks.

At Audi, these robots have already performed complex manipulation tasks “that would have been simply impossible with conventional robotics,” according to Christoph Schneider of the Audi Production Lab. Mimic co-founder and head of product Stephan-Daniel Gravert describes Audi as “exactly the manufacturing partner for whom we developed FLUX-mimic.” Their production environments require automation that is flexible enough to handle unstructured tasks, reliable enough for continuous operation, and can be integrated without months of development work. Through this partnership with Audi, we aim to demonstrate that FLUX-mimic meets these requirements in real production lines.” ce/mm