mimic-video is the pioneering work by mimic robotics that introduced the concept of Video-Action Models (VAMs). The model combines pre-trained video generation with robot action prediction, so that knowledge of the world's dynamics and behavior — acquired during large-scale video pre-training — is used to teach a robot manipulation tasks.
mimic-video is the direct predecessor of FLUX-mimic, which was subsequently developed together with Black Forest Labs on the FLUX 3 video-model architecture. The Video-Action approach allows industrial robots to be trained far more efficiently than Vision-Language-Action models based on static images.
The robot learning role describes software for training a robot's control policies and manipulation/locomotion skills using machine learning methods. It covers: reinforcement learning in simulation with sim-to-real transfer, imitation learning and learning from demonstration, training Vision-Language-Action (VLA) models, and fine-tuning robotics foundation models. It typically uses massively parallelized simulation environments (Isaac Lab, MuJoCo) to generate training data, then deploys the trained policies on a physical robot.
Robot Control denotes the role of software responsible for motion control, command execution, coordination of actuating elements and the direct operational logic of the robot.
