This course teaches how to learn humanoid robot movements directly from human videos and deploy them on real humanoid robots. Students will start by capturing and analyzing videos of human motion, then converting those videos into structured motion datasets compatible with the Unitree G1 humanoid. The course covers motion extraction, data cleaning, and dataset formatting for whole-body control. Students will retrain these datasets into movement policies using modern robot learning techniques, visualize and debug motions in simulators such as MuJoCo or Isaac Sim, and iteratively refine performance. The course concludes with deploying trained movement policies onto real Unitree G1 hardware, bridging simulation and real-world humanoid control. By the end of the course, students will understand the full pipeline from human video → motion dataset → learned policy → simulated validation → real humanoid deployment.