FlowHMR: Mocap from Monocular Video
Official implementation of a generative model that turns monocular video into realistic, physically valid 3D motion capture.
Official code for a generative model that gets physically plausible motion capture from one video. In the demo an LLM agent builds the scene, retargets the motion, sets the camera and renders. The code is Apache-2.0. I did not check the weights license.