Monocular video motion capture that combines flow matching with physics-based RL for plausible 3D human motion.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
FlowHMR is a research model for recovering 3D human motion from monocular video. It combines flow matching with physics-based reinforcement learning (GRPO) so that the recovered motion is both faithful to the video and physically plausible. The project page is titled "Physically Plausible Motion Capture from Video" and an arXiv preprint (2026) is cited.
The project page shows applications: retargeting captured motion to a Unitree G1 humanoid robot (using GMR, with SONIC tracking the motions in physics simulation), animation creation with scene reconstruction and retargeting, and text-to-motion by generating a video from a prompt and then capturing the motion from it. A YouTube creator summary says code, models and a Hugging Face demo were released, but the provided pages did not include those links, so they are not listed here.
It is aimed at researchers and developers working on character animation, motion capture and humanoid robotics. The vendor and licence are not stated in the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Reconstructs physically realistic 3D human body motion from regular video, usable for humanoid robot transfer and character animation. Code, models and a Hugging Face demo released.
Closest neighbours by meaning, not just by category.
Research
Research framework that retargets human motion to humanoid robots using learned point cloud correspondences.
Model
Text-to-human-motion flow-matching model that generates full-resolution motion directly, with training-free joint constraints.
Research
Diffusion framework that learns text-to-3D human motion generation from 2D poses extracted from monocular videos.
Model
Research method that recovers 6-DoF object and camera pose from monocular video via proxy-video translation and PnP.
Library
Training-free staged sampling method that speeds up pretrained flow-matching text-to-image diffusion models.
Model · Nvidia
Feed-forward model that repairs lifted 2D tracks and depth into accurate dense 3D trajectories from monocular video.