Beta

xEmbody explorer

From egocentric human demonstrations to robot-native trajectories across humanoid embodiments.

Raw egocentric RGB, stereo and IMU recordings become metric depth, SLAM camera poses and 3D hand poses, which drive full-body motion and dexterous hand retargeting in simulation.

Source dataset

xEmbody works with any data source that records stereo video and IMU.

3D hand keypointsHuman demo with tracked hand pose
Robot handsMotion retargeted onto the H2's hands
Stereo depthPer-pixel depth heatmap
Robot · left viewWhole-body H2 render
Robot · right viewWhole-body H2 render
0:00 / 0:00