CalBear turns video of people doing skilled work into robot training data, in minutes. Head-mounted, chest-mounted or third-person — any camera that sees the hands.
See how it worksSix stages run on our GPUs: blur, segment, reconstruct hands, retarget, align, score. You review the episodes; we package LeRobot, RLDS, HDF5 or Zarr.
Explore the pipelineHead-mounted, chest-mounted or third-person footage goes straight from your browser to storage.
Faces and screens are blurred first. Then segmentation, hand reconstruction and retargeting to your robot.
Every episode gets a quality score and an instruction. Keep what's good, drop what isn't.
The dataset in the formats you chose, plus an offline toolkit, in one kit with checksums.
Replay, train and evaluate on your own machine. Nothing about your robot connects to us.
Arms, bimanual rigs, humanoids and mobile manipulators. Pick the embodiment when you create a dataset; the same footage can be retargeted to another robot later without re-uploading. Joint names, limits and gripper conventions match each robot's LeRobot layout.
Dobot Magician (suction), ROBOTIS OpenManipulator-X (OMX), Koch v1.1, PingTi arm (human-scale SO-100), SO-100 arm, SO-101 arm and 16 more
ABB GoFa CRB 15000 + gripper, AUBO i5 + gripper, Dobot CR5 + gripper, Dobot Nova 5 + gripper, Doosan A0509 + gripper, Doosan H2017 + gripper and 31 more
AgileX Cobot Magic (two Piper followers), ALOHA 2 (two ViperX 300 S), I2RT YAM bimanual (two arms + linear grippers), Trossen AI Stationary (two WidowX AI arms), ABB YuMi IRB 14000 (two 7-DoF arms), OpenArm bimanual (two 7-DoF arms) and 1 more
AgiBot X1 (open-source), Booster T1 (4-DoF arms) + grippers, Generic humanoid, 4-DoF arms + grippers (Digit class), EngineAI PM01, PAL TALOS + grippers, Unitree H1 (4-DoF arms) + Inspire hands and 8 more
Hello Robot Stretch 3, LeKiwi (SO-101 on an omni base), Toyota HSR, Boston Dynamics Spot Arm, Clearpath Ridgeback + UR5e, Unitree B2 + Z1 arm and 7 more
Missing yours? A profile is a 20-line file. Request a robot and we usually add it within a few days.
Faces, badges and screens are blurred at ingest, before any model sees a frame, and every episode carries a chain of custody back to its source video.
The download includes an offline toolkit: verify the files, replay trajectories with safety limits, train with LeRobot, and evaluate on your own hardware.
Create a workspace, upload a short clip from any camera, and review your first episodes a few minutes later.