what we capture
Dexterous manipulation
Hands doing real work — picking, placing, tool use, two-handed coordination. The fine-motor data simulation can't fake and labs struggle to source at volume.
REC 02:14:08
consent-clean
manipulation
what you get
captured to spec, audited, and delivered to your schema.
Contact-rich interaction
Grasps, regrasps, tool use, and two-handed coordination — the contact events manipulation policies need.
Fine-motor detail
Close, hands-visible framing so keypoints and object pose can be recovered from real footage.
Hard to source, at volume
The data labs struggle to gather — captured to spec by trained contributors, not scraped.
Where it's used: Robot grasping and manipulation, dexterous policies, imitation learning, teleop bootstrapping.
get started
tell us what your models are missing.
We'll scope a capture program and show you sample data — usually within a week.