A gripper in a human hand, in a room a robot has never seen.
Handheld UMI-style grippers with a wrist-mounted fisheye camera let people demonstrate grasps anywhere — kitchens, shops, workshops — without a robot in the room. Manipulation data with the diversity a lab cannot stage.
- Environment
- Lived-in kitchen counter
- Captured with
- Handheld UMI-style gripper, wrist fisheye camera
- Task
- Pick a tomato from a bowl, set it on the board
Perit has shipped 4,000+ hours of speech through this bench. Physical-AI capture runs on the same recruiting, consent and QA machinery, and the first collections are being scoped with early-access partners now. Nothing on these pages is a delivered volume — the first hours will be listed here the way our speech hours are, after they ship.
What gets captured, and why it is worth having.
The gripper is the robot's end-effector, carried by a person. What the wrist camera sees is what a policy will see; what the on-device tracking records is where the gripper was. Because no robot is needed at capture time, the pool is wide and the rooms are real.
Single-hand and bimanual variants, 2-finger and custom jaws. Episodes are short and dense — a grasp, a carry, a place — and the diversity across sites is the point.
- Grasp policies that generalise across scenes
- Bimanual coordination
- Object-centric manipulation priors
- Cross-embodiment transfer — human hand to robot gripper
- Primary rig
- UMI-style handheld gripper, single or bimanual
- Also
- Wrist-mounted fisheye action camera
- Also
- On-device pose tracking (visual-inertial)
- Also
- 2-finger and custom jaw variants
- Video
- H.264 in MP4, fisheye
- Resolution
- 4K sensor-native or 1080p
- Frame rate
- 30 fps
- Pose
- 6-DoF end-effector trajectory per frame
- Gripper
- Opening width per frame
- Episode
- 10 seconds – 2 minutes
Where it is recorded.
What people are asked to do.
Tasks are recorded where they actually happen — not in a studio dressed to look like a kitchen. Hover a room to see the kind of setting the protocol calls for.
- Fisheye video per episode
- End-effector trajectory and gripper width per frame
- Task and environment metadata per episode
- QC status and reviewer chain per episode
- Consent reference on every file
- Quality report per batch: accepted, rejected, and why
- Weekly drops to your bucket, region pinned at kickoff
- Grasp success labels
- Object category per episode
- Contact events
- Trajectory smoothing and re-timing
Labelled on the same bench as our speech work, to a written guide, behind the same calibration test.
Small first, on purpose.
- 01Briefday 0
The task family, the environments, the rig, the sensors, the episode count. One page, agreed before anything is bought or anyone is recruited.
- 02SOP and kitweek 1
A written capture protocol — framing, lighting, where a task starts and ends, what counts as a failed take — and the kit list. Operators are trained and tested on it before a single episode is paid for.
- 03Pilot episodesweek 2
A small batch, ten to thirty episodes, delivered with QC status and a quality report. This is where the protocol breaks — on purpose, and cheaply.
- 04Review and scaleongoing
You review the pilot, we fix the protocol, then the run scales on a weekly delivery cadence with the same report attached to every batch.
Scope a umi gripper pilot.
The task family, the environment, the rig and the episode count — one page, and we come back with a protocol and a quote.