Action camera
Chest or head mount. 4K or 1080p, 30 fps, on-device audio. The workhorse of egocentric capture.
EgocentricHand PoseTactile
Physical Intelligence · Samples
Watch it, download it, then decide.
10 sample episodes across four lines, each with the rig, the room, the task and the file list of the full download. The other four lines are cut to your spec on request.
Eight lines
Pick a line to see its samples.
Hover a card to play the capture. Lines marked on request are confidential by nature or carry per-session calibration, so their samples are cut to your spec.
Egocentric
Monocular first-person RGB with IMU from chest and head mounts, in real homes, shops and depots.
Watch and downloadDepth
RGB with an aligned per-frame depth map from stereo headsets, LiDAR phones and head-mounted depth cameras.
Watch and downloadUMI Gripper
Handheld wrist-camera gripper demonstrations with a 6-DoF trajectory and jaw width per frame.
Watch and downloadTeleoperation
Leader–follower and VR teleoperation episodes with synchronised cameras and per-step state and action logs.
Watch and downloadHand Pose
Head and wrist video with 21 measured keypoints per hand per frame, occlusions included.
Available on requestTactile
Video with a synchronised per-taxel pressure stream and contact events per finger or jaw.
Available on requestBring Your Own Rig
Your hardware and format, our crews, rooms and QC — under NDA.
Available on requestVideo Annotation
Segments, captions, masks and success labels per episode, in JSON or your schema.
Available on requestFrom brief to pilot
A sample in your format is the second thing you get after the quote, and it is cut from the same protocol the pilot will run on — so what you see here is what ships.
- 48 hrs
- Written spec and quote
- 5 days
- Sample episode in your format
- 2 weeks
- Pilot batch with its QA report
- Weekly
- Delivery cadence once the pilot clears
Capture hardware
What every line is recorded with.
Off-the-shelf where it is good enough, custom where it is not, and yours when the program calls for it. Each card says which lines the device runs on.
LiDAR phone
Chest harness or head strap. RGB plus metric depth and a timestamped IMU stream.
EgocentricDepth
Android phone
Head strap or chest mount. RGB with IMU, the cheapest way to put a wide pool on camera.
Egocentric
Custom monocular rig
A single camera module in a printed housing, built to your field of view and mount.
EgocentricBring Your Own Rig
Custom stereo rig
Calibrated stereo pair on a head strap or cap brim, for depth without a headset.
DepthBring Your Own Rig
Depth camera
RealSense-class RGB-D on a head mount, aligned depth at sensor-native resolution.
Depth
Stereo passthrough headset
Quest-class headset with front cameras, for depth capture and VR teleoperation.
DepthTeleoperation
Halo-strap headset
Pico-class headset with a rigid halo, comfortable over long teleoperation sessions.
Teleoperation
UMI-style gripper
Handheld parallel-jaw gripper with a wrist fisheye camera and on-device pose tracking.
UMI Gripper
Tactile glove
Pressure pads on each fingertip and the palm, logged at up to 500 Hz alongside video.
TactileHand Pose
Leader–follower arms
ALOHA-class bimanual rig with wrist, front and overhead cameras and per-step logs.
Teleoperation
Before an episode counts
What QC checks on every sample and every delivery.
The same ten checks, the same thresholds written into the protocol, on the sample you download and on the ten-thousandth episode.
- Framing
The task stays in frame; hands and workspace inside the protocol's bounds for the whole episode.
- Exposure and focus
No blown highlights or focus hunting beyond the tolerance set for the room.
- Motion
No camera whip, no dropped frames; IMU continuous where the rig has one.
- Sync
Every stream aligned within one frame; timestamps monotonic across cameras, state and action logs.
- Task compliance
Start and end states match the SOP; no skipped or reordered steps unless the protocol allows it.
- Duration
Inside the episode window for the task family; overruns are trimmed to the protocol, not the clip.
- Metadata
Environment, task, rig, operator id, session and consent reference present and valid on every episode.
- Privacy
No bystander faces, screens or documents without consent; blurred where the protocol flags it.
- Calibration
Intrinsics and extrinsics on file per rig and session for depth, stereo and multi-camera capture.
- Labels
Where annotation is ordered, agreement against the gold set above the threshold before the batch ships.
Need a sample in your task family, on your rig?
Send the task, the room and the format. A sample episode comes back within five working days, cut from the protocol the pilot will run on.