High-fidelity human motion. Robot-ready data. Anywhere.
Motion capture has always demanded studios, markers, and controlled lighting. Our hardware is worn, mobile, and built for ordinary environments — the places robots actually need to work.
A wearable capture stack designed to leave the lab.
Motion capture has always demanded studios, markers, and controlled lighting. Our hardware is worn, mobile, and built for ordinary environments — the places robots actually need to work.
Head-mounted sensing
Egocentric vision and head pose, time-locked to every other stream. Sees what the wearer sees.
High-fidelity gloves
Per-finger joint articulation and contact events at the resolution manipulation policies require.
Body cuffs
Lightweight tracking units on arms, legs, and torso. Full-body kinematics without a capture volume.
Edge compute
On-body fusion, calibration, and compression. Hours of untethered capture, no infrastructure.
From raw sensor streams to trained behavior.
01. Acquire
Synchronized multi-modal streams: pose, articulation, contact, egocentric video.
02. Fuse
Drift-corrected sensor fusion into a canonical, physically plausible body model.
03. Label
Automatic segmentation into tasks, interactions, and object-centric events.
04. Retarget
Morphology-aware transfer onto humanoid and manipulator embodiments.
05. Serve
Versioned, queryable datasets for imitation learning and VLA model training.
From raw signal to structured dataset.
Capture
Multi-modal, on-body, real-world
Synchronize
Hardware-level temporal alignment
Structure
Labeled, indexed, versioned
Deliver
Formats built for learning pipelines