Synchronized
Visual and inertial streams share an explicit episode timeline.
ZL Capture hardware
Wearable systems built for Physical AI data.
Capture synchronized ego data from real-world human activity with hardware designed around visual, inertial, depth, pose, and training-data requirements.
Two collection profiles
Choose the signal density your program needs.
ZL Capture V1 prioritizes high-dimensional sensing. ZL Capture L1 reduces the operating footprint for portable human demonstration collection. Both feed the same processing and episode pipeline.
ZL Capture V1
High-dimensional wearable capture.
Built for programs that require dense multimodal observation, high-frequency inertial sensing, and explicit depth from the point of action.
- Vision
- Fisheye grayscale VGA x2
RGB HD x2 - Motion
- IMU 1000Hz x2
- Geometry
- Depth x1
- Pose
- 6DoF head pose x1
ZL Capture L1
Lightweight egocentric capture.
Designed for portable collection programs where natural human movement, repeatable operation, and core visual-inertial signals are the priority.
- Vision
- RGB HD x2
- Motion
- IMU 200Hz x1
- Geometry
- Synthetic depth x1
- Pose
- 6DoF head pose x1
System comparison
Match hardware to the episode contract.
| Signal | ZL Capture V1 | ZL Capture L1 |
|---|---|---|
| RGB | HD x2 | HD x2 |
| Grayscale fisheye | VGA x2 | — |
| IMU | 1000Hz x2 | 200Hz x1 |
| Depth | Native x1 | Synthetic x1 |
| Head pose | 6DoF x1 | 6DoF x1 |
| Best fit | Dense multimodal programs | Portable demonstration programs |
Beyond the device
Hardware is the first stage of the data system.
Sensor output becomes useful when timing, calibration, processing, and task context remain connected through the full episode pipeline.
Calibrated
Sensor relationships and confidence travel with the captured data.
Processable
Raw signals enter ZL Core for cleaning, slicing, fusion, and reconstruction.
Trainable
Outputs are organized into reusable task-ready Physical AI episodes.
Build with Zerolaw