Robotics and physical AI
Characters your models have never seen.
Varied 3D humans for perception training and testing, specified by you and generated inside the simulation you already use.
The problem
Robots, drones and vehicles must recognize people, and they fail on the ones they rarely see: wheelchair users, unusual clothing, poor light. Real capture is slow and privacy-bound, and generated video cannot repeat a scene or stay inside a secure network.
What you get
- Populations specified by ethnicity, age group, body type, clothing and mobility aids
- Same input, same people, so test sets are reproducible
- Full 3D, rigged characters that move, not flat images
- Thousands of characters per run, on your own hardware
- No third-party models in the core engine. On-premises available for enterprise.
Pilot program open
Become a first partner. Help us build these.
- Custom ground truth per character: attributes, measurements and joints
- Coverage reports: what each dataset contains against what you asked for
- Simulation-native output for the tools and engines your team already runs
- Your own motion capture applied across a whole population at once
Tell us the people your system must handle.
A six-week pilot puts a specified population in your simulation, with one success metric agreed up front.
‹ All use cases