Demonstrate naturally
Move naturally, without markers or suits.
Robotics & Physical AI
A unified 3D ground truth capture stack.
RayKnot transforms human-demonstration data collection by capturing body motion and hands in one single synchronised 3D world.
No markers, mocap suits or VR headsets on human subjects. Demonstrators move naturally while RayKnot produces structured, time-aligned data for the purpose of annotation, retargeting, simulation and robot learning.
Robot learning has a data productivity problem.
Fragmented capture systems add setup, synchronisation and annotation work. RayKnot unifies movement data in one time-aligned 3D pipeline.
One seamless data pipeline.
Capture the whole scene
Capture humans, hands and objects together.
Rich ground-truth capture.
Reconstruct and align.
Build synchronised 3D motion and interaction data.
Efficient downstream integration
Reduces manual labelling and need for cleaning large volumes of low-information data
Expand with grounded synthetic data.
Expand datasets using grounded synthetic variations.
One capture. Multiple possibilities.
Demonstration capture
Collect task demonstrations in natural movement.
Annotation-ready datasets
Prepare structured data for review and labelling.
Enable automated cleaning and annotation
Increase productivity in downstream data preparation
Dataset expansion
Expand datasets from measured movement.
More useful data from every demonstration
-
Full body to fingertips
Full-body and hand detail in one capture. Rich ground truth capture.
-
Human and object together
One shared 3D view of interactions.
-
Unconstrained capture.
No markers, suits or VR headsets.
-
Structured, time-aligned outputs
Structured data for downstream workflows
Transform the productivity of demonstration-data collection
Tell us your robotics workflow. Explore a capture setup for reusable training data.
Book a Demo