Introducing SimReadyGen
Agentic Simulation Generation for Physical AI
From a text prompt to physically accurate SimReady assets.
Physical AI runs on simulation — and robot training now demands physically accurate worlds at a scale and speed traditional asset creation can't match. Today we're launching SimReadyGen, Lightwheel's agentic simulation-generation engine. Built on OpenUSD and integrated with NVIDIA Omniverse Libraries, SimReadyGen helps agents generate structured, simulation-ready assets from a text prompt.
Measured Physics, Generated at Scale
And those worlds are physically accurate, not estimated. That comes from SimReady Foundry, our measured-physics pipeline. It begins in our Physics Measurement Factory, where robotic rigs measure how real objects actually behave — including contact, friction, and dynamics — and convert them into ground-truth physical parameters for simulation.
Those measurements calibrate our physics solver directly, so simulated behavior tracks the real thing: on our fidelity benchmark, simulated force response matched measured behavior at over 99%. The result is a large, growing library of validated, physically measured SimReady assets. SimReadyGen generates on top of Omniverse Libraries — so the assets and scenes it produces carry real, measured physics, not guesses, at generation speed.
Built for OpenUSD Workflows
We integrated NVIDIA Omniverse Content Agents to support automated 3D content workflows for USD files, including material assignment, physics property classification, texture generation, and content validation. SimReadyGen is built on OpenUSD — so every asset and environment it creates can move through robotics and simulation workflows, including Isaac Sim or Isaac Lab.
Continuous Learning Infrastructure for Physical AI
SimReadyGen feeds the rest of Lightwheel's platform — RoboFinals for evaluation, RoboStack for deployment — building toward the continuous learning infrastructure for Physical AI.