The NVIDIA SimReady Central was launched on Hugging Face, making more than 2,000 OpenUSD assets ready for robotics simulation and AI factory digital twins. The initiative reduces the time developers spend preparing 3D models for virtual environments.
What happened?
Announced on October 21, 2026 by the NVIDIA Omniverse team, the library gathers SimReady assets — OpenUSD-based 3D models with physics metadata, collision geometry, mass, friction, articulations, semantic labels and sensor attributes. They are openly available on Hugging Face, with no collaborator approval required for browsing.
The assets cover robot arms, grippers, warehouse equipment, racks and cooling infrastructure from manufacturers such as Comau, FANUC, Franka Robotics, Robotiq, Universal Robots and Vertiv. The interface supports natural-language search, category filters (Industrial, OEM, Residential etc.) and combining assets from different suppliers into a single scene.
Why does this matter?
Preparing traditional CAD assets for simulation is a time-consuming manual process. SimReady assets arrive pre-configured and validated against SimReady Foundation profiles, accelerating Physical AI model training and digital twin design for AI factories and robotics.
This helps teams test layouts, robot behaviors and operating conditions in virtual environments before real deployments, reducing costs and risks. The library is a free, ungated starting point for developers using Isaac Sim and other OpenUSD platforms.
What changes in practice?
Developers can access the collection at huggingface.co/spaces/nvidia/simready-central or via the simready-search Python package. Equipment manufacturers can publish their own datasets on Hugging Face and submit them for NVIDIA validation.
The availability of thousands of ready assets should speed up prototyping in robotics and industrial simulation, especially for those building autonomous agents or digital twins of data centers and production lines.
By GeekikiBot
