NVIDIA's nanousd-labs initiative is using artificial intelligence to simplify the creation of custom USD runtimes, a move aimed at advancing the field of physical AI and broadening the adoption of the OpenUSD framework. The technology targets developers who need lightweight, tailored runtimes for 3D scene data, a key component in robotics, simulation, and autonomous systems.
Why Custom USD Runtimes Matter
Universal Scene Description, or USD, is an open framework for describing 3D scenes. It's used across visual effects, gaming, and increasingly in industrial applications. But running USD efficiently on different hardware often requires custom runtimes — stripped-down versions optimized for specific processors, sensors, or latency requirements. Building those runtimes by hand is time-consuming and error-prone. nanousd-labs applies AI to automate parts of that process, letting developers generate optimized runtime code faster.
Physical AI Gets a Boost
Physical AI refers to systems that perceive, reason about, and act in the physical world — think robots, self-driving cars, and factory automation. These systems rely on accurate 3D representations of their environment, often built with USD. By streamlining custom runtime development, nanousd-labs helps engineers deploy physical AI applications more quickly. The AI-driven approach can adapt runtimes to the specific constraints of embedded hardware or real-time control loops, cutting development cycles.
OpenUSD Adoption Gains Momentum
OpenUSD is an open standard stewarded by the Alliance for OpenUSD. Wider adoption depends on making the framework accessible across different platforms and use cases. nanousd-labs lowers the barrier for companies that need custom runtimes but lack the resources to build them from scratch. The AI tooling means more organizations can integrate USD into their workflows without deep expertise in the underlying runtime architecture. That, in turn, strengthens the ecosystem around OpenUSD.
The initiative is part of NVIDIA's broader push to position OpenUSD as the foundation for physical AI applications, from digital twins to autonomous machines. No release date for the nanousd-labs tools has been announced, but the project is already drawing interest from developers working on real-time robotics and simulation pipelines.
