NVIDIA has announced advancements in Universal Scene Description (OpenUSD) that extend its application to robotics, industrial design, and engineering. The new services include NVIDIA NIM microservices, which offer generative AI models capable of generating OpenUSD language, Python code, and applying materials to 3D objects. These services are designed to enhance the development of digital twins and robotics by enabling more accurate virtual simulations and workflows.
The offerings include USD Code, USD Search, and USD Validate microservices. These tools allow for generating code based on text prompts, searching through extensive data libraries, and validating file compatibility with OpenUSD standards. Additional microservices in development, such as USD Layout, USD SmartMaterial, and fVDB Physics Super-Res, aim to further enhance 3D scene assembly, material application, and high-resolution physics simulations.
NVIDIA’s collaboration with companies like Foxconn and WPP illustrates the practical use of these technologies. Foxconn is employing these tools to create digital twins of factories, while WPP is incorporating them into its content creation pipeline for clients such as The Coca-Cola Company. Additionally, NVIDIA is partnering with Siemens to integrate OpenUSD pipelines into Siemens’ Simcenter portfolio, improving real-time visualization and simulation for industrial applications.
These developments are part of NVIDIA’s efforts to expand the OpenUSD ecosystem, including co-founding the Alliance for OpenUSD with Pixar, Adobe, Apple, and Autodesk. New USD connectors for robotics and streaming to Apple Vision Pro are also being introduced, broadening the scope of OpenUSD’s application.
The USD Search, USD Code, and USD Validate microservices are currently available in preview, with additional tools available through NVIDIA’s API catalog. Further details will be shared by NVIDIA’s founder and CEO, Jensen Huang, at the SIGGRAPH computer graphics conference.
