Integrate NVIDIA Omniverse RTX Sensor Simulation Into Existing Apps

At SIGGRAPH 2026, NVIDIA announced that the Omniverse libraries are now part of the NVIDIA Agent Toolkit, with ovrtx serving as the RTX sensor simulation layer, providing modular APIs that allow developers to embed physically accurate sensor simulation into existing applications. The core principle of ovrtx is to let the host application maintain control over its architecture, UI, data model, and workflow, while leveraging NVIDIA RTX's RT Cores and Tensor Cores for real-time, physically accurate rendering and sensor output.

ovrtx supports generating outputs for camera, lidar, radar, ultrasonic, semantic segmentation, and more from OpenUSD scenes. A typical integration flow includes: creating an ovrtx renderer, loading or referencing OpenUSD content, defining RenderProducts and RenderVars, stepping the simulation from the application loop, and mapping results to CPU or GPU memory for display, storage, or analysis. Pseudocode examples show that the host application owns the main loop, and ovrtx generates outputs on demand.

ovrtx runs as a lightweight C/Python SDK, relying on the Omniverse RTX renderer, which uses RT Cores for ray tracing and Tensor Cores for AI acceleration to achieve physically accurate light transport simulation. In the Omniverse library stack, ovstage provides a shared runtime stage that manages OpenUSD scene data, allowing libraries such as ovphysx (physics simulation) and ovrtx to exchange simulation state and scene changes through a common data substrate.

This architecture supports multimodal simulation pipelines for synthetic data generation, digital twins, and physical AI validation. Developers can integrate ovrtx into CAD, Blender, PLM/PDM workflows, as well as cloud-native design processes such as PTC Onshape Render Studio.

Target applications for ovrtx include design, robotics, autonomous vehicles, industrial digital twins, and physical AI. Developers need to visualize OpenUSD scenes, generate physically accurate data, test perception systems, validate behaviors, and view results within existing tools. ovrtx's modular APIs make these capabilities embeddable into CAD systems, Blender workflows, PLM/PDM workflows, OpenUSD asset pipelines, and internal simulation stacks.

The article showcases two integration examples: PTC Onshape Render Studio and SimReady Blender workflows. These examples illustrate how to incorporate ovrtx's sensor simulation capabilities into existing design processes without restructuring the application architecture.

ovrtx is currently available as pre-release software on GitHub. Developers can inspect the C and Python APIs, run examples, test OpenUSD scene loading, and evaluate how RTX sensor outputs fit into their own application loops. NVIDIA emphasizes that this library-first approach gives teams more flexibility in controlling where simulation capabilities run within their own tools and services.