NVIDIA Omniverse
NVIDIA Omniverse is a set of GPU-accelerated libraries, APIs and services for building physically based 3D simulations and digital twins on OpenUSD data. NVIDIA states it has been free for development, production and redistribution since May 2026; enterprise support needs an AI Enterprise license.1
Also known as Omniverse libraries, Omniverse Kit SDK, Omniverse Enterprise
At a glance
- What is it?
- Omniverse is NVIDIA's toolkit for simulating physical spaces in software. NVIDIA no longer presents it as one desktop application. It now describes Omniverse as a collection of libraries and microservices: ovrtx for RTX rendering and sensor simulation, ovphysx for physics, ovstage for shared scene data, ovstorage, ovstream and ovui for storage, streaming and user interfaces, plus the Kit SDK for complete applications. All of them read and write OpenUSD scene description. NVIDIA lists the Omniverse libraries as part of NVIDIA Agent Toolkit, so AI agents can call them as tools.24
- What does it do?
- Omniverse turns 3D data into a running simulation. A team converts CAD files, robot descriptions or other 3D formats into OpenUSD, adds materials and physics, and then renders camera, lidar and radar output on RTX GPUs. Common results are a digital twin of a factory or warehouse, synthetic images for training vision models, and virtual test environments for robots and vehicles. Supporting tools convert formats, check assets against SimReady specifications, search USD content in natural language, and stream rendered output to local or browser clients.4
- Who needs it?
- Teams that build simulation software, industrial digital twins, synthetic data pipelines, or robot and vehicle test environments. Independent software vendors that want to add physically based rendering or physics to their own products. The work suits developers with Python or C and C++ skills and access to NVIDIA RTX GPUs.
- What does it need?35
- An NVIDIA RTX GPU; NVIDIA lists a GeForce RTX 3070, 16 GB RAM and 250 GB storage as the Kit minimum
- Windows 11 or Ubuntu 22.04 or 24.04
- An NVIDIA driver from a branch validated in the Omniverse technical requirements (R570, R580 or R595)
- For ovrtx: Python 3.10 to 3.13, or a C++17 compiler with CMake 3.16 or later
- 3D assets in OpenUSD or a format the converters accept (common 3D formats, URDF, MuJoCo, Gaussian splats)
- Outbound HTTPS to PyPI, GitHub and NVIDIA content hosts for installs and sample scenes
- What it is not
- Omniverse is not OpenUSD. OpenUSD was created by Pixar and is developed under the Alliance for OpenUSD, which NVIDIA co-founded with Pixar, Adobe, Apple and Autodesk; Omniverse builds on that open standard. Omniverse is also no longer a launcher with ready-made desktop apps: NVIDIA lists the Launcher, USD Composer and USD Explorer as no longer supported. Free use does not include enterprise support, and many of the new libraries are pre-release.16
Availability and licensing. NVIDIA's documentation states that from 1 May 2026 Omniverse is free for development, production and redistribution, with community support through the NVIDIA Developer Forums and Discord. Enterprise Support requires an NVIDIA AI Enterprise license, and the Omniverse License Server is no longer needed. The new libraries are marked pre-release and not enterprise-supported, and each repository carries its own license (ovrtx, for example, uses an NVIDIA proprietary license), so check each component.157
The problem it solves
Planning a factory line, a warehouse layout or a robot cell usually involves several design tools that each store 3D data in their own format. Moving data between them is slow, and testing a change on the real site costs time and interrupts operations.
Teams that train vision or robot models face a second problem: real sensor data for rare events, such as defects or near misses, is hard to collect and label.
Omniverse addresses both by putting scene data into OpenUSD and offering GPU libraries that render sensors and simulate physics on that shared data, so changes can be tested and training data generated in software first.
How it works
- Bring data in as OpenUSD. Converters turn common 3D formats, URDF robot files, MuJoCo models and Gaussian splats into OpenUSD; a CAD-to-SimReady agent skill adds geometry, materials and physics.
- Check the assets. simready-validate and usd-validation-nvidia test assets against SimReady and core USD rules.
- Load the scene once. ovstage loads USD into a runtime representation and shares state between tools through zero-copy tensors.
- Simulate. ovphysx handles USD-native physics; Newton and PhysX are available for robotics workloads.
- Render and sense. ovrtx produces camera, lidar and radar output on RTX GPUs for synthetic data and visual checks.
- Deliver. ovstream sends GPU output to local or browser clients, ovui builds inspection tools, and the Kit SDK builds complete applications.
Blueprints, such as the DSX Blueprint for AI factories, show how the parts fit together in a full workflow.4
Diagram as a list
Applications & solutions
- CAD, URDF and other 3D sourcesExisting design data that feeds the twinConnects to Converters and SimReady validation
- Kit apps, agent skills and blueprintsWhere users and AI agents work with the simulation
Models & frameworks
- Converters and SimReady validationTurn source data into checked OpenUSD assetsConnects to OpenUSD scene (open standard, not NVIDIA's)
- OpenUSD scene (open standard, not NVIDIA's)Common scene description every library readsConnects to ovstage scene data layer
Inference & runtime software
- ovstage scene data layerLoads USD once and shares runtime stateConnects to ovphysx / Newton / PhysX, ovrtx RTX rendering and sensors
- ovphysx / Newton / PhysXPhysics simulation on the shared sceneConnects to ovrtx RTX rendering and sensors
- ovrtx RTX rendering and sensorsCamera, lidar and radar outputConnects to ovstream and ovstorage, Kit apps, agent skills and blueprints
Operations & orchestration
- ovstream and ovstorageStreams output to clients and syncs asset dataConnects to Kit apps, agent skills and blueprints
Accelerated computing
- NVIDIA RTX GPUs (workstation, RTX PRO, OVX, cloud)Runs rendering and physicsConnects to ovrtx RTX rendering and sensors, ovphysx / Newton / PhysX
Capabilities
RTX rendering and sensor simulation (ovrtx)35
A C and Python library that embeds physically based camera, lidar, radar and other sensor simulation in your own application.
Why it matters: Produces synthetic sensor data and visual checks without a full desktop application.
Limits: Pre-release; the repository is under an NVIDIA proprietary license. The libraries need an RTX-capable NVIDIA GPU and a compatible driver to initialize.
USD-native physics (ovphysx, Newton, PhysX)4
Physics libraries that run collision, motion and dynamics checks directly on OpenUSD scenes.
Why it matters: Lets a twin or robot scene behave physically, not just look right.
Limits: ovphysx is pre-release. Simulated behavior still needs checking against real measurements.
OpenUSD conversion and validation4
Converters for common 3D formats, URDF, MuJoCo and Gaussian splats, plus validators for SimReady and core USD rules.
Why it matters: Most projects start with data in other tools; conversion and checks decide whether the scene is usable.
Limits: Output quality depends on the source data; converted assets still need review.
Shared scene data, storage and streaming14
ovstage loads USD once and shares runtime state; ovstorage manages and syncs asset data; ovstream delivers GPU output to local or browser clients.
Why it matters: Several tools and users can work on one scene and view results remotely.
Limits: ovstorage and ovstream are marked pre-release and not enterprise-supported.
Agent skills with human review4
NVIDIA-verified skills for tasks such as CAD to SimReady conversion, defect image generation, neural reconstruction and USD performance tuning.
Why it matters: Coding agents can run defined Omniverse workflows as tools.
Limits: NVIDIA describes human review at decision points; agents do not replace engineering sign-off.
Kit SDK and app streaming1
An SDK for building full Omniverse-based applications, with a Kit App Streaming API and preconfigured cloud workstations on AWS, Azure and Google Cloud.
Why it matters: For teams that need an interactive application rather than individual libraries.
Limits: Earlier ready-made apps such as USD Composer and USD Explorer are no longer supported, so the app must be built.
Reference blueprints14
Documented end-to-end workflows, including the DSX Blueprint for AI factories, synthetic manipulation motion generation and a digital twin for interactive fluid simulation.
Why it matters: Shortens design work by showing a complete pipeline that uses several libraries together.
Limits: Blueprints are starting points; some are marked early access, and each needs adapting to your data.
Practical use cases
A plant or warehouse layout change has to be tested before equipment is moved.
- Approach
- Convert CAD and equipment models to OpenUSD, assemble the facility scene, simulate physics and check sight lines, clearances and robot paths in software.
- Role of NVIDIA Omniverse
- Provides the scene format, physics and rendering for the facility twin.
- Data, infrastructure and skills
- Clean CAD data, RTX workstations or servers, and engineers who know OpenUSD.
- Type of benefit
- Lower planning risk
- Caveats
- A twin is only as accurate as its source data; keeping it current is ongoing work.
- First step
- Convert one production cell to OpenUSD and validate it with simready-validate.
A visual inspection model lacks examples of rare defects.
- Approach
- Use the defect image generation agent skill and ovrtx sensor rendering to create labeled synthetic defect images, then evaluate the model on real samples.
- Role of NVIDIA Omniverse
- Generates and renders the synthetic training images.
- Data, infrastructure and skills
- 3D models of the parts, defect definitions, and a held-out set of real images for testing.
- Type of benefit
- More training data for rare cases
- Caveats
- Synthetic images can differ from real ones; always measure on real data.
- First step
- Run the defect image generation skill on one part type.
Robot or vehicle software needs repeatable tests in many conditions.
- Approach
- Build the test scene in OpenUSD, simulate sensors with ovrtx and physics with ovphysx or Newton, and run tests in the loop, often through Isaac Sim.
- Role of NVIDIA Omniverse
- Supplies the simulation libraries that robot and vehicle tools build on.
- Data, infrastructure and skills
- Robot or vehicle models, sensor specifications and a test plan.
- Type of benefit
- Safer, repeatable testing
- Caveats
- Simulation results do not replace field validation.
- First step
- Convert your robot's URDF to OpenUSD with urdf-usd-converter.
Who uses it
BMW Group · Automotive manufacturing
BMW Group: planning car plants in an Omniverse-based Virtual Factory
BMW Group plans and checks its car plants in a Virtual Factory built on NVIDIA Omniverse and OpenUSD. BMW says digital twins now cover more than 30 production sites, cut collision checks for new models from almost four weeks to about three days, and are projected to lower planning costs by up to 30 percent.
Scaling
Foxconn (Hon Hai Technology Group) · Electronics manufacturing
Foxconn: digital twins for new server plants with Omniverse, Isaac and Metropolis
Foxconn uses NVIDIA Omniverse digital twins to plan production lines, Isaac to simulate robots and Metropolis for camera-based monitoring, from Hsinchu to new server plants in Mexico and the US. Most published results are expectations, such as a forecast energy cut of over 30 percent in Mexico.
In production
PepsiCo, with Siemens · Food and beverage manufacturing and logistics
PepsiCo and Siemens: plant and warehouse twins with Digital Twin Composer on Omniverse
PepsiCo is converting selected US plants and warehouses into 3D digital twins with Siemens Digital Twin Composer, which Siemens builds on NVIDIA Omniverse libraries. PepsiCo and Siemens report a 20 percent throughput gain on the first deployment; the program is in early pilots.
Pilot
Unilever · Consumer goods (marketing content production)
Unilever: product digital twins for marketing imagery with Omniverse and OpenUSD
Unilever builds photoreal 3D twins of its products with NVIDIA Omniverse and OpenUSD, working with creative technology partner Collective World, and renders marketing images from them instead of running repeated photo shoots. Unilever and NVIDIA report imagery made twice as fast at half the cost.
In production
Deutsche Telekom (T-Systems) · Telecommunications and cloud services
Deutsche Telekom Industrial AI Cloud: an NVIDIA-based AI factory in Munich
Deutsche Telekom opened its Industrial AI Cloud in Munich on 4 February 2026, with nearly 10,000 NVIDIA Blackwell GPUs in DGX B200 systems and RTX PRO Servers. Telekom says it was over a third utilized at opening, with Agile Robots and PhysicsX among early users.
In production
Works with
Optional integration
- NVIDIA AI EnterpriseEnterprise Support for Omniverse requires an NVIDIA AI Enterprise license.
Complementary tools
- NVIDIA IsaacIsaac Sim is a robotics framework built on Omniverse libraries.
- NVIDIA CosmosNVIDIA describes Omniverse as the simulation environment and Cosmos as the models that turn simulations into photoreal synthetic data.
- NVIDIA cuOptNVIDIA describes cuOpt combined with Omniverse digital twins for logistics planning.
- NVIDIA DSXThe DSX Blueprint for AI factories is listed among Omniverse blueprints.
- NVIDIA Holoscan SDKIsaac for Healthcare combines Omniverse simulation with the Holoscan runtime.
Required by
- NVIDIA IsaacIsaac Sim is built on NVIDIA Omniverse libraries.
Optional integration for
- NVIDIA RTX RemixThe RTX Remix Toolkit is built on NVIDIA Omniverse and stores assets in OpenUSD.
- NVIDIA DSXThe Omniverse DSX Blueprint is part of DSX Sim.
Relationship labels follow NVIDIA's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.
Getting started
Check hardware and drivers
Compare your GPU, OS and driver with the Omniverse technical requirements page.
Check: The GPU is an RTX model and the driver version is on the validated list.
Install ovrtx and ovstage
Clone the ovrtx repository and install the packages with uv add ovrtx ovstage (or pip), then run the minimal Python example.
Check: The example loads the sample robot scene and produces rendered output.
Convert one of your own assets
Use usd-convert-asset or urdf-usd-converter on a real part or robot, then run simready-validate.
Check: The validator reports no blocking errors for the asset.
Try an agent skill
Run one Omniverse agent skill, such as CAD to SimReady, from the NVIDIA skills repository with your coding agent.
Check: The skill output is reviewed and accepted by an engineer.
Pick a reference blueprint
Choose the blueprint closest to your goal, for example the DSX Blueprint for AI factories, and follow its guide.
Check: You can map each blueprint component to a part of your own pipeline.
Decide on support
Choose between community support (forums, Discord) and Enterprise Support through an NVIDIA AI Enterprise license.
Check: We recommend agreeing the support route before production use.
Official resources
Could this technology help you?
Describe your project to the Solution Architect. It starts with NVIDIA Omniverse as context but recommends independently, including when you do not need it.
Sources
Each statement above links to the source it comes from. Labels say who reported it.
- NVIDIA Omniverse documentation (opens in a new tab)
- NVIDIA Omniverse (product page) (opens in a new tab)
- Omniverse technical requirements (opens in a new tab)
- NVIDIA Omniverse for developers (opens in a new tab)
- ovrtx GitHub repository README (opens in a new tab)
- Alliance for OpenUSD (opens in a new tab)
- Omniverse License Server Overview (Legacy) (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.