Omniverse's expansive new features, revealed at this morning’s GTC keynote, include Omniverse Cloud managed services, Omniverse Kit connected tools and microservices, and Omniverse Exchange, a publishing portal for industry and purpose-built third-party solutions; the online conference runs through September 22.
NVIDIA’s GTC, the online AI developer conference, kicked off yesterday with a complete program of sessions slated to bring together students, developers, researchers, creators, IT decision-makers, and business leaders who shape the world of AI, computer graphics, and data science. The event runs through September 22; the full schedule is available here.
This morning during NVIDIA CEO Jensen Huang’s latest GTC keynote, the largest ever release of new features for Omniverse was revealed, including Omniverse Cloud managed services and container deployments, new developer toolkits, and an open publishing portal for developers. The newly expanded capabilities promise to optimize the ease with which developers can build, extend, and connect 3D tools and platforms to the Omniverse ecosystem.
Cloud services for building and operating metaverse applications:
NVIDIA’s first SaaS offering, Omniverse Cloud, is an infrastructure-as-a-service that connects Omniverse applications running in the cloud, on-premises, or on edge with users able to create and collaborate across devices. With the Omniverse Nucleus Cloud, Omniverse App Streaming provides shared access to virtual worlds for editing and the scalability of 3D workloads with Omniverse Farm.
Omniverse Cloud runs on the planetary-scale Omniverse Cloud Computer, powered by NVIDIA OVX for graphics-rich virtual world simulation, NVIDIA HGX for advanced AI workloads, and NVIDIA Graphics Delivery Network to enable low-latency delivery of interactive 3D experiences to edge devices.
Applications for industry workflows, such as DRIVE Sim for testing and validating autonomous vehicles and Isaac Sim for training and testing robots, are packaged as containers for simple deployment. For synthetic data for industry use cases, Omniverse Replicator in the cloud enables synthetic 3D data generation for perception networks.
Developer experiences with Omniverse Kit:
Designed as the foundation of Omniverse-connected tools and microservices, the Omniverse Kit is a complete toolset for building applications, extensions, and connectors for Omniverse. Its modularity offers flexibility in assembling tools to suit users’ individual needs.
- Viewport 2.0 is now the default, along with the omni.ui.scene, and with the new Viewport Menu, users can create their own workflow and build their own tool menu.
- Kit Core now supports third-party extensions in C++ and has several new features, including Kit Actions for easier scripting and hotkeys and Kit Activity Monitor for a full timeline of load activity. In addition, users can easily bring their own C++ library into Kit and build performance-critical code.
- Kit Runtime introduces RTX performance and quality improvements, and Action Graph has UI/UX improvements across the board, including specific nodes for creating UI.
Publishing portal on Omniverse Exchange:
With the upcoming release of the Omniverse Exchange, developers will have a new channel to expand the user audience for connectors, extensions, and asset libraries. The publishing portal will provide a workflow for partners and community members to publish applications, connectors, and extensions that offer industry and purpose-built third-party solutions. All content undergoes security vetting and quality assurance before being published.
Developers can be among the first to upload extensions or connectors to the NVIDIA Omniverse Exchange Publishing Portal through its early-access program, available by application.
Advances to USD:
Currently used in a wide range of industries, the open and extensible ecosystem USD is, according to NVIDIA, the best candidate to serve as the HTML of the metaverse. To support its evolution towards satisfying the requirements of the metaverse, NVIDIA supports the USD ecosystem in all areas, from education to building custom schemas for specific industry use cases and providing free USD assets and resources to all audiences. The support aims to accelerate its path to becoming the standard language of virtual worlds.
Key updates for NVIDIA’s work in USD:
- Asset Revolver (Ar) 2.0 support, enabling Omniverse Connector interoperability with any build of USD
- New and improved USD code snippets with documentation for common workflows
- USD C++ extension examples for Omniverse Kit, including open-source USD schema examples and tutorials
- Support for UTF-8 identifiers for full interchange of content encoded with international character sets
- An open-source text render delegate enabling human-readable debugging and more efficient unit and A/B testing of scene delegate implementations
NVIDIA has developed Omniverse Connectors for a variety of top industry applications.
Introduced at GTC 2022:
- Connectors for Autodesk Alias
- The Siemens JTTM data format
- Open Geospatial Consortium (OGC) Web Map Service and APIs (now available in beta)
Connectors are also in development for Unity, SimScale, and some of the Siemens Xcelerator portfolio.
In addition, new Omniverse Extensions extend the functionalities of Omniverse Apps, including Motionverse from Beijing Deepscience Technology Ltd, in3D, and SmartCow. Prevu3D, and Move.AI will soon have extensions available, as well.
SimReady assets for simulation workflows:
NVIDIA has released SimReady, a new category of 3D asset. 3D assets for digital twins and AI training workloads require specific, USD-based properties. Thousands of SimReady assets are currently in development as well as the specifications for building them, tailored to specific simulation workloads like manipulator bot training and autonomous vehicle training.
SimReady goes beyond full-fidelity visuals for 3D art assets. It includes content with attached, robust metadata that can be inserted into any Omniverse simulation and behave as it would in the real world. For developers and technical artists, SimReady content can act as a standard baseline or starting point and evolve as simulation tools like Omniverse become more robust.
The assets provide full-fidelity visualization with the intent of photorealism. Core simulation metadata is always included in the assets and can be easily accessed on import of the art asset. SimReady assets also leverage the modular nature of USD to allow flexibility for technical artists generating synthetic data.
Watch a video example of Omniverse SimReady assets:
SimReady asset attributes for developers:
- Semantic labels help train simulation algorithms by identifying the various components of a 3D model in a predictable and consistent way. These labels provide ground-truth arbitrary output variables for annotations within training.
- Physical materials that bring simulations closer to reality and allow the fine-tuning of how computers see the world with non-visible sensor support like lidar and radar.
- Rigid body physics for accurate mass and defined center of gravity to help technical artists create simulations that reflect real-world behavior.
- Defined taxonomy with consistent tagging for search and discoverability, making assets usable across many domains.
- Robust kinematics and constraints to define complex, multi-part relationships and behaviors.
- Advanced EM materials to unlock behaviors across the light spectrum.
Download Omniverse to access the first collection of free SimReady assets.
To learn more, watch the GTC session, “How to Build Simulation-Ready USD 3D Assets:”
Generate synthetic datasets with Omniverse Replicator:
NVIDIA Omniverse Replicator is now deployable in the cloud through containers hosted on NVIDIA NGC, and through Omniverse Cloud early access. It features a new Replicator Insight app for enhanced viewing and inspecting generated data.
Partners, including Siemens SynthAI, SmartCow, Mirage, and Lightning AI, are integrating Omniverse Replicator into their existing tools to extend their synthetic data workflows. More information about how Omniverse Replicator is accelerating AI training with custom physically accurate synthetic data generation pipelines is available here.
A new era of neural graphics:
Neural graphics can change the way content is created and experienced. AI-powered tools on NVIDIA AI ToyBox now allow creators to experiment with the latest research in their 3D workflows. AI Animal Explorer is available now, with updates coming soon, including:
- GANverse3D with GAN-based and diffusion model-based experimental AI tools
- AI Car Explorer with meshed/textured car models by attributes like make and style
- Canvas360 with 360-degree landscapes and backgrounds for 3D design
Procedural behavior in Omniverse:
Omniverse now enables proceduralism for behavior, movement, and motion in Omniverse worlds. Behavior and physics simulation technologies have improved ease of use, refinement, and functionality. With the new OmniGraph node-based physics systems, behavior, motion, and action can be entirely procedural without manual animation.
NVIDIA PhysX is an advanced real-time physics engine in Omniverse.
Other key updates announced at GTC:
- Improved automation and a physics toolbar for object creation
- Support for multiple scenes with the ability to put individual assets and stages into their own physics scene
- Audio for physics collisions
- Ability to author vehicles and force fields in Omnigraph
- Reduced CPU bottlenecks and GPU improvements for soft body simulation
- Ability to simulate collisions for particle systems
Expanding ecosystem of extensions with Omniverse Code Contest:
This summer, worldwide developers from around the world used Omniverse Code to build extensions for 3D worlds as part of the NVIDIA #ExtendOmniverse Contest. New tools were submitted for layout and scene authoring, Omni.ui, and scene modifier and manipulator tools.
Winners of the contest will be announced at GTC during the NVIDIA Omniverse User Group on September 20 at 3:00 PM (Pacific time).
NVIDIA GTC offers panels, talks, and hands-on labs for developers who want to learn more about Omniverse and other tools used to build metaverse extensions and applications.
Watch the GTC keynote by NVIDIA founder and CEO Jensen Huang to hear about the latest advancements in AI and the metaverse.