Digital Twin Physics Simulation Platform
It covers the full workflow including simulation asset creation, simulation scene construction, simulation data collection, model training and deployment verification. Digital twin technology bridges the physical and simulated worlds to break through the data bottleneck of embodied intelligence.
Through its self-developed Connector, the Synthesis platform enables the import, lightweight processing, structuring, and conversion to formats like USD for most mainstream 3D software formats. Leveraging the Multiverse engine, it achieves ultra-fast loading and browsing of massive 3D models.
"The platform features a vast library of 1:1 scale, high-fidelity reconstructed assets (most including hinges and covering both rigid and soft bodies). These include 3D engineering models, 3D Gaussian Splatting (3DGS) reality models, and fully interactive physics-based scenes built as 1:1 digital twins. Every asset is strictly reconstructed to match physical counterparts, ensuring simulation environments align perfectly with real-world physics, construction standards, and regulatory codes."
"Beyond achieving physical-grade geometric precision, all assets utilize PBR (Physically Based Rendering) materials to ensure visual consistency between simulation and reality. At the physical level, detailed parameters for dynamics, colliders, and joints ensure complete behavioral fidelity. Furthermore, unified standards for semantic labeling, data organization, and coordinate systems ensure all assets are fully optimized for physics-based simulation."
"Synthesis provides an online viewer, allowing users to browse 3D assets, buildings, and 3DGS models directly in a web browser, with full access to attribute, semantic, and joint information. All assets and models support online conversion to the OpenUSD format for seamless integration with NVIDIA Isaac Sim. Select Sim-Ready assets, architectural models, and fully interactive scenes are open-sourced under the CC BY-NC 4.0 license and are available for direct download via Synthesis."
Synthesis provides an online scene assembly tool with LLM-powered generalization, enabling users to quickly construct physically accurate and interactive scenes that mirror the real world through a drag-and-drop interface directly in a web browser. Users can seamlessly incorporate Sim-Ready assets, architectural models, and 3D Gaussian Splatting captures into their scenes and adjust their placements intuitively.
Leveraging Synthesis' generalization capabilities, users can rapidly generate a large volume of realistic and functionally valid scene variations through logical adjustments to lighting, weather conditions, material properties, and asset positioning.
- Residence
- Factory
- Subway Station
- Hospital
- Laboratory
Rapid Construction of Simulation Environments
With self-developed tools, users can easily load simulation assets and scenes across various simulators and rapidly initialize simulation environments to meet the demands of simulation data collection and task verification.
Meanwhile, we have deeply customized the WebRTC streaming solution based on NVIDIA Isaac Sim and built our proprietary Isaac Sim cluster system. Users can load simulation scenes, collect data, and verify tasks and models simply via a web browser. This improves GPU sharing utilization and significantly cuts down hardware costs for simulation.
Simulation Data Collection
With our self-developed data acquisition kit, users can remotely operate to complete data collection and augmentation within simulation environments. We fully leverage the strengths of simulation in generalization and randomization to achieve low-cost coverage of long-tail scenarios and large-scale data expansion.
To facilitate seamless access to assets from the Synthesis platform, we have developed and open-sourced the Synthesis Asset Explorer extension for Omniverse and Isaac Sim(https://github.com/Extwin-Synthesis/Synthesis-Assets-Explorer) .
With this extension, you can browse and utilize high-quality USD assets hosted on Synthesis just as you would with NVIDIA's official repositories — directly within your Omniverse workflow.
