NVIDIA releases SimFoundry to turn real-world videos into physics-ready simulations
zhengyiluo · x · 2026-08-16
NVIDIA's GEAR team has released SimFoundry, a real2sim2real framework that automatically converts real-world scenes from images or videos into simulation-ready worlds. It can transform a short video into a physics-ready simulation scene in under an hour, automatically segmenting objects, reconstructing geometry, and generating textured 3D meshes. The output is compiled into an OmniGibson scene with physical parameters. The code is now open-sourced to facilitate foundation model training using synthetic data.
More from Embodied
- Seeed Unveils reComputer RK3576 Edge AI Module — ___Mufasaa · 2026-08-16
- Source Robotics Open Sources 3D Printed Differential Robot Wrist — blaizedsouza · 2026-08-16
- After reviewing 1,000+ physical AI companies, here is how to build a great robotics startup — lukas_m_ziegler · 2026-08-16
- Robot fight competition scheduled for tonight in SF, streaming on X — beffjezos · 2026-08-16
- Dyna-2 World-Action Model: 1M Hours of Human Video Pre-training Yields Human-to-Robot Transfer Scaling Law — chris_j_paxton · 2026-08-16
- HUSKY Enables Humanoid Robots to Skateboard via Physics-Aware Control — tom_doerr · 2026-08-16