Hyper3D WorldGen: Generating physically correct 3D worlds from images
APPSO · wechat · 2026-09-01
Hyper3D released WorldGen, a model that converts a single RGB image into a complete 3D world composed of independent assets. Based on the CAST method (SIGGRAPH Best Paper), it infers spatial relationships to fix floating/penetrating bugs and exports assets with topology and PBR materials.
Key Features
- Physics-Aware Reconstruction: Infers contact/support relationships to build a graph and optimize geometry.
- Production-Ready Assets: Outputs assets ready for Blender/Unreal, used in VFX, gaming, and robotics.
- Workflow Integration: Connects with video models like Seedance 2.5 to restore 3D scenes from 2D frames.
Use Cases
- Gaming/VFX: Accelerates environment construction; enables scalable asset generation for open-world games.
- Embodied AI: Provides low-cost, diverse simulation environments for robot training.
- Spatial Computing: Converts real photos into roamable 3D scenes for XR devices.
Related event: Deemos Releases WorldGen: One Image Generates Fully Editable 3D Worlds(8 posts)→
More from Embodied
- TRACE 开源：8 小时单卡把 G1 控制器迁移到全尺寸人形 H2 — zhengyiluo · 2026-10-18
- MIT's SANDO Drone Planner Avoids Every Moving Obstacle Across 12 Real Flights — Olivier__OG · 2026-10-18
- SimInfinity: open-source robotic simulation benchmark with 10K real-world tasks — joemeno · 2026-10-18
- Robi: a German/English voice companion demo built on Gemma, Tyger and Humaneness Voice TTS — realmrfakename · 2026-10-18
- Munich's Plim unveils wheeled home-care humanoid controllable from 9,000 km away — CyberRobooo · 2026-10-18
- Musk says Optimus and Cybercab will make up the vast majority of Tesla's value — XFreeze · 2026-10-18