princeton-vl/infinigen

Infinite Photorealistic Worlds using Procedural Generation

What it solves

Infinigen creates photorealistic 3D worlds and assets procedurally, removing the need for manual 3D modeling or the reliance on limited, existing 3D datasets. It provides a scalable way to generate infinite variations of nature, indoor scenes, and articulated simulation assets for use in computer vision and robotics research.

How it works

The project uses procedural generation—creating content via algorithms and rules—rather than manual artistry. It integrates with Blender to generate 3D scenes, materials, and objects. The system is split into specialized modules:

  • Infinigen Nature: Generates photorealistic natural environments.
  • Infinigen Indoors: Creates photorealistic indoor scenes.
  • Infinigen Articulated: Produces simulation-ready assets with articulated joints (e.g., objects that can move or open).
  • ProcFunc: A Python-based abstraction layer for 3D procedural generation.

Who it’s for

It is primarily designed for researchers in computer vision, robotics, and simulation, who need high-quality, diverse, and ground-truth annotated 3D data for training AI models.

Highlights

  • Infinite Variety: Generates an endless stream of unique 3D environments and objects.
  • Multi-Environment Support: Covers nature, indoor spaces, and articulated objects.
  • ** uma Simulation-Ready**: Assets can be exported to external file formats (OBJ, OpenUSD) and physics simulators.
  • Ground Truth: Provides extended ground-truth annotations for the generated data.

Related

  • Project
    HorizonRobotics/EmbodiedGen

    A generative 3D world engine that converts text and images into simulation-ready 3D assets and environments for embodied AI, deployable across multiple robotics simulators.

  • Project
    ZiYang-xie/WorldGen

    WorldGen is a tool that generates immersive 3D scenes from text prompts or images in seconds, supporting both Gaussian Splatting and mesh outputs for VR, games, and simulations.

  • Project
    xandergos/terrain-diffusion

    A learned successor to Perlin noise that uses diffusion models to generate infinite, deterministic, and randomly-accessible real-time terrain and climate data.

  • Project
    LiteReality/LiteReality-Agent

    LiteReality‑Agent is an open‑source toolkit that converts a phone‑captured RGB‑D scan (via the LiteReality iOS app) into a fully‑articulated, graphics‑ready indoor 3D scene. An LLM‑driven agent coordinates vision models (TRELLIS, GroundingDINO) and Blender to generate, place, and polish objects, producing GLB/BLEND assets usable in game engines, AR/VR, or robotics. It runs on macOS or Linux, can use Modal’s cloud GPUs or a local ≥24 GB GPU, and requires API keys for image generation and the LLM.

  • Project
    Tencent-Hunyuan/HY-World-2.0

    A multi-modal world model framework that generates and reconstructs editable 3D assets (meshes and Gaussian Splattings) from text, images, or video for use in game engines.