HunyuanWorld 1.0: Generating Immersive, Explorable, and Interactive 3D Worlds from Words or Pixels

Creating immersive and playable 3D worlds from texts or images remains afundamental challenge in computer vision and graphics. Existing worldgeneration approaches typically fall into two categories: video-based methodsthat offer rich diversity but lack 3D consistency and rendering efficiency, and3D-based methods that provide geometric consistency but struggle with limitedtraining data and memory-inefficient representations. To address theselimitations, we present HunyuanWorld 1.0, a novel framework that combines thebest of both worlds for generating immersive, explorable, and interactive 3Dscenes from text and image conditions. Our approach features three keyadvantages: 1) 360{\deg} immersive experiences via panoramic world proxies; 2)mesh export capabilities for seamless compatibility with existing computergraphics pipelines; 3) disentangled object representations for augmentedinteractivity. The core of our framework is a semantically layered 3D meshrepresentation that leverages panoramic images as 360{\deg} world proxies forsemantic-aware world decomposition and reconstruction, enabling the generationof diverse 3D worlds. Extensive experiments demonstrate that our methodachieves state-of-the-art performance in generating coherent, explorable, andinteractive 3D worlds while enabling versatile applications in virtual reality,physical simulation, game development, and interactive content creation.