Download
HunyuanWorld-1.0 is an open-source, simulation-capable 3D world generation model developed by Tencent Hunyuan that creates immersive, explorable, and interactive 3D environments from text or image inputs. It combines the strengths of video-based diversity and 3D-based geometric consistency through a novel framework using panoramic world proxies and semantically layered 3D mesh representations. This approach enables 360° immersive experiences, seamless mesh export for graphics pipelines, and disentangled object representations for enhanced interactivity. The architecture integrates panoramic proxy generation, semantic layering, and hierarchical 3D reconstruction to produce high-quality scene-scale 3D worlds from both text and images. HunyuanWorld-1.0 surpasses existing open-source methods in visual quality and geometric consistency, demonstrated by superior scores in BRISQUE, NIQE, Q-Align, and CLIP metrics.
Features
- Generates immersive, explorable, and interactive 360° 3D worlds from text or image inputs
- Uses panoramic world proxies combined with semantically layered 3D mesh representations
- Supports mesh export for seamless integration with existing computer graphics pipelines
- Employs hierarchical 3D reconstruction for coherent and high-quality scene generation
- Outperforms competing open-source panorama and 3D world generation methods in visual and geometric metrics
- Compatible with Flux and adaptable to other image generation models like Stable Diffusion
- Provides ready-to-use scripts for text-to-world and image-to-world generation workflows
- Includes a web-based ModelViewer tool for quick visualization and interaction with generated 3D worlds
Project Samples
