HunyuanWorld

HunyuanWorld generates explorable 3D environments from text prompts or images. It's for game developers, VR creators and researchers who need scene geometry for interactive content or physical simulation.

• Text and image inputs: Create a 360° environment from a written scene description or a reference image.
• Mesh export: Export generated geometry for use in existing computer graphics workflows.
• Separate scene objects: Keep foreground elements distinct within layered scene geometry to support interaction with individual objects.
• Panorama generation: Produce panoramas from text or images, with dedicated inpainting models for scene content and skies.
• Browser viewer: Inspect and explore generated scene files in the supplied web viewer without importing them into another graphics application.

Generation runs locally with a FLUX-based model using Python and PyTorch. The quantized Lite variant reduces memory use and supports consumer NVIDIA GPUs such as the RTX 4090; optional caching can speed up generation. The browser viewer displays the generated files, while the generation code runs separately on GPU hardware.