Hunyuan 3D

Tencent's high-performance generative framework for creating production-ready 3D meshes from text or images in under 10 seconds.

Hunyuan 3D leverages a two-stage generation process (a multi-view diffusion model followed by a sparse-view reconstruction model) to bridge the gap between 2D synthesis and 3D geometry. It outputs high-quality assets in standard formats like .obj and .glb, featuring precise topology and vertex coloring suitable for game engines and AR/VR workflows. By utilizing a DiT-based (Diffusion Transformer) architecture, the system achieves state-of-the-art results on the OmniObject3D benchmark, outperforming previous models in both geometric fidelity and inference speed.

https://github.com/Tencent/Hunyuan3D-1