Tencent has recently released five open-source models based on Hunyuan3D-2.0, including the Turbo series, multi-view versions, and the lightweight mini series. The Turbo series leverages FlashVDM, a 3D generation acceleration framework developed by Tencent Hunyuan, to boost geometry generation speeds by several dozen times. This advancement reduces the creation time of large-scale 3D models to mere seconds, with the entire process completing within 30 seconds.
The multi-view version models, such as Hunyuan3D-2-MV, integrate input information from multiple views to enhance detail capture and generate 3D assets that better align with user needs. Meanwhile, the lightweight mini series optimizes model architecture and boosts operational efficiency, reducing computational costs. Its geometry models can be deployed on hardware like the NVIDIA 4080 GPU or Apple M1 Pro chip, broadening their range of applications.
Additionally, Tencent Hunyuan's 3D AI creation engine has been upgraded. The new version supports multi-view input functionality, allowing users to upload 2-4 standard perspective images to quickly generate high-precision, high-quality 3D models. This feature significantly cuts production costs for scenarios like game development and 3D UGC content creation by streamlining the transition from multi-view concept art to 3D models.
The upgraded engine also enhances 3D intelligent mesh reduction, adaptively generating hundreds to thousands of triangular faces to improve geometric edge smoothness while maximizing detail representation in low-poly models. Furthermore, improvements in PBR material generation deliver more realistic color and texture expression, achieving cinematic-level quality.
In terms of compatibility, the updated engine supports popular formats such as OBJ, GLB, and FBX, as well as STL, USDZ, and MP4 formats. This ensures seamless integration with 3D printing tools, fast model previews, and real-time interaction on mobile devices, catering to diverse business needs for various 3D model formats.
