Tencent HunYuan recently announced the open-source release of five new 3D generative models. Built upon Hunyuan3D-2.0, these models boast faster generation speeds, richer details, and more realistic material representation. Simultaneously, Tencent's self-developed 3D AI creation engine has been upgraded, adding features like multi-view input, intelligent model reduction, and full format compatibility. It's now fully open to C-end users.
Among these new models, the Turbo series achieves a significant speed increase (tens of times faster) in geometric model generation while maintaining high accuracy and quality, completing the entire process within 30 seconds. This acceleration is powered by Tencent HunYuan's FlashVDM, a 3D generation acceleration framework that breaks through traditional efficiency barriers, bringing large-scale 3D model generation down to the second level.
The multi-view model, Hunyuan3D-2-MV, leverages information from multiple input views to better capture details and generate 3D assets that meet user expectations. The lightweight mini series, through model architecture optimization and performance improvements, significantly reduces computing costs. Its geometric models can be deployed on 4080 GPUs or even Apple M1 Pro chips, greatly expanding application scenarios.
The upgraded Tencent HunYuan 3D AI creation engine supports multi-view input. Users only need to upload 2-4 standard-view images to quickly generate high-precision, high-quality 3D models. This significantly reduces the production cost of converting multi-view artwork into 3D models for game production and 3D UGC creation.
The new engine also features intelligent 3D reduction, adaptively generating hundreds to thousands of triangular faces. This improves the smoothness of geometric edges, maximizing model detail with fewer polygons, resulting in cleaner meshing, reduced rendering difficulty, and increased model usability and production efficiency.
In terms of material representation, this upgrade achieves enhanced PBR (Physically Based Rendering) material generation. Through physical property simulation, the models are given more realistic colors and material expressions, showcasing cinematic quality in light and shadow interactions.
Regarding compatibility, in addition to common OBJ, GLB, and FBX formats, the new engine can output STL, USDZ, and MP4, seamlessly connecting with 3D printing tools and supporting quick model previews and real-time mobile interaction, meeting the format requirements of various business scenarios.
Tencent HunYuan 3D generative models are reportedly being used in user-generated content (UGC), product material synthesis, and game 3D asset generation. In game development, the large model-generated 3D models already meet the standards for some game assets, including the rationality of geometric meshing, texture accuracy, and skeletal skinning.
