Tencent Hybrid launches 5 3D generated models, all open source

March 18, 2011 - In today'sTencent Hunyuan 3D Open SourceDay event.Tencent Hybrid Announces 5 New 3D Generative Models, in the generation of speed, details and material expression have been improved, and all open source.

at the same time,Its self-developed 3D AI creation engine has also received an upgradeIn addition, we have added the ability of multi-view input, intelligent model surface reduction, and full compatibility of formats, which are fully open for C-users to use.

All 5 open source models are based on Hunyuan3D-2.0:

• Turbo series models:On the basis of ensuring high precision and quality, the geometry generation model is accelerated by tens of times to ensure that the entire generation process can be completed within 30 seconds. This acceleration capability comes from FlashVDM, a 3D generation acceleration framework proposed by Tencent Hybrid, which breaks through the efficiency barriers of traditional models and improves the generation time of large-scale 3D models to the second range.
• Multi-view versioning model:Hunyuan3D-2-MV, for example, is able to better capture details and generate 3D assets that meet user expectations by combining input information from multiple views.
• Lightweight mini series models:By optimizing the model architecture and improving the operation efficiency, the arithmetic cost can be further reduced, and its geometric model can be deployed on 4080 graphics cards or even Apple M1 Pro chips, which extends the scenarios for the application of the model.

The latest open source models include the multi-view shape model Hunyuan3D-2mv and the 0.6B shape model Hunyuan3D-2mini, etc. 1AI attaches the open source address as follows:

https://github.com/Tencent/Hunyuan3D-2