3D-GPT-Avatar

A multimodal framework generating high-fidelity 3D digital humans from text prompts using Large Language Models.

3D-GPT-Avatar streamlines the creation of animatable 3D assets by leveraging the reasoning capabilities of GPT-4 to interpret complex textual descriptions. The system orchestrates specialized tools (like SMPL-X for body modeling and TeCH for texture synthesis) to produce 2K-resolution avatars complete with realistic clothing and facial expressions. By automating the technical pipeline from prompt to mesh, it reduces production time from hours of manual sculpting to minutes of automated generation: a critical leap for developers in gaming, VR, and digital commerce.

https://github.com/yushuang-wu/3D-GPT-Avatar