The article introduces 3D-GPT, a framework that uses large language models (LLMs) for instruction-driven 3D modeling. This approach simplifies the complex and time-consuming 3D modeling process. The 3D-GPT framework includes three core agents: the task dispatch agent, the conceptualization agent, and the modeling agent. These agents help to enhance initial scene descriptions and integrate procedural generation. The framework not only interprets and executes instructions effectively but also collaborates well with human designers. Moreover, it integrates seamlessly with Blender, a 3D software, making it easy to create 3D assets. The authors see potential in using LLMs in 3D modeling and believe their work can serve as a basic framework for future advancements in scene generation and animation.
Publication date: 20 Oct 2023
Project Page: https://chuny1.github.io/3DGPT/3dgpt.html
Paper: https://arxiv.org/pdf/2310.12945
