From Text to Physical Prototypes: An AI-Driven 3D Printing System

Abstract

Much of current literature on artificial intelligence has focused on large language models for natural language generation. However, rapid advancements have also been made in AI-driven 3D model generation. Recent developments have enabled the generation of detailed 3D models from single source images. When combined with modern diffusion-based image generation models, this allows the construction of automated text-to-3D generation pipelines. These models have primarily been designed for video game asset generation, but have exciting potential in engineering design and prototyping. This thesis presents the design of a fully integrated generative AI 3D printing system capable of direct text-to-prototype fabrication. The system eliminates the conventional multi-step 3D printing workflow by automating the entire pipeline from natural language input to physical object output. The proposed platform combines text-to-2D and 2D-to-3D generative models (namely Hunyuan3D 2.0) with automated mesh processing, slicing, and G-code generation, all embedded within a custom hardware and software architecture. In the final implementation, these components are integrated into a custom-built FDM 3D printer supported by a touchscreen interface linked to a single-board computer and cloud-based GPU resources. Key contributions of this work include the development of a robust automated pipeline for converting text prompts into manufacturable geometries, the design of a user interface enabling interaction with generative AI for physical fabrication, and the construction of a fully functional 3D printer with integrated AI capabilities. Experimental results demonstrate the system’s ability to reliably generate and fabricate complex geometries without user intervention. This work highlights the potential of generative AI to transform rapid prototyping by lowering technical barriers and accelerating the concept-to-prototype cycle. The proposed system serves as a proof of concept for future intelligent manufacturing platforms, with possible applications in engineering design expected to increase with the continued development of more robust AI models.