High-precision improved text prompt 3D model generation and printing method

Purpose This study aims to improve the method of generating 3D models from text descriptions and apply it to additive manufacturing. The objective is to overcome the limitations of traditional additive manufacturing models, such as slow design and rigid model lines, to enhance efficiency, increase customization and obtain realistic 3D models that combine texture and color features with smooth curves, thereby promoting intelligent and personalized manufacturing. Design/methodology/approach The method consists of three stages. First, a neural network is constructed to incorporate lighting vectors to predict the signed distance function, using a diffusion model to guide the 2D mapping image of the 3D model based on text descriptions. Second, differentiable rasterization and anti-aliasing are used to refine the texture, with the diffusion model being replaced, and the neural network inherits weights to optimize memory usage. Finally, the 3D model data is converted and exported as an stereo lithography file, sliced and printed using 3D printing technology. Findings This method significantly improves design and production in additive manufacturing. Through three steps – model generation, texture refinement and additive manufacturing – a high-quality 3D object is obtained. It breaks the limitations of traditional model generation and manufacturing, enabling faster and more flexible production. Originality/value This study innovatively applies text-to-3D generation to additive manufacturing. By combining advanced model generation and 3D printing technologies, it provides a new design approach for the field of additive manufacturing, improving design efficiency and customization capabilities and paving the way for personalized and intelligent manufacturing.