Prompt-to-fabrication: a proposed multimodal process for the automated fabrication of facade elements

Abstract

Architects increasingly use generative AI to produce novel façade concepts quickly, yet translating AI imagery into buildable components remains difficult. This paper proposes a prompt-to-fabricate workflow that couples text-to-image diffusion, AI-based 2D-to-3D mesh generation, and algorithmic geometry processing in Rhino/Grasshopper to convert visual intent into fabrication-ready panel geometry. The pipeline uses controllable image synthesis to explore design variations, then reconstructs a selected concept as a detailed mesh and rationalizes it through cleanup, topology control, thickness enforcement, and panelization. Physical production is achieved through a hybrid method in which CNC-milled XPS foam molds are used to cast or spray fiber-reinforced concrete such as GFRC. Preliminary results indicate that the approach can generate intricate, customized concrete façade elements with relatively rapid turnaround. The main advantages are fast concept iteration, continuity from generative design to fabrication constraints, and the ability to realize complex forms without fully manual modeling. Key limitations include partial controllability of AI outputs, uncertainty in mesh fidelity, the effort required for rationalization, and constraints imposed by materials and fabrication processes. The work demonstrates an integrated multimodal bridge from prompting to physical prototypes and motivates future research in performance-driven generative workflows and more automated construction pipelines.