Computing Colour and Narrative in Architectural Sketches: A model framework to support sketching in computational design

Sketching encourages exploration in architectural design and narratives accompany sketches. Colour, an essential design element, has often been disregarded in research on sketching. In computational design, sketches are only used as inputs for generating a final image or 3D model, and investigating the process is disregarded. The aim of the study is to develop a model framework that facilitates the relationship between colour and narrative in sketching processes. The research questions focus on how AI can support the exploratory nature of sketching without compromising its ambiguity and how meaningful AI suggestions can be generated using visual and textual modalities. Through protocol studies, the study examined colours in conveying physical and atmospheric qualities and the complementary role of narratives in expressing design intentions. A domain-specific dataset of watercolour sketches paired with narratives was created and used to fine-tune a Stable Diffusion model. The model’s performance was evaluated against a base version, showing its ability to generate suggestions based on colour-narrative relationships while maintaining sketch ambiguity. This research contributes a novel dataset and framework for integrating AI into exploratory architectural sketching, offering new insights for human-machine collaboration.