Lumos: AI-driven prompt optimisation tool for assisting conceptual design

AI drawing tools, which rely on high-quality prompts to guide and refine the creative process, have the potential to enhance efficiency and quality during the conceptual design stage. However, addressing the inherent accuracy issues of AI is crucial. This study explores the efficiency and usability of integrating AI drawing with prompt optimisation tools and introduces a GPT-based prompt optimisation tool for validation. Experiment results indicate that, during the divergent thinking stage of concept design, both junior and senior designers using this tool reduced their average time consumption by 35.6% and 21.6%, respectively. Expert evaluations of the generated images reveal higher selection rates for images produced after prompt optimisation with this tool compared to those generated using natural language. These findings underscore the tool's potential to boost productivity and improve the usability of generated images, thereby expanding designers' imaginative boundaries and enhancing their ability to leverage AI for streamlined workflows.