Generative Image AI Using Design Sketches as input: Opportunities and Challenges
Chengzhi Zhang, Weijie Wang, Paul Pangaro, Nikolas Martelaro, Daragh Byrne
Abstract
Generative image AI has sparked heated discussion among creative professionals. However, how it might be a tool for design practice is understudied, and it is not yet understood how designers may find image AI helpful across the stages of design practice. To address this, our preliminary study explores how designers use generative image AI accompanied by design sketches to inform early-stage 3D design. Further, we also examine the perceived limitations of text-to-image models. To do this, we recruited 11 Architecture graduate students with a median work experience of 2 years. Participants completed a design task using generative image AI packages and incorporated design sketches as inputs. The study findings provide insights into how image AI can or can not be a valuable resource for architectural design practitioners. Further, findings suggest possible directions for future image AI-assisted design tools and workflows.