Investigating The Relationship Between Hand Sketch Quality and AI-Assisted Concept Render Quality in Product Design

Abstract Sketching is seen as a key part of the early-stage design process, enabling designers to communicate conceptual ideas to one another and to external stakeholders. The introduction of image-generative AI opens up new possibilities for AI-assisted sketching, which may be a useful tool for designers. In the present study, undergraduate product design students received introductory training in an image-generative AI tool, Vizcom, which uses a multimodal text- and sketch-based input to create more photorealistic rendered outputs. The purpose of the study was to characterize the relationship between the qualities of the sketch inputs and the AI-generated outputs. The input sketches and output renders were also compared in terms of understandability. Investigating what kinds of sketches are most useful to use as inputs for generative AI tools is important for both design instruction and practice. Our study found that line quality, proportionality, and understandability of the input sketches were all positively correlated with the proportionality and understandability of the AI-generated outputs from Vizcom. We also found no significant differences in level of understandability between the input sketches and the AI-generated outputs. Our findings indicate that input sketches with correct proportions and clean lines are ideal inputs for AI-aided sketching that utilizes Vizcom or similar programs to represent design concepts.