While car exterior designers are adopting 2D generative AI in their workflows, interior designers still face significant challenges. AI often fails to capture intended forms and layouts from interior sketches typically drawn in extreme perspectives and containing multiple components. Moreover, designers must rely on extensive manual work to explore numerous combinations of various components across different views. Thus, we propose a novel workflow in which designers sketch components individually, generate refined renderings for each component, and compose them to pre-visualize concepts. This pre-visualization then serves as input for generating high-quality renderings. We first interviewed interior designers to identify key challenges. We next developed a proof-of-concept system and conducted a pilot study to test its usefulness. We finally improved the system based on additional requirements and conducted a formal user study, demonstrating that our simple yet effective workflow supports designers in exploring, composing, and refining cabin designs into high-quality results.
ACM CHI Conference on Human Factors in Computing Systems