Generative Large Model-Driven Methodology for Color Matching and Shape Design in IP Products

The rise in generative large models has gradually influenced traditional product design processes, with AI-generated content (AIGC) playing an increasingly significant role. Globally, tourism IP cultural products are crucial for promoting sustainable tourism development. However, there is a lack of practical design methodologies incorporating generative large models for tourism IP cultural products. Therefore, this study proposes a methodology for the color matching and shape design of tourism IP cultural products using multimodal generative large models. The process includes four phases, as follows: (1) GPT-4o is used to explore visitors' emotional needs and identify target imagery; (2) Midjourney generates shape options that align with the target imagery, and the optimal shape is selected through quadratic curvature entropy method based on shape curves; (3) Midjourney generates colored images reflecting the target imagery, and representative colors are selected using AHP and OpenCV; and (4) color harmony calculations are used to identify the best color combination. These alternatives are evaluated quantitatively and qualitatively using a color-matching aesthetic measurement formula and a sensibility questionnaire. The effectiveness of the methodology is demonstrated through a case study on the harbor seal, showing a strong correlation between quantitative and qualitative evaluations, confirming its effectiveness in tourism IP product design.