Fashion Prompt Engineering: Analyzing and Deriving Components of Fashion Prompts through Text Mining

This study conducted text mining on fashion prompts and derived their components to develop a systematic approach to fashion prompt engineering, with the aim of producing results that clearly reflect the designer&quots intent through human-AI interaction in generating fashion images. As a research method, text mining was performed on 947 fashion prompts shared in a text-to-image AI online database, with the components of the fashion prompts appropriately organized. The results of this study are summarized as follows: First, word frequency and TF-IDF analysis showed that words related to high-fashion photos capturing full-body shots of fashion models or women; highly detailed images on the theme of complex fashion and realistic depictions; image quality; and MidJourney parameters were frequently used in fashion prompts. Second, network visualization of a CONCOR analysis revealed four clusters: &quotwomen&quots fashion and style,&quot &quotmen&quots fashion and style,&quot &quotphotorealistic shooting,&quot and &quotfashion portrait.&quot It was also found that fashion prompts use words to generate photorealistic images and fashion portrait images containing elements of women&quots or men&quots fashion styles as well as elements of shooting. Although there are elements of both women&quots and men&quots fashion styles, it was confirmed that the expression of men&quots styles often involves elements of photorealistic image expression. Third, the components of fashion prompts are broadly categorized into perceptual and technological aspects. The perceptual aspects include subject, model, fashion, background and setting, style, atmosphere, and tone, while the technological aspects include medium, layout, depth and camera, lighting, quality and resolution, and ratio components. This study serves as foundational research for effective fashion prompt engineering, considered as communication between humans and AI. It provides basic data for optimized prompt design towards creative fashion design generation and can be used as educational material for AI-integrated fashion education; moreover, it is expected to contribute to future fashion idea generation and design development.