Pusan National University Researchers Explore How Generative AI Can Streamline Fashion Design
Generative AI models such as ChatGPT and DALL-E offer promising ways to improve efficiency in fashion design and to identify emerging trends. By analyzing data patterns, these AI tools can generate new text and images that assist designers in creating fresh fashion catalogues, helping accelerate product development.
Large language models (LLMs) like ChatGPT, alongside AI image generators like DALL-E, have demonstrated value across various industries. In fashion, LLMs can analyze past styles and forecast upcoming trends, generating detailed prompts that image generators use to visualize new collections. Integrating AI effectively into fashion design requires a clear understanding of these model capabilities and limitations.
Study Overview
Researchers from Pusan National University’s Department of Clothing and Textiles, Professor Yoon Kyung Lee and Master’s student Chaehi Ryu, recently investigated how generative AI can visualize seasonal fashion trends. Their study, published in the Clothing and Textiles Research Journal (June 22, 2025), focused on the role of precise prompt engineering for generating realistic fashion images with AI.
Using ChatGPT-3.5 and ChatGPT-4, the team analyzed men’s fashion trends up to September 2021 and predicted fall/winter 2024 trends. They categorized design elements from these predictions, Vogue’s 2024 Fall/Winter Men’s Fashion Trend data, and fashion literature into six groups: trends, silhouette elements, materials, key items, garment details, and embellishments.
Generating Fashion Images with DALL-E 3
Based on these categories, the researchers crafted 35 detailed prompts for DALL-E 3, each describing a unique male outfit for a 2024 Fall/Winter runway show. These prompts included variables such as aspect ratios, camera angles, model appearance, runway design, and mood. Each prompt was run three times, producing 105 images in total.
DALL-E 3 accurately implemented the prompts 67.6% of the time, with prompts containing descriptive adjectives performing best. Some generated images closely resembled actual 2024 Fall/Winter men’s collections. However, the model struggled with certain trend elements like gender fluidity, and many images leaned toward ready-to-wear fashion. The study found that trend keywords alone were insufficient for precise image generation, highlighting the importance of expert input.
Implications for Fashion Design
The findings emphasize the critical role of expertly crafted prompts and fashion expertise in achieving accurate AI-generated designs. With ongoing improvements, generative AI tools like DALL-E 3 could support designers by speeding up the creation of complete collections and enhancing creative possibilities. Additionally, these tools can help non-experts better understand and engage with fashion trends.
This research illustrates how generative AI can serve as a valuable assistant in the design process, making it easier to visualize and experiment with upcoming fashion seasons.
About Pusan National University
Founded in 1946, Pusan National University in Busan, South Korea, is recognized as the leading national university in research and education. The multi-campus institution includes sites in Yangsan, Miryang, and Ami, hosting approximately 30,000 students and over 1,200 professors. The university operates 14 colleges and 103 departments, committed to principles of truth, freedom, and service.
About Professor Yoon Kyung Lee
Assistant Professor Yoon Kyung Lee specializes in creativity and sustainability within fashion design, focusing on AI, digital technology, and neuroscience. She holds an MSA and Ph.D. in dress aesthetics from Seoul National University. Her dissertation examined Eastern and Western philosophies through the “Hyperspace Paradigm.” With experience as a Milan-based designer and founder of the brand UginiO, she has showcased at Seoul Fashion Week and Prêt-À-Porter Paris. Professor Lee has also taught at major Korean universities and held postdoctoral and visiting scholar roles at the University of Minnesota.
Reference
DOI: 10.1177/0887302X251348003
Your membership also unlocks:
