Overview

• DALL-E 3 achieved perfect implementation for 67.6% of the 105 runway images generated from 35 AI-crafted prompts.
• Researchers organized predicted menswear trends into six design codes—trends, silhouettes, materials, key items, details and embellishments—to structure AI outputs.
• Generated images predominantly reflected ready-to-wear styles and often failed to capture nuanced elements such as gender fluidity.
• Higher fidelity correlated with prompts featuring descriptive adjectives, highlighting the necessity of precise, expert-crafted instructions.
• Pusan National University team recommends further model training and refined prompt techniques to enhance inclusivity, accuracy and adoption in fashion workflows.