Design and Product Development

Integration of Large Language Models in Fashion Storytelling for Crowdfunding Campaigns

---

Abstract

Crowdfunding in fashion offers a co-creative service ecosystem, addressing challenges such as product oversupply, uncertain demand, and environmental concerns. Successful crowdfunding campaigns leverage compelling narratives to mobilize resources and engage supporters. This study investigates a framework for applying large language models (LLMs) in crafting entrepreneurial stories within crowdfunding campaigns, focusing on the collaboration between fashion domain experts and IT professionals. A content analysis of recent Korean crowdfunding campaigns and two rounds of expert interviews—comprising fashion R&D specialists and LLM technology experts—identified key requirements for fashion-focused LLMs. These requirements informed a six-stage LLM modeling process: data collection and preprocessing, pre-training, fine-tuning, reward modeling, reinforcement learning, and aligned LLMs construction. This collaborative process ensures that LLMs accurately reflect the unique values and complexities of fashion storytelling. Findings highlight the critical role of fine-tuning and reinforcement learning stages, where fashion experts and IT professionals jointly integrate domain-specific knowledge into LLM development.

Keywords: fashion product, storytelling, crowdfunding campaigns, large language model, AI, Storytelling

How to Cite:

An, H., choi, y., Kim, S., Ji, D., Kim, S. & Park, M., (2025) “Integration of Large Language Models in Fashion Storytelling for Crowdfunding Campaigns”, International Textile and Apparel Association Annual Conference Proceedings 81(1). doi: https://doi.org/10.31274/itaa.18561