Research on AIGC-Based Generative Design of Jiaotai Porcelain Pattern Styles

Abstract

With the increasing demand for digital preservation of intangible cultural heritage and intelligent design, the transformation of traditional Jiaotai Porcelain patterns in contemporary design faces several challenges, including limited innovation in visual expression, heavy reliance on expert experience in the design process, and difficulty in controlling stylistic continuity. To address these issues, this study proposes a Kansei Engineering–based framework that integrates artificial intelligence–generated content (AIGC) for style classification and generative design of Jiaotai Porcelain patterns.First, a card sorting method was conducted with an expert panel to systematically organize the formal characteristics and affective impressions of Jiaotai Porcelain patterns. Based on this analysis, a structured style recognition system was constructed, and the patterns were categorized into four distinct style types. Second, a controllable generation pipeline was developed by fine-tuning a Flux model using Low-Rank Adaptation (LoRA), combined with a ComfyUI workflow to enable flexible and controllable pattern synthesis. Finally, a fuzzy comprehensive evaluation method was introduced for empirical validation, assessing the generated results in terms of stylistic consistency, product feasibility, and image completeness.The proposed approach provides not only an operational analytical framework for the study of Jiaotai Porcelain pattern styles, but also a reusable methodological and technical pathway for design research on intangible cultural heritage within the context of AIGC.