To address the challenges of detail loss, stylistic generalization, and limited controllability in the digital generation of Chinese intangible cultural heritage “shadow puppetry”, this study proposes a generative pipeline developed through self-directed inquiry-based learning. The pipeline employs Flux.1 as the base model and integrates Low-Rank Adaptation (LoRA) fine-tuning within a ComfyUI-based workflow. Experimental results demonstrate that the proposed approach enables high-fidelity and controllable generation of shadow puppetry styles. Furthermore, an analysis of the learning process indicates that systematic tool application and reflective technical journals contribute to the development of learners’ computational thinking, cognitive transformation, and complex problem-solving abilities. This research not only presents a feasible technical solution for the digital preservation and creative reinterpretation of specific intangible cultural heritages but also provides an operable practical paradigm for empowering digital humanities education through AI-generated content (AIGC) technologies.
Zhang et al. (Tue,) studied this question.