Answers are grounded in this paper and cite their sources.
In professional artistic creation with Generative AI, creators struggle to translate their multi-dimensional, implicit artistic intentions into the linear instructions required by generative models, creating a significant semantic gap. This challenge is particularly acute in traditional Chinese figure painting, which emphasizes sophisticated brushwork, stylistic lineage, and cultural conception. To address this, we present PromptTCP, a domain knowledge-driven human-AI co-creation system. By integrating a domain-specific knowledge graph (TOBEST-SemKG) encoding six artistic dimensions with a visual“Semantic Canvas” for hierarchical concept organization, PromptTCP transforms prompt engineering from a high-load linguistic task into a co-creative process of conceptual orchestration and visual feedback. A mixed-methods study with 12 participants demonstrates that PromptTCP significantly reduces user frustration and effort while enhancing satisfaction. Furthermore, the generated prompts and images surpass a text-based baseline in richness, cultural relevance, and artistic conception. Our findings provide empirical support for culture-specific AIGC applications and offer a transferable methodology for designing co-creative systems for professional domains.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: