Participatory Landscape Design and Tool Innovation through Natural Language Processing:

User participation in spatial design is often limited by insufficient design knowledge, low proficiency with professional software, and constrained design options that fail to meet personalized needs. This study proposes an AI-based participatory landscape design workflow that integrates natural language processing to capture user intent, provide design knowledge, and expand design possibilities. We implement the workflow using Whisper, GPT-4o, and Stable Diffusion, and validate it through a landscape renovation case study in the Qingnian East Road Community, Jinan, China. System testing shows that NLP-driven interaction can effectively support non-expert users: the average Conversations Per Session (CPS) reached 13 and over 85% of user–agent exchanges were assessed as effective. The system reduced operational barriers and offered knowledge-informed suggestions (e.g., spatial configuration and material selection). However, the text-to-image module was constrained by the style and scale limitations of the current Stable Diffusion model. Future work will develop domain-specific models to better support diverse, personalized landscape design needs.