UGC driven generative 3D urban screens for urban interfaces

Generative artificial intelligence offers new possibilities for producing cultural content for urban public screens, yet systematic empirical research remains limited on how public digital expressions can be translated into locally grounded generative content and evaluated in real public spaces. Using Hubin Pedestrian Street in Hangzhou as a case study, this research develops a workflow integrating UGC corpus analysis, urban image and visual motif selection, generative visual production, corner screen based glasses free 3D presentation, and in situ evaluation. A corpus of 6708 deduplicated Weibo posts was used to identify tourism related urban images in Hangzhou. Structured prompts and a locally deployed Stable Diffusion workflow were then used to generate visual assets, which were further adapted into glasses free 3D content for the corner screen. Field evaluation included 100 post exposure questionnaires and 12 fixed camera behavioral observation sessions conducted over three consecutive weekdays. Participants reported generally positive evaluations of cultural representativeness, overall satisfaction, and visual comfort, and several post exposure perception dimensions were positively associated with sharing, recommendation, and revisit intentions. Stopping rates on the two screen on days were 17.4% and 16.1%, respectively, both significantly higher than the 3.5% observed on the screen off day (both p < 0.001), while no significant difference was found between the two screen on days. These findings indicate that the UGC driven generative workflow provides a traceable basis for urban cultural visual design and, in this study context, is associated with positive post exposure evaluations and higher public stopping proportions, offering a reproducible framework for generative urban screen design and in situ evaluation.