Research-Acoustic Urbanism of Generating Spatial Scenes from Sound

Our project explores how urban soundscapes can act as generative drivers for spatial and visual reconstruction. Using street audio recordings from multiple cities, the system translates acoustic signals into visual representations through a cross-modal pipeline. Audio features are extracted using pretrained models and transformed into conditioning inputs for diffusion-based image generation, enabling dynamic visual outputs driven by sound. The project investigates how different acoustic environments influence spatial perception, comparing urban contexts across cities such as London, Beijing, and Singapore. By linking sound to visual structure, it reveals how auditory cues contribute to the identity and interpretation of urban space. Positioned between generative AI and spatial design, the work reframes sound as an active medium in shaping and reimagining urban environments.