IN-PERSON London. Section A

Abstract

Since 2022, there has been an extraordinary surge in AI image generators, showcasing remarkable innovations such as DALL-E, Midjourney, and Stable Diffusion. Among these advancements, the Generative Fill function within AI image generators has emerged as a groundbreaking tool for quickly visualizing design concepts, underscoring its practicality and potential for application in participatory urban design. This research seeks to address current challenges in AI-participatory urban design, focusing on two main issues: the visibility of imported streetscape and the lack of professional logic in AI-generated designs. By proposing criteria for evaluating visibility to maintain information integrity when importing 2D media into AI image generators, and emphasizing the importance of continuity and coherence in shaping urban environments and public perceptions through Urban Design Qualities, we aim to provide a framework for users to engage in AI-participatory urban design. Additionally, we link the Physical features involved in “Urban Design Qualities” with “The Image of the City”, aiming to establish a classification system and propose a conceptual framework for the operation process of participatory design, providing reference for users in the implementation and development of AI tools. This elevates “Generative fill” beyond a mere image stitching application, ensuring certain professional standards and logic while making it more readily accepted by the public. This helps bridge the cognitive gap between the public and professionals in participatory design processes, ultimately enhancing the role of artificial intelligence in future practices.