Generative Images: Integrating Ai Into Architectural Narrative

This study explores the integration of AI in undergraduate architecture education, specifically through the use of diffusion model technologies like Midjourney, for generating images from text, a process commonly known as "text2image". Over a 10-week course, fourth-year architecture students tested a workflow that integrates AI in the visual narrative production of architecture. The exploratory-descriptive methodology employed a case study where students used textual prompts to generate preliminary images and developed strategies for customized datasets. Preliminary findings suggest that AI integration in architectural design can enrich initial creative processes by facilitating the rapid exploration of multiple visual alternatives. Furthermore, developing a systematic vocabulary for composing descriptive texts or prompts can enhance skills in conceptualizing and communicating specific spatial ideas. However, challenges were identified in the spatial consistency of the generated images, highlighting the importance of critical thinking in interpreting and improving AI-generated results.