AI for Architects

Course Co-ordinator: Subin Jameel

Drashti Thumar

This elective reframed AI as a conceptual tool for architectural thinking rather than as a production-speed boost. The course, which was organized around Midjourney and ComfyUI, went beyond mere image creation and instead positioned AI as a system that translates words and pictures into spatial propositions. The focus was on comprehending how prompts, parameters, and procedures influence architectural outcomes rather than just software expertise.

One of the course's exercises required students to choose an architectural precedent and intentionally change its meaning by changing the context, materials, or program. This speculative reimagination was not about stylistic exaggeration but about reframing architectural logic. Midjourney imported precedents to severe climates or other metropolitan environments, while ComfyUI workflows like ControlNet Canny and Depth maintained structural integrity during the change in setting. The experiment improved our understanding of how prompts, denoising strength, and edge or depth maps affect spatial continuity. Rather than producing separate visuals, the technique promoted iterative design thinking, with AI serving as a tool for testing spatial hypotheses while maintaining architectural integrity. (One example of the above exercise is attached below)

ComfyUI_Workflow

ComfyUI_Workflow

Finally, the workshop underlined a key point: AI should be viewed as a creative collaborator rather than just an optimization tool. When approached wisely, it changes the way architectural ideas are created, expressed, and developed, broadening the area of conceptual exploration while requiring greater clarity of intent from the designer.