18: Deepmirror: Enhancing Architectural Design Processes Through Real-Time Image Synthesis With Artificial Intelligence

In this paper, we will describe the design process and premises of DeepMirror, our interactive platform under development that enables the integration of generative image, voice recognition, and computer vision with conventional architectural representation techniques—such as hand drawing, digital 3D modeling, and physical models—generating variations of the user input that could inform subsequent design decisions. However, unlike existing AI applications for architectural image generation that are limited to producing isolated visualizations, our platform establishes an instantaneous feedback loop between user input and system-generated output. We will present how these inputs and outputs are orchestrated within the multimodal DeepMirror platform to enable this circularity and thus foster unconventional experimentation in the early stages of architectural design.