Enhancing Architectural Image Consistency: A Study of Image-to-Image Workflows by Generative AI Models

This study explores the integration of generative AI into architectural design workflows by leveraging the FLUX.1 model and Low-Rank Adaptation (LoRA). It proposes scalable image-to-image (I2I) pipelines that combine semantic image recognition, targeted fine-tuning, and stylistic control using FLUX.1 Redux. Through quantitative evaluation using MSE, SSIM, and PSNR metrics, the approach demonstrates significant improvements in image quality, structural fidelity, and visual coherence. The results highlight the potential of open-source Generative Artificial Intelligence (GAI) technologies to streamline design processes, reduce manual effort, and foster innovative, collaborative design paradigms in architecture.