Generative Artificial Intelligence As A Tool for Designing Contemporary Saudi House Facades Inspired By Salmaniya Architecture

This study examines the use of generative artificial intelligence, specifically diffusion models fine-tuned with the Low-Rank Adaptation (LoRA) and DreamBooth techniques, to produce contemporary Saudi housing facades guided by the value-based framework of Salmaniya Architecture. A design-based research methodology was adopted with four integrated phases: curating a local dataset that encodes Saudi identity, privacy, proportional harmony, rhythm, and spatial organization; selecting and fine-tuning the model; crafting prompts and conducting text-to-image generation with experimental deployment; and conducting expert evaluation and validation. A set of facade models was generated and assessed using a multi-criteria matrix spanning cultural, functional, and environmental indicators. Quantitative analysis, including means, standard deviations, and ranking visualized through bar charts with error bars, was integrated with qualitative expert commentary that identified criterion highs and lows through radar plots. The results indicate that generative AI can design culturally consistent, climate-responsive, and constructable facades for enhanced livability, as well as reveal inadequacies of proportional balance and contextual continuity in poorly performing submissions. A number of designs came forward as standards, reflecting a harmonious mix of identity, human-centeredness, and environmental performance. The research contributes to a reproducible framework that integrates artificial intelligence modeling, expert judgment, and standards-based evaluation. Consistent with the King Salman Charter for Architecture and Urbanism, it provides a culture-driven route for Saudi housing development and the wider regional practice.