Within the domain of AI-driven architectural design, the task of faithfully depicting specific architectural styles, especially in the design of building facades, continues to be a substantial difficulty. This work introduces a novel method that combines shape grammars, which is known for representing design rules, with stable diffusion models to tackle this problem. The specific focus of this study is on the Ionic style. The research pioneers a way to generate building facades by merging the technological capabilities of AI models like LoRA and Dream-Booth with the image-generating abilities of Stable Diffusion. This entails a demanding procedure of creating specialized datasets, training AI models using these datasets, and doing a thorough comparative analysis to assure the accuracy and visual authenticity of the designs in the Ionic style. The main contribution of this study is its illustration of how shape grammar may direct AI models to generate architectural facades that are of excellent quality and consistent in style. The revised model shows the ability to create Ionic-style facades with higher accuracy and AI that follows conventional architectural rules. This study highlights the capacity of AI to enhance the visual elements of architectural design, closing the divide between contemporary computational methods and conventional architectural sophistication.