This paper explores the synthesis of house floor map layouts through the implementation of the Stable Diffusion model augmented with ResNet50 and DINO to enhance and guide image quality. The generation of architectural layouts is inherently complex due to issues such as spatial efficiency functional zoning and aesthetic coherence. These challenges demand innovative methodologies to overcome conventional constraints and open creative possibilities in floor layout design. The proposed approach leverages the generative capabilities of the Stable Diffusion model which is particularly suited for high-resolution and contextually coherent image generation tasks making it an ideal candidate for architectural visualization. The integration of ResNet50 and DINO enhances the system’s perceptual understanding by enabling detailed comparisons and refining outputs to align with both structural logic and artistic vision. This project provides a powerful tool to architects and designers facilitating the automated creation of diverse and optimized floor layouts tailored to specific needs and preferences. The employment of state-of-the-art diffusion models in tandem with advanced feature extraction frameworks like ResNet50 and DINO signifies a notable advancement over traditional CAD systems and heuristic design tools. Furthermore this study discusses the applicability of complementary technologies including CLIP-based text-to-image models and autoencoder-driven refinement pipelines to enhance the creative and practical utility of the proposed system. This interdisciplinary convergence not only enriches the design process but also underscores the potential of machine learning models to revolutionize the field of architectural planning and visualization.