This study examines the automated creation of spatial visualizations for interior design, emphasizing user preferences over precision. Recognizing design as a reflection of personal identity, we utilize domain-specific, image-fine-tuned AI models to capture the qualitative aspects of various design styles. In interior architecture, design styles are often categorized by shared visual features—like material use, color combinations, and furniture arrangement—based on tacit consensus rather than explicit data. These features significantly impact both the aesthetic and functional aspects of spaces, influenced by historical, cultural, and personal factors. We advanced the field with a text-to-image model that translates descriptive text into visual representations. An extensive evaluation of the default model was conducted, generating over 15,000 images across 25 design styles, which informed the subsequent integration of detailed design knowledge into the model’s training. The refinement process included data preparation, textual alignment with image content, and hyperparameter optimization to develop fine-tuned models. Implemented across multiple scenarios, this approach proved successful in combining the nuanced models with the default, creating images that align with user-defined styles. This methodology serves as a tool for generating spatial visualizations that align with user requirements, providing a range of styles that cater to diverse preferences. It highlights the potential of AI in enhancing design visualization and the shift towards personalized, user-centric design solutions.