EditPanorama: An Innovative Approach to Indoor Space Renovation Design Interaction and Presentation Through Diffusion Models

With the accelerating pace of urban renewal in China big cities, the demand for renovating existing residential indoor space has surged, while traditional design workflows face bottlenecks in creative expression and client communication efficiency. Generative artificial intelligence (GenAI) technologies, such as Stable Diffusion models, offer novel approaches for design visualization through rapid text-to-image responses. However, their limitations in architectural 3D spatial coherence and real-time interactivity remain unresolved. This paper proposes EditPanorama, an interactive architectural interior design framework that integrates Panoramic Neural Radiance Fields (PERF) with Stable Diffusion model, achieves rapid and high quality generation of spatial renovation proposals. Experimental results demonstrate that EditPanorama preserves architectural structural constraints while enabling real-time adjustment of design elements through natural language instructions and facilitating multi-scheme comparisons within seconds. In case studies of aging residential renovation projects, the framework significantly shortened design feedback cycles and improved client satisfaction compared to traditional workflows.