Comparing Generative AI Models in Spatial in in Spatial Storytelling

Abstract

Generative artificial intelligence tools are increasingly used in architecture and interior design to visualize spatial narratives; however, how different models interpret the same narrative remains underexplored. This study comparatively examines visual outputs produced by multiple generative AI models using identical spatial storytelling prompts. Narratives derived from Engereğin Gözü by Zülfü Livaneli and Parasite directed by Bong Joon-ho were transformed into standardized prompts and visualized through Playground AI, DALL·E, ChatGPT, Stable Diffusion, Runway ML, Leonardo, and OpenArt AI. The images were qualitatively analyzed based on spatial storytelling capacity, text–image coherence, and contextual accuracy. Findings demonstrate clear differences among models in preserving narrative intent and atmospheric consistency. While some models generate visually compelling but contextually limited results, others align more closely with narrative structure and spatial references. Overall, generative AI models reconstruct rather than merely represent spatial narratives, highlighting the importance of informed model selection in AI-assisted spatial design processes.