AI Design Face-Off: An Interior Designer's Test of Google's Gemini vs. OpenAI (www.fcilondon.co.uk)

FCI London put three leading image generators—Google’s recently revealed Gemini 2.5 Flash Image (nicknamed “Nano Banana”), OpenAI’s Sora, and ChatGPT’s image model—head-to-head using identical prompts and seven luxury furniture pieces to see which tool could produce true, magazine-quality interior visualisations. The practical tests showed Sora clearly led: it delivered realistic proportions, coherent spatial relationships, natural lighting and “designed” compositions that work for client-facing visuals. ChatGPT was useful for early ideation but fell short on dimensional accuracy and polish. Nano Banana, despite impressive marketing and anonymous buzz on LMArena, repeatedly altered the supplied products, misread spatial cues (struggling with open-plan instructions) and offered limited aspect-ratio control. Technically, the comparison highlights a trade-off between iterative consistency and precise spatial reasoning. Nano Banana’s strengths—multi-image fusion, character consistency and text-based editing—make it promising for mood boards, cross-iteration coherence and style exploration, but its current spatial-awareness limits its use for exact product placement or architectural fidelity. Sora’s superior geometry and lighting handling points to better internalised scene understanding, a crucial capability for professional interior work. The takeaway for designers: AI is an accelerating design assistant, not a replacement—tools that combine robust multi-image handling with reliable spatial models will be the most valuable for high-end visualisation.