Nano Banana is the informal, community-assigned name for an image generation and editing model developed within Google's Gemini family, which became widely known under that nickname before Google publicly confirmed its identity. The model was later formally released and marketed as Gemini 2.5 Flash Image.
Origins of the name
The model first appeared anonymously as a mystery entry on LMArena, the community model-comparison platform, where new systems are sometimes tested under codenames before an official announcement. Users on the platform and social media began referring to the unidentified model as "nano-banana" based on its listed codename, and the nickname stuck through weeks of speculation about which lab and which model version was behind it. When Google DeepMind confirmed the model's identity in 2025, it embraced rather than dropped the nickname in some of its own marketing and developer communications, an unusual instance of a major lab adopting community-generated branding for a serious product release.
Capabilities
Nano Banana distinguished itself from earlier Text-to-image generation and image-editing tools through strong instruction-following on edits to existing photographs, including consistent identity preservation of people and objects across multiple edited outputs, a capability many prior diffusion-based editors handled poorly. It could perform targeted edits described in natural language, such as changing an object's material, altering lighting, or compositing elements, while keeping the rest of an image visually consistent, and it supported multi-image inputs, allowing several source photos to be blended into a single generated result.
The figurine trend
The model became a mainstream viral phenomenon in 2025 through a specific use case: transforming a photo of a person into an image resembling a collectible action-figure or toy figurine, complete with blister packaging, accessories, and a display box, rendered with photorealistic materials and lighting. The trend spread rapidly across social platforms, with millions of users generating and sharing figurine-style self-portraits, drawing comparisons to earlier viral AI image trends such as Studio Ghibli-style portraits generated with OpenAI's image tools. The trend's scale illustrated both the improved photorealism and identity-consistency of contemporary image generation models and the continuing appetite for accessible, shareable AI-generated content, a dynamic also discussed in the context of platform virality and broader concerns about AI-generated content flooding social feeds.
Reception and significance
Nano Banana was widely covered as a milestone for consistency and controllability in AI image editing, an area where earlier models, including versions of Midjourney and FLUX, had been comparatively weaker relative to their strength in pure text-to-image generation. Its integration into Google's broader Gemini app and API made high-fidelity, consistent image editing accessible to a mass consumer audience rather than remaining a specialist tool, and its release intensified competitive pressure on other image-generation labs to match its editing precision. The episode also highlighted a recurring pattern in AI releases during this period, in which unbranded or anonymously tested models generate significant public interest and speculation on evaluation platforms before their developer is officially confirmed.