Nano Banana

Nano Banana is the brand name of a family of image generation and editing models developed by Google DeepMind within the Gemini model line. The name began as an anonymous codename when the model appeared for public testing on the LMArena leaderboard on 12 August 2025, and became official branding when it launched on 26 August 2025 as Gemini 2.5 Flash Image.^[1][2] Rather than a standalone text-to-image system, Nano Banana is a native multimodal capability of Gemini: images are generated and edited conversationally, and the same subject can be kept consistent across multiple turns and images.^[3]

History

Nano Banana drew attention before it had a name, when a mysterious model codenamed "nano-banana" appeared in anonymous testing on LMArena in August 2025. Google released it as Gemini 2.5 Flash Image on 26 August 2025 across the Gemini app and Google AI Studio, with general availability in the Gemini API on 2 October 2025.^[1] Its photo-editing features, including a viral trend of turning personal photos into collectible-style figurines, made it one of the fastest-spreading consumer AI tools of 2025. In November 2025, Google added Nano Banana Pro (Gemini 3 Pro Image), aimed at professional asset production with search-grounded accuracy, an internal "thinking" pass and up to 4K output. Nano Banana 2 (Gemini 3.1 Flash Image) followed on 26 February 2026 and became the default image model in the Gemini app, AI Mode in Search and Google Lens, with Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) arriving on 30 June 2026 as the family's fastest and cheapest tier at about USD 0.034 per 1K-resolution image.^[4][5] Alongside the rollout, Google retired its earlier Imagen models from the Gemini API in favour of the Nano Banana family.^[6]

Key Concepts and Technology

Nano Banana models generate and edit images through Gemini's multimodal architecture. Key capabilities include multi-turn conversational editing, where successive instructions refine a picture without losing the subject; character and object consistency, allowing the same person or product to appear across scenes; multi-reference composition, blending up to fourteen input images on the Gemini 3 generation of models; and search grounding, drawing on Google Search for real-world accuracy in generated scenes. Higher tiers add a "thinking" step that refines composition before generation and 4K output, while lighter tiers optimise latency — Nano Banana 2 Lite produces images in about four seconds. Generated images carry Google's SynthID imperceptible watermark, and the family is available through consumer surfaces including the Gemini app, AI Mode in Search, Photos and NotebookLM, as well as the paid Gemini API with per-image pricing.^[3][4][6]

Applications and Impact

The family is used for social media content, advertising mock-ups, product photography, storyboards and everyday photo retouching, with conversational editing lowering the skill barrier compared with earlier text-to-image tools. Its consumer-first distribution, free to use inside the Gemini app, drove mass adoption, while API tiers let developers build image features into their own products. The family competes with Midjourney, FLUX, Stable Diffusion and Adobe Firefly, and its rise has intensified debate about provenance, watermarking and the use of edited likenesses in advertising.^[3][5]

>See Also

• Gemini
• Text-to-Image Model
• FLUX
• Midjourney
• AI Watermarking