Nano Banana

basicwidespread

Nano Banana is the nickname for Google Gemini's image generation and editing model: it draws images from text prompts or transforms your existing photos.

In plain terms

Nano Banana is an image model belonging to Google's Gemini family. It is like a very fast graphic designer: you simply tell it what you want, such as 'move my dog to the seaside' or 'change the coat to red', and it does it.

Its greatest strength is editing, not drawing from scratch. You upload a photo and request changes using everyday sentences, while the subject's face and the rest of the image remain recognizably the same.

With earlier image generators, a different face often appeared with every new request. Nano Banana better maintains the same character and object across multiple images, can merge multiple photos, and its newer versions also add readable text to the image.

The name was originally a test codename, but it stuck so well that Google officially adopted it.

How it works

You describe in text what you want, and if needed, upload one or more images.
The model interprets text and image simultaneously because the same Gemini system processes both. That is why it understands requests like 'leave the person on the left, change the background.'
It generates a new image where only the requested parts change, while striving to preserve the rest.
If you do not like the result, you refine it through conversation: "make it darker", "bring it closer". You do not need to start everything over.
Google places an invisible watermark (SynthID) on the generated image, indicating that the image was created or modified by AI. Newer versions may also supplement this with a standard provenance tag (C2PA) attached to the image.

An everyday example

You photograph your living room, upload it to the Gemini app, and ask it to show you a version with a green sofa and lighter walls. Within seconds or minutes, you see the new version of your room before buying anything.

Who is affected

everyoneAnyone can transform a photo with a few sentences, so an image alone is increasingly less proof that something happened.

marketerCreate product photos, ad variations, and social media images in minutes without a photoshoot. → Marketing

entrepreneur, executiveAs a small company, you can obtain acceptable-quality visual material cheaply, but you must pay attention to legal and authenticity issues.

parent, student, teacherUseful for illustrative graphics and projects, but also increases the risk of misuse of photos of children. → Education & Family

A common misconception

MythNano Banana is a standalone application that must be downloaded separately.

In realityThis is a nickname for a model, not a separate program: you access it through the Gemini app, other Google products, and developer interfaces. Behind the name, there are also several successive versions.

Limits and risks

The image may be convincing but not necessarily accurate: small details, hands, captions, and the data in maps and charts can be wrong.
Creating misleading images of real people with it is easy, and this can have legal and ethical consequences.
The invisible watermark is not a perfect defense: not every platform checks it, and the marking can be damaged when the image is heavily altered. Moreover, it only marks images created with Google's own tools, not forgeries made with other tools.
Your uploaded photos go to the provider's server, so do not upload sensitive images or other people's personal photos thoughtlessly.

Similar, but not the same

Gemini – Gemini is Google's full AI model family and assistant; Nano Banana is only its image generation and editing part.
Imagen – Imagen is Google's dedicated text-to-image model, while Nano Banana is part of Gemini and excels mainly in conversational image editing.
Deepfake – Deepfake is the term for deceptive, falsified content, while Nano Banana is a tool capable of creating both harmless and deceptive images.

Related concepts

Broader

Generative AI, Image generation, multimodal model

Narrower

Nano Banana Pro, Nano Banana 2

On the edge

Gemini, Imagen, Diffusion model, Deepfake, SynthID

A brief history

The model appeared anonymously in August 2025 under the cover name "nano-banana" on the LMArena model comparison site, where it quickly rose to the top. Google officially released it that same month as Gemini 2.5 Flash Image. In November 2025, the Nano Banana Pro based on the Gemini 3 Pro model was released, which, among other things, writes text appearing in images much more accurately. At the beginning of 2026, the faster, cheaper Nano Banana 2 appeared, bringing most of the Pro version's capabilities to the Flash series.

Where things stand

Nano Banana is the built-in image model of the Gemini app and several Google products, with multiple successive versions. Due to mass usage, the issue of image authenticity and labeling is also on the agenda.

Status: 2026 November

Also known as

Nano Banana · nano-banana · Gemini 2.5 Flash Image · Gemini Image · Nano Banana Pro · Gemini 3 Pro Image · Nano Banana 2 · Google image generator

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.