Nano Banana is the playful name Google gives its Gemini image models: the AI that makes and edits pictures in the Gemini app, Google Search and Google Photos. Google's developer blog introduced the first one on August 26, 2025 as Gemini 2.5 Flash Image, "aka nano-banana", and the name stuck. The sections below explain each Nano Banana model, the story behind the name, common uses, how to try one, and what costs nothing.
What is Nano Banana AI, in plain terms
Nano Banana is a group of Google DeepMind image models that take instructions in ordinary language. You describe a picture or upload one, then keep talking to it about what to change until the result is right. Google builds them on Gemini, so they bring Gemini's general knowledge to pictures: a request for a water cycle diagram or a street in a real city draws on what the model knows about those things.
Editing works through the same conversation. Instead of starting over, you can say "make the jacket red" or "add a second chair", and the model changes that part of the same picture. It can also blend several photos into one image and keep a character looking the same across a series of pictures.
Why is it called Nano Banana?
Nano Banana began as a placeholder. In late July 2025, Google's team needed a public codename for LMArena, a site where people compare answers from two unnamed AI models and vote for the better one. At 2:30 a.m., with the submission due, product manager Naina Raisinghani merged two of her nicknames: friends call her Naina Banana, and some call her Nano. She has said the name suited it because it was a Flash model.
The model showed up on LMArena in early August 2025, and the odd name spread quickly on social media. Google launched it officially as Gemini 2.5 Flash Image, but people kept saying Nano Banana, so Google adopted it, down to a banana emoji on the Gemini app's image chip.
The Nano Banana models, compared
As of September 2026, Google has released these Nano Banana models:
| Model | Official name | Released | What Google built it for |
|---|---|---|---|
| Nano Banana | Gemini 2.5 Flash Image | August 26, 2025 | The first version; Google now recommends 2 Lite to developers replacing it |
| Nano Banana 2 | Gemini 3.1 Flash Image | February 26, 2026 | Pro-level quality at Flash speed; the Gemini app's default |
| Nano Banana 2 Lite | Gemini 3.1 Flash-Lite Image | June 30, 2026 | The fastest, lowest-cost model in the family, for quick drafts and high volume |
| Google's Pro image model | Gemini 3 Pro Image | November 20, 2025 | Precise, detailed work; in Gemini it redoes images that need extra detail, such as infographics |
Google DeepMind's model page lists only the last three. Which of them draws your picture in the Gemini app depends on the model menu, covered in the Gemini image generator guide.
What is Nano Banana 2
Nano Banana 2 is the model the Gemini app uses by default. When Google launched it on February 26, 2026, it replaced the Pro model across the app's Fast, Thinking and Pro settings. What it adds over the original:
- Web knowledge. It can look up live search results, images included, which Google says helps with infographics, diagrams and data charts.
- Text in images. Its lettering comes out readable, and it can translate the wording already in a picture into another language.
- Consistency. It keeps up to five characters and 14 objects looking the same through one workflow, so a set of images can keep the same cast.
- Sizes. Developers can request anything from 512 pixels up to 4K, including the extra-wide and extra-tall ratios 4:1, 1:4, 8:1 and 1:8. In the Gemini app, you download at 1K on the free tier and at 2K on a paid plan.
What is Nano Banana used for
Google's own examples fall into five groups:
- Fixing a photo. Google's launch examples include blurring a background, removing a stain from a T-shirt, taking a whole person out of a shot, changing a pose and colorizing a black-and-white picture.
- Restyling. In Google Photos, Help me edit uses Nano Banana to repaint a snapshot as a Renaissance portrait, a colorful tile mosaic or a children's picture-book page.
- Combining pictures. Merging two photos into one scene, or trying different hairstyles and looks on a selfie.
- Figurine images. Google's Gemini page shows photos turned into figurine-style pictures. The 3D figurine prompt page has copyable versions.
- Infographics and text. Diagrams, labeled posters and signs, including translated ones.
Two prompts that test the text and knowledge side. Paste either one into Gemini or another image generator:
Create a labeled infographic of the parts of a honeybee, with thin leader lines and short labels: "Antenna", "Compound eye", "Wing", "Stinger" and "Pollen basket". Clean scientific illustration on a white background, warm yellow and brown tones, tall 3:4 layout, no other text.A birthday card with a watercolor bouquet of sunflowers and blue cornflowers on cream paper. Across the top in elegant dark green script: "Happy Birthday, Rosa". Soft natural light, lots of empty paper around the flowers, square 1:1 format.There are more Nano Banana prompts, sorted by job, to copy.
How to use Nano Banana in Gemini
The Gemini app is the simplest way to use Nano Banana, and the steps match on a computer and a phone.
- Sign in at gemini.google.com, or open the Gemini app on Android, iPhone or iPad.
- Open Images from the sidebar or menu, or choose Create images in the tools menu.
- Leave the model menu on any setting except Flash-Lite to get Nano Banana 2. Pick Flash-Lite for quick drafts with 2 Lite.
- Tap a template, or write a request that starts with an action word like create, generate or draw.
- To change a photo, upload it and describe the edit. Photo editing requires age 18 or older.
Downloads, Redo with Pro and linking Google Photos have their own steps in how to use Gemini's image generator.
Other places to use Nano Banana
Google has put Nano Banana models into many of its products, so you may already have it:
- Google Search. AI Mode and Lens, in the Google app and in mobile and desktop browsers.
- Google Photos. Help me edit for restyling, and AI templates for things like a professional headshot or a winter holiday card, which started on Android in the US and India for users 18 and older.
- Flow. When Nano Banana 2 launched in February 2026, Google made it the default image model in Google Flow and priced it at zero credits for all Flow users.
- NotebookLM, Google Ads and Stitch. Google listed these among the products getting Nano Banana 2 Lite.
- For developers. Google AI Studio, the Gemini API and Google Cloud. Using Nano Banana 2 in Google AI Studio needs a paid API key.
For other tools that edit and generate images, see Nano Banana alternatives.
Is Nano Banana free?
Yes, within limits. Nano Banana is free in Gemini for anyone signed in: Google's plan table lists Nano Banana 2 image generation on the no-cost tier as well as on every paid one. To use Nano Banana for free, sign in to Gemini with a personal Google Account. Generating pictures is open from age 13 (or your country's minimum), and editing from 18.
The allowance is the catch. Gemini's limits are compute-based, so pictures use them up faster than text chats, and they refill every five hours up to a weekly ceiling. Nano Banana 2 images also have a daily quota, and free downloads come in at 1K. As of September 2026, Google AI plans start at $4.99 a month in the US, and every paid tier raises the limits, lifts downloads to 2K and adds Redo with Pro. College students can get AI Pro free for one year, and Google says limits can change without notice.
Outside the Gemini app, Google launched Nano Banana 2 in Flow at zero credits, while building with it in Google AI Studio takes a paid API key.
How to check whether Nano Banana made an image
Pictures from the Gemini app normally come with two marks: a visible watermark and SynthID, an invisible one. Google said in November 2025 that images made by Ultra subscribers would drop the visible mark, so SynthID is the more dependable sign. Anyone can test a suspect image by uploading it to Gemini and asking whether Google's AI produced it. Google pairs SynthID with C2PA Content Credentials, an industry format that records how a file was made or edited, and Nano Banana 2 Lite output carries SynthID as well.
Try an idea in Image Generator
If you'd rather work outside Google's apps, Image Generator in Toybox AI makes one image per prompt from the same kind of plain description, including the honeybee and birthday card prompts on this page. Paste a prompt into the "Describe your image" box, then attach up to 3 photos if you want to steer the style or layout. It offers five shapes, from Square to Wide (16:9), and a "More options" switch called "Enhanced quality" for pictures with words.
Each image is currently 50 credits, or 95 with "Enhanced quality" on. A new account opens with 20 free credits, which won't buy a single image, so check Toybox pricing before you start. If you add a photo of a person as a reference, the face in the result may not match exactly, so check the likeness before sharing.