How to Use Gemini Nano Banana: Create and Edit Images

Nano Banana is Google’s name for the image models inside Gemini, and learning how to use Gemini Nano Banana starts with one fact: there is no single model called Nano Banana. Five models share the name, and the one that draws your picture depends on the Gemini model you picked before you asked. Get that choice right and the rest is simple. Open the Gemini app, sign in with a Google account, pick Create images, and describe the picture you want in plain words. To change a photo, upload it and describe the change the same way.

Which Nano Banana model are you using?

Google’s developer docs list five models under the Nano Banana name. The Gemini app hides most of this behind its model picker, so the table below maps what you see in the app to what is actually drawing the image.

Model you get	Gemini model setting	What it is for
Nano Banana 2 Lite	Flash-Lite	Quick images and basic ideas. Fastest and cheapest. Not built for several reference images or repeated edits.
Nano Banana 2	Flash or Pro	The all-rounder. Better quality, more world knowledge, steadier text. Accepts several reference images.
Nano Banana 2.1	Rolling out in the app	The current workhorse. An update to Nano Banana 2 with better visual quality, mask editing and steadier faces between edits.
Nano Banana Pro	Pro, then “Redo with Pro”	The premium option. Paid Google AI plan only. Best for text-heavy images and infographics.
Nano Banana (original)	—	The legacy model from August 2025. Google’s docs recommend moving to 2 Lite instead.

Two things in that table trip people up. First, Google’s help page describes Nano Banana 2 as the default in the app, but its developer docs now call Nano Banana 2.1 the primary workhorse and recommend it for new projects. Google announced 2.1 on October 6, 2026, and the rollout is gradual, subscribers first — so your app may still run Nano Banana 2 while someone else’s already runs 2.1. Second, Nano Banana Pro is not in the model picker at all. Paid subscribers reach it by making an image first, then choosing More and “Redo with Pro”.

To pick a different model, switch your Gemini model before you send the image request. If you only see Flash-Lite in your picker, you are on Nano Banana 2 Lite. Reports in October 2026 said Google moved free Gemini accounts to Flash-Lite, which would put free users on 2 Lite by default; other guides written the same week say Flash is still open to free accounts. Google’s help page describes the mapping without tying it to a plan, so check your own picker rather than trusting either report.

How to use Gemini Nano Banana in the Gemini app

The whole workflow is a conversation: prompt, attach, refine, download. It works the same on the web at gemini.google.com and in the Gemini app on Android and iPhone.

1. Sign in to Gemini. Open the Gemini app or gemini.google.com and sign in with a Google account. You need to be 13 or over to generate images, and 18 or over to edit them.
2. Pick your model first. If you want Nano Banana 2 rather than 2 Lite, set the Gemini model to Flash or Pro before you start. This is the step most people skip, and it decides the quality of everything after it.
3. Ask for an image. Pick Create images from the tools menu, or type a request that starts with “Create an image of…”. Gemini sends the job to the Nano Banana model behind your setting.
4. Refine in follow-up messages. Keep the same chat going: “warmer light”, “zoom out a little”, “make the headline bigger”. Each turn builds on the latest image. Nano Banana 2 and 2.1 hold a face steady across those turns; 2 Lite is not built for repeated edits, so expect it to drift.
5. Download the result. Hover over the image and click Download full size. Without a Google AI plan, downloads come out at 1K resolution; with a plan, they come out at 2K.

Choose the aspect ratio before you generate — the ratio menu sits under the message box. A square image cropped into a banner afterwards loses the edges you actually wanted.

How to edit a photo with Nano Banana

Editing is where Nano Banana earns its reputation. You describe the change in ordinary sentences — no brushes, masks or layers. Google’s help page lists three kinds of edit: change an image Gemini made, upload your own photo and ask for changes, or upload several photos and ask Gemini to combine them into one.

Start with one change per message. A prompt like “Remove the boxes on the floor and the cables behind the TV. Keep everything else exactly as it is” gives the model one job and tells it what must stay untouched. That second sentence does real work: name what should stay the same, or the model will “improve” parts of the photo you liked. Once the first edit lands, send the next change in the same chat — “Now paint the back wall a soft sage green. Keep the lighting, the furniture and the window unchanged.” Chain edits this way and you can tell which instruction caused a problem when something goes wrong.

Blending photos works the same way. Upload two or more pictures and ask for one new scene with both in it. Nano Banana 2 accepts several reference images; 2 Lite does not handle this well. If the blend keeps dropping one subject, switch to Flash or Pro mode and try again.

To go back to an earlier version, find the image in your Library and open it in a chat. If an image is nearly right but the text is wrong or the detail is thin, paid subscribers can redo it with Nano Banana Pro from the More menu. One catch: if you have already hit your daily Nano Banana 2 quota, you cannot redo any more images with Pro either.

Prompting that actually works

Google’s own help page gives the prompting advice in four lines: start with “draw”, “generate” or “create”; name the style; describe the subject, what it is doing, and the background. A working formula is subject, then details, then style, then light, then constraints — the same five slots from our guide to writing AI image prompts.

Compare these two prompts for the same product shot:

• Weak: “A beautiful coffee mug, amazing, aesthetic, high quality, stunning.”
• Strong: “Create a product photo of a ceramic coffee mug with the word LAB printed on it, on a wooden desk next to a green plant, soft morning light. Photorealistic, no text except LAB.”

The strong version says what the object is, where it sits, what the light does, and what style to use. It also puts the exact word in capitals, which helps the model spell it right. For edits, the same rule holds: say what changes, then say what stays. “Edit this image: make the mug matte dark green, with the text LAB in white. Keep everything else exactly the same.”

Small text is the known weak spot. Short words in capitals usually come out right; small lettering often does not. Google lists small text as a limitation, so zoom in and check every character before you publish an image that has to carry words.

What is free and what costs money

Nano Banana is free in the Gemini app — Google’s own image page says “Available for free” — but free does not mean unlimited.

The Gemini app gives free accounts a daily quota of Nano Banana images, but Google does not publish the number for the app. Its limits page says the app runs on compute-based limits, which weigh how complex your prompt is, which models and features you use, and how long the chat gets. When the quota is gone, generation stops until it resets, and Pro redos stop too. Two numbers Google does publish for its other surfaces: AI Mode in Google Search allows up to 25 images in 24 hours, and Flow allows up to 20 Nano Banana Pro images a day as of November 1, 2026. Those are the figures for those surfaces, not for the Gemini app.

What costs money: a Google AI plan. It raises the limits, gives 2K downloads instead of 1K, and opens Nano Banana Pro redos. The paid Gemini API has no free tier for any of the five models — images through the API are charged per image, with reports putting Nano Banana 2 Lite at $0.034 per 1K image.

The Google Photos connection, which lets Gemini reach your own photos for edits, is US-only and 18-plus, and the Personal Intelligence features that tailor images to you are US-only as well.

Limits worth knowing before you rely on it

A few limits only show up after you have built something on Nano Banana, so here they are up front.

Every image Gemini makes carries a visible watermark and an invisible one called SynthID, a signal hidden in the pixels that marks the picture as AI-made. A Media Watermark setting under Settings on gemini.google.com can turn the visible watermark on or off, but SynthID stays. In the EU, machine-readable AI labelling has been a legal requirement since August 2026.

The daily quota is real but unpublished for the app, as covered above. The 2.1 rollout is gradual, so two people on the same plan can be running different models on the same day. Nano Banana 2 Lite cannot do multi-turn edits or multiple reference images, so a workflow built on those features breaks if your account is on Flash-Lite. And the Pro redo is a redo, not a starting choice: you cannot pick Pro from the model picker and generate with it directly.

Google’s terms forbid some uses outright, including sexual images of people made without their consent, and the help page asks users to use discretion before relying on or publishing what Gemini generates.

When Nano Banana gets it wrong

Most failures have boring causes:

What you see	What to do
The image looks cheap or ignores your references	Check the model picker. You may be on Flash-Lite (Nano Banana 2 Lite). Switch to Flash or Pro and run the same prompt again.
“Create images” is missing from the tools menu	Update the Gemini app, then check you are signed in to the right Google account. The feature rolls out by account and region.
Images fail to generate at all	Check your connection first — image generation runs on Google’s servers. Then check whether you have hit the daily quota.
The face changes between edits	You are probably on 2 Lite, which is not built for repeated edits, or on the original Nano Banana. Switch to Flash or Pro mode for the edit chain.
Text in the image is misspelled	Short words in capitals work best. Regenerate, or generate the image without text and add the lettering afterwards in an editor.
“Can’t redo with Pro”	Your daily Nano Banana 2 quota is used up. Wait for the reset, or accept the standard version.

If an edit goes wrong in a way you cannot explain, start a fresh chat with the same prompt. Long chats carry more context than you think, and a stale instruction from ten turns ago can steer the next image.

Try one edit tonight: take any photo on your phone, upload it to Gemini, and ask for a single change with the words “keep everything else exactly the same” at the end. If the result keeps the photo recognisable and changes only what you asked for, you are on the right model. If it rebuilds the whole scene, check the model picker first — that one setting decides most of what Nano Banana can and cannot do for you.

Sources

• Google Gemini Apps Help, “Generate & edit images with Gemini Apps” — https://support.google.com/gemini/answer/14286560?hl=en (model mapping, age rules, Pro redo steps, download resolutions, Personal Intelligence, Google Photos connection)
• Google Developer Docs, “Nano Banana image generation” — https://ai.google.dev/gemini-api/docs/image-generation (five-model list, 2.1 as primary workhorse, SynthID on all generated images; fetched 2026-11-07)
• Independent guides read for cross-checks (secondary): novoads.ai “Is Nano Banana Free?” (Oct 26/Nov 3 readings of Google’s pages); aialleyway.com “How to Use Nano Banana” (model mapping, age rules); deyronlabs.com “How to edit images with Nano Banana in the free Gemini app” (Oct 8; small-text limitation); blog.paulbreit.com “Nano Banana 2” (Oct 13; 2.1 release date, free quota reporting); ai-xblog.com “Nano Banana 2.1: API Costs & Migration Guide” (Oct 7; deprecation note, no shutdown date announced); theleveragewire.com (Oct 9; free-accounts-to-Flash-Lite reports, 2 Lite API price)