Start here
How to use Nano Banana to create and edit images
Start with one image and one clear instruction. This guide follows the Gemini web workflow, with separate steps for creating something new and changing a photo you already have.
Before you start
You need a Google account with access to Gemini image features, a supported country and language, and an image you are allowed to upload if you want to edit a photo. Availability also depends on age and account type. Google's current help page restricts image editing to users aged 18 or over.
This walkthrough is based on Google's public Gemini help documentation, checked September 6, 2026. It is not a recording of a signed-in session. Google currently documents more than one interface, so the labels on your account may differ.
The video above is Google's Nano Banana Pro demonstration. It shows the model's published examples, not the current app interface or a result produced by one of this site's prompts.
1. Open image creation in Gemini
- Open gemini.google.com and sign in to your own account.
- Look for Images in the sidebar. In some layouts, open Tools and choose Create images.
- Start a new request or choose an available template. A new conversation makes it easier to understand which instructions affect your first result.
The current Help Center describes Nano Banana 2 as the image-creation model. It also documents a Flash-Lite option using Nano Banana 2 Lite. Do not assume that selecting “Pro” in the text-model menu means your first image was generated by Nano Banana Pro: the paid image redo is a separate action.
No image option? Check Google's image help page for account and regional availability. A third-party site's “free access” button cannot change your Google account's eligibility.
2. Create your first image
Start with an easy-to-inspect scene. Name the subject, environment, composition, and light. Avoid requesting a dozen unrelated styles in the same sentence.
A simple still life
Start with: A new image request; no uploaded photo required.
Create a realistic photograph of a blue ceramic cup beside an open notebook on a wooden desk. Show the scene from a slightly elevated angle. Use soft daylight from a window on the left, with natural shadows. Keep the background simple and include no lettering or logos.
Follow-up correction: Keep the composition and objects unchanged. Make the cup a darker blue without changing the lighting.
This is an original starting example, not a pre-tested output recipe. Submit it, then check whether the image actually matches each part: one cup, one notebook, the requested angle, and light from the left. You can learn more from one clear mismatch than from adding “ultra-detailed” repeatedly.
3. Edit a photo you already have
Upload a suitable image using the attachment control in your Gemini conversation. For an image already in Gemini, Google's current instructions describe opening Library, selecting the image, and choosing Chat.
Ask for a single change and say what must remain unchanged. For example, a background edit should not also change a person's face, clothing, pose, and lighting unless that is really your intention.
Change the background, keep the subject
Start with: A clear photo with one person, used with their permission.
Replace only the background with a plain light-gray wall. Preserve the person's identity, facial expression, hairstyle, clothing, pose, and body proportions. Match the existing light direction and keep the edges around the hair natural. Do not add text or accessories.
Follow-up correction: The hairstyle changed. Return to the original reference photo and preserve the hair exactly; change only the background.
If the model changes too much, return to the original reference rather than repeatedly editing an already altered face. Use a conventional selection or masking tool when exact pixel preservation matters more than creating a plausible new image.
4. Refine, compare, and save
- Choose one correction. Describe the actual problem: the text is wrong, the subject moved, or the shadows conflict.
- Keep the successful parts. State the objects, framing, or colors you do not want changed.
- Compare with the original. Inspect faces, hands, product details, labels, and backgrounds at full size.
- Download the image. On the web, Google's help page describes hovering over the result and selecting Download full size.
Eligible paid Gemini subscribers can use the image's More menu and Redo with Pro. This consumes available usage and is not a guarantee that the new version will be better for every task. Compare the output instead of choosing by model name alone.
Before paying just for a redo, read the free versus Pro guide. For a different task, try the prompt notebook rather than forcing this example into every kind of image.
When a result goes wrong
| Problem | Useful next step |
|---|---|
| The model describes an image instead of making one | Check that image creation is available and selected, then explicitly request an image. |
| The edit changes the whole scene | Return to the original. Name the one object or region to change and list the details to preserve. |
| Text contains spelling errors | Supply the exact wording in quotation marks. Reduce the amount of text and inspect every character. |
| You reach a usage limit | Check Settings > Usage Limits and the reset information in your account. Limits depend on the task and plan. |
| A request is refused | Read the service's policy and use a permitted request. Do not try to bypass account or safety restrictions. |
For information about model names and where they fit, return to the Nano Banana overview.
FAQ
Common questions
Where do I find Nano Banana in Gemini?
Open the official Gemini app and look for Images in the sidebar. Some versions show Tools, then Create images. Google changes interface labels and rolls features out across accounts, so use the linked official help when your screen differs.
Do I need to write a long prompt?
No. Begin with the subject, the desired change or scene, and a few important constraints. Add detail in follow-up requests only when the result shows what is missing. Contradictory instructions usually make the task harder.
Why did the face or text change?
Image editing is generative, not a guaranteed pixel-preserving operation. Ask for one change at a time, explicitly protect identity and exact text, and compare the result with the original. Repeat or use a conventional editor when exact preservation is essential.
Can I use the guide without paying?
The guide itself is free to read. Access to Google image models depends on your account, age, location, plan, and current limits. This website cannot grant or guarantee free model access.
Sources & editorial notes
Access information and model names were checked against the official sources below. Interfaces, limits, and offers can change.
- Google Help: generate and edit images in Gemini Apps
- Google: image generation with Gemini
- Google: Nano Banana image generation
Independent editorial writing. Original prompt examples are not a model benchmark; external demonstrations carry their own attribution.
Keep exploring