Write a scene, then specify its purpose
Describe the subject, action, setting, light and composition in sentences. Say whether you need a product listing, poster or website hero. For a hero image, leave space for live website copy.
From a brief to an image
How to use Nano Banana 2.1
Start with a clear scene, choose the right format, and refine one detail at a time. Follow the steps below or open a ready-to-use brief from the prompt library.
Updated 2026-10-09 · Independent guide
Explore 16 image examples ↗The latest model
What changed in 2.1?
Google released Nano Banana 2.1 on October 6, 2026. Its official API model name is gemini-nano-banana-2.1. Google describes improvements in visual design, image quality, text and subject consistency. It supports image generation and editing at 1K, 2K and 4K through the API. Official model reference ↗
The model still makes mistakes in wording, spatial placement and likeness. Review the image before using it. Model card and limitations ↗
Create an image on this site
01
Choose a look
Browse the images in the prompt library. Pick a result you like, then choose Use prompt to load its full prompt into the generator. Replace the subject and quoted wording with your own.
02
Set the frame
Choose one of this site’s five aspect ratios and 1K, 2K or 4K. For editing prompts, upload the required references. Check the selected size and its credit cost.
03
Generate and review
Sign in, check the service status and your balance, then submit. Inspect the wording, hands, product details and composition. Download the result or upload it for another edit.
A prompt link only fills a draft. It never submits a task or spends credits. Reference uploads require sign-in and accept PNG, JPEG or WebP, up to 5 MB per image and 10 MB in total. Generation depends on the live service status shown in the generator.
Six useful prompting habits
Quote the exact words
Settle the copy first. Put each short headline or label in quotation marks, specify its position and lettering style, and proofread the result. Choose 2K or 4K for finer detail after reviewing the credit cost.
Edit one thing and name what stays
Upload a clear reference, describe the change, and list the face, product shape, label, lighting or background that must remain. Compare the output with the reference after every edit.
Give each reference a role
For a composite, say which product or logo comes from each numbered reference. Upload the images in the same order. This site accepts up to four references per edit; start with two when possible.
Use a marked image for local instructions
Draw an outline on the source photo in an image editor, upload it here, then request a change inside the outline and removal of the marks. The model can miss a boundary, so inspect surrounding areas.
Refine the last result in small steps
Download a result you like, upload it in Edit an image, and ask for one revision. This site sends a fresh request each time; it does not carry a Gemini chat history. Repeat identity and preservation details in the new prompt.
Other ways to use the model
In the Gemini app
Open Tools, choose Create images, and describe your scene or attach an image to edit. Google currently lists Nano Banana 2.1 for its AI Plus, Pro and Ultra plans; availability and limits depend on your account. Gemini image guide ↗
In Google AI Studio or the API
Select Nano Banana 2.1 and use the platform’s ratio and size controls. The API offers more aspect ratios and reference inputs than this site exposes. Search grounding and conversational history are platform features; this site currently uses a prompt and up to four reference images per request. Official generation documentation ↗