Nano Banana 2.1 Generator

Make pictures from a prompt or edit up to 10 of your photos with Google’s newest Gemini image model: readable infographics, consistent characters, panoramas from 8:1 to 1:8 and up to 4K. Online, with no Google AI plan, from 1 credit an image.

Make a Nano Banana AI image online

Model

Prompt

0/20,000

Format

Aspect ratio

Size

2K · Auto · No watermark

A man in a dark jacket standing in a meadow of wildflowers at golden hour, seen from behind, made with Google’s Nano Banana modelExample

What is Nano Banana 2.1?

Nano Banana is Google’s name for its Gemini image models, and version 2.1, released on October 6, 2026, is the newest. It is built on Gemini 3.6 Flash and updates Nano Banana 2 while keeping Flash-level speed and cost. One model makes pictures from a prompt and edits your photos. On Melies AI it runs in your browser, with up to 10 images per edit and no Google AI plan, app or API key.

Google lists five updates in version 2.1. Pictures look better and more realistic at 1K, 2K and 4K. Wide and tall panoramas (1:4, 4:1, 1:8 and 8:1) no longer show tiling artifacts at 2K and 4K. Text and infographic layouts are more accurate. And one picture can fuse up to 14 references in Google’s own API, keeping up to 4 characters and 10 objects consistent. Google’s API also offers Search grounding and thinking levels; this site does not expose those settings.

Google’s model card scores version 2.1 in human side-by-side tests. With thinking on, it reached an Elo of 1050 for overall text-to-image preference, against 990 for Nano Banana 2 and 935 for Nano Banana Pro. For keeping several characters consistent in an edit it scored 1106, against 978 and 1011. These are Google’s own tests, not independent ones.

On this site you pick Auto or one of 14 shapes, and 1K, 2K or 4K. A square is 1024 × 1024, 2048 × 2048 or 4096 × 4096 pixels, and an 8:1 panorama at 4K is 12288 × 1536. Prompts run up to 20,000 characters. Save as PNG or JPEG. Google embeds an invisible SynthID watermark in every picture; no visible watermark is added.

Aspect ratios + Auto
14
Largest size
4K
Reference images here
10
Pixels · longest side
12,288
Prompt characters
20,000
Credit · one 1K image
1

By Google · run through a third-party API · not affiliated with Google · Last updated

What Google’s newest image model does well

Five things the 2.1 release improves, from infographics to panoramas, and where it still has limits.

A dog food ad for PUGGY with the headline Good Food Good Mood Every Day, a Pug Fuel bag, benefit icons and a pug, every line of text readable
01

Infographics and posters that read right

Google improved text rendering and infographic layout in version 2.1: titles, labels, charts and step-by-step diagrams land where you ask. Put every word in quotes. Google notes that small text can blur at 1K, so render text-heavy pictures at 2K or 4K.

The same red-haired woman in four scenes, a café, a rooftop at sunset, a forest trail and an office, with the same face in each
02

The same characters in every picture

Upload up to four people or characters and put them in new scenes, outfits and poses while their faces stay the same. Google scored version 2.1 well ahead of earlier Nano Banana models on multi-character consistency. Name each upload @Image1, @Image2 in the prompt.

Six product photos, sneakers, a puffer vest, sunglasses, a silver bag, cargo jeans and a grey hoodie, worn together by one woman crossing a rainy Shibuya street at night
03

Fuse up to 10 images into one

Combine products, props, a location and a style into one picture: a sofa, rug and lamp in one room, or a product with its packaging and a backdrop. Say which image each part comes from and what must stay exactly as shown.

A huge ocean wave towering over green hills while a person walks along a path below
04

Panoramas from 8:1 to 1:8

Four extreme shapes, 8:1, 4:1, 1:4 and 1:8, sit next to the usual ten. Google fixed the tiling artifacts these shapes showed at 2K and 4K, so a website banner, a scroll or a tall infographic comes out in one piece, up to 12288 × 1536 pixels.

A sculpted silver ring on charred wood in a beam of warm light, with fine metal and bark texture
05

4K detail at Flash speed

Version 2.1 keeps the speed of Google’s Flash models while improving quality at every size. Draft at 1K for 1 credit, then render the keeper at 4K, about 17 megapixels, for 3 credits.

Google’s Nano Banana models compared

Google runs three main Nano Banana models. Version 2.1 and Pro are both on this site. The figures come from Google’s docs and model card.

Version 2.1Nano Banana 2Nano Banana Pro
Google modelgemini-nano-banana-2.1Gemini 3.1 Flash ImageGemini 3 Pro Image
Google’s role for itFast, efficient workhorsePrevious fast workhorsePremium, reasons through complex tasks
Sizes1K, 2K, 4K512 px, 1K, 2K, 4K1K, 2K, 4K
Aspect ratios14, from 8:1 to 1:814, from 8:1 to 1:810, from 21:9 to 9:16
Overall preference (Elo)1050 (thinking on)990 (thinking on)935
Multi-character consistency (Elo)1106 (thinking on)978 (thinking on)1011
On this siteYesNoYes

How to use the Nano Banana AI image generator

  1. 01

    Pick a tab

    Text to image to start from words, or Image edit to start from up to 10 of your own pictures.

  2. 02

    Write the prompt

    Describe the subject, layout, light and style, and put any text in quotes. In an edit, name your images @Image1, @Image2, say what to change and what to keep.

  3. 03

    Choose shape and size

    Auto or one of 14 shapes from 8:1 to 1:8, and 1K, 2K or 4K. The price is on the button: 1, 2 or 3 credits.

  4. 04

    Generate and download

    Press Generate and save as PNG or JPEG. The picture stays under “My images” for 14 days.

Nano Banana pricing per picture

One price per picture, set by its size: 1 credit at 1K, 2 at 2K and 3 at 4K. Edits cost the same however many images you upload. Packs from $9.90, no subscription, failed images refunded.

Model1K2K4K
Nano Banana1 credit2 credits3 credits

Nano Banana FAQ

Something else? Email support@meliesai.com.

In Google’s Gemini app, the free plan uses an earlier Nano Banana, and Google keeps version 2.1 for Google AI Pro, Plus and Ultra plans. On Melies AI, new accounts that sign up with Google get 4 free credits, enough for 4 Nano Banana pictures at 1K. After that you pay per picture with one-time packs from $9.90.

A Nano Banana picture costs 1 credit at 1K, 2 credits at 2K and 3 credits at 4K. Edits cost the same, whether you upload 1 image or 10. The $9.90 pack has 100 credits, enough for 50 pictures at 2K. There is no subscription, and failed pictures are refunded.

Google released the 2.1 version of Nano Banana on October 6, 2026, as a generally available model in the Gemini API and Google Cloud. The same day Google published its model card. It reached the Gemini app, Google AI Studio, Google Flow, Search AI Mode, Google Ads and Stitch.

In Google’s own human tests, yes, for most jobs. Version 2.1 with thinking on scored higher than Nano Banana Pro on overall preference, infographic design, general editing and multi-character consistency. Google still calls Pro the premium choice for complex instructions and brand consistency. Both are in the image generator here, so you can run one prompt on each.

Google lists better quality and realism at 1K, 2K and 4K, no more tiling artifacts on 8:1, 4:1, 1:4 and 1:8 panoramas at 2K and 4K, more accurate text and infographic layout, and stronger multi-image fusion with up to 4 consistent characters. Nano Banana 2 also offers a 512-pixel size, which version 2.1 does not.

No independent benchmark covers both, and each maker reports its own tests. Pick Google’s model for panoramas, consistent characters and fast 4K. Pick GPT Image 2.6 for quality levels, transparent backgrounds, GIF or SVG output, or a painted mask over the exact area to change. Both are in the same image generator here, so you can run one prompt on each.

Nano Banana is Google’s name for the image models in Gemini. The family has the original Nano Banana (Gemini 2.5 Flash Image), Nano Banana 2 (Gemini 3.1 Flash Image), Nano Banana 2 Lite, Nano Banana Pro (Gemini 3 Pro Image) and, newest, version 2.1. Each one makes pictures from text and edits images in conversation.

Open the Image edit tab, add 1 to 10 photos and describe the change. Name each photo @Image1, @Image2 in the order you added them, and say which part comes from which photo, such as the person from @Image1 and the jacket from @Image2. Pick Auto or a shape and a size, then press Generate.

Up to 10 images per edit on this site, as JPG, PNG or WebP. Google’s own API takes up to 14 and keeps up to 4 characters and 10 objects consistent. Images are numbered in the order you add them and called @Image1, @Image2 and so on in the prompt. The price stays the same however many you add.

Three sizes: 1K, 2K and 4K, where a square is 1024, 2048 or 4096 pixels a side and every 4K shape is about 17 megapixels. 14 shapes plus Auto: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 5:4, 4:5, 21:9 and the panoramas 4:1, 1:4, 8:1 and 1:8. At 4K an 8:1 picture is 12288 × 1536.

No visible watermark is added to Nano Banana pictures made on Melies AI. Google embeds an invisible SynthID watermark in every picture its image models make, which tools such as the Gemini app can detect. In the Gemini app itself, Google also adds a visible mark. Each picture stays under “My images” for 14 days.

No, Nano Banana makes still pictures. To animate one, use a video model on this site, such as Gemini Omni (Google’s own video model), Seedance or Kling. Make the picture here, then upload it as the first frame in the video generator’s Image to video tab.

Yes. Google offers it in the Gemini API and Google AI Studio as gemini-nano-banana-2.1, and in Google Cloud. That route needs your own Google account, billing and code. Melies AI has no API of its own: it is a website where you make pictures in the browser and pay with credits.

Google’s model card lists them. Small text is often blurry at 1K, and long paragraphs do not render well. Characters are not always perfectly consistent. Edits occasionally keep the original pose. Left and right can get confused. For dense small text, render at 2K or 4K, or try Seedream 5.5, which ByteDance tuned for small text.

Yes. Pictures you generate are yours to use, including in ads, client work and products, as long as your prompt and uploaded images do not infringe anyone else’s rights and follow our content standards. Edits of real people need their consent.

No. Melies AI is an independent platform that gives you access to Google’s Nano Banana model through a third-party API provider. It is not affiliated with, endorsed by or sponsored by Google. Google’s own places to use it are the Gemini app, Google AI Studio and Google Flow.

Make your first Nano Banana picture

Write a prompt or upload your photos, pick a shape from 8:1 to 1:8 and a size up to 4K, and download the result.

Create an image