Gemini Image Generation (Nano Banana) Explained

Gemini image generation is Google's ability to create and edit images from text and photos. Google calls its Gemini image models "Nano Banana". As of October 2026, the current API lineup is Nano Banana 2.1, dated 6 October 2026, alongside Nano Banana 2, Nano Banana 2 Lite and Nano Banana Pro. Nano Banana 2.1 costs $0.0336 per image at 1K.

What Gemini image generation is

Nano Banana is Google's nickname for its Gemini image models. The first one was announced on 26 August 2025 as Gemini 2.5 Flash Image, in preview on the Gemini API, Google AI Studio and Vertex AI. The name stuck, and later models carry it too. Google's API docs now describe the original as legacy.

You can use the models in several ways: type a prompt to generate an image, upload a photo and describe a change, or combine several images. The Gemini API docs describe them as able to add, remove or modify elements, change style and adjust colour grading.

Nano Banana models compared

| Model | API model ID | Google's positioning | Price per image at 1K (standard) | |---|---|---|---| | Nano Banana 2.1 | gemini-nano-banana-2.1 | "Primary high-efficiency workhorse" | $0.0336 | | Nano Banana 2 Lite | gemini-3.1-flash-lite-image | "Fastest and cheapest"; 1K only | $0.0336 | | Nano Banana 2 | gemini-3.1-flash-image | Previous generation; 2.1 recommended for new projects | $0.067 | | Nano Banana Pro | gemini-3-pro-image | "Premium choice for the most complex visual tasks" | $0.134 | | Nano Banana (original) | gemini-2.5-flash-image | Legacy | $0.039 |

Sources: the Gemini API image generation docs and the pricing page, both read on 17 October 2026.

A short timeline: from Gemini 2.5 Flash Image to Nano Banana 2.1

26 August 2025: Nano Banana, the Gemini 2.5 Flash Image model, is announced. Google's developer post gives its launch price as $0.039 per image.
20 November 2025: Nano Banana Pro (Gemini 3 Pro Image) launches. It adds infographics and diagrams, area edits, camera-angle and focus changes, legible text in several languages, and 2K and 4K output. It can blend up to 14 input images and keep up to five people consistent.
26 February 2026: Nano Banana 2 (Gemini 3.1 Flash Image) is announced as a preview, with Pro-style capabilities at Flash speed, web-search grounding and output from 512 px to 4K.
6 October 2026: Nano Banana 2.1 is announced, built on Gemini 3.6 Flash. Its model card lists the Gemini app, Google AI Studio, the Gemini API, Google Search AI Mode, Google Ads, Google Flow and Google Stitch among its places of use.

What it can do

Editing: add, remove or modify elements, change style, and adjust colour grading. Google recommends multi-turn editing through the conversation as the way to iterate.
Multiple references: up to 14 reference images in one request. Nano Banana 2.1 and 3.1 Flash can keep up to 10 objects and 4 characters consistent.
Text: the models offer "advanced text rendering" for infographics, menus, diagrams and marketing assets. Google's model card lists limits on small and long text, covered below.
Reasoning: Gemini 3 image models reason before they render, and that thinking cannot be turned off in the API. For Nano Banana 2.1 the levels are minimal, medium (the default) and high.
Search grounding: supported on Gemini 3 and newer models, except Lite. Nano Banana 2.1 and 3.1 Flash also support Image Search grounding. Google's pricing page lists 5,000 free search requests per month, shared across Gemini 3 and newer models, then $14 per 1,000 requests.
Video input: Nano Banana 2.1, 3.1 Flash and 3.1 Flash Lite can generate images from YouTube URLs or uploaded video files.

Resolutions, aspect ratios and reference images

Output comes in 1K, 2K and 4K. Nano Banana 2.1 defaults to 1K, Nano Banana 2 also supports 512 px (0.5K), and Lite supports 1K only. The size value uses an uppercase K, so write "2K" not "2k".

The docs list these aspect ratios: 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9 and 21:9. The Nano Banana 2.1 model page also names 1:4, 4:1, 1:8 and 8:1. The full list differs by model, so check the model page for the one you use. Up to 14 reference images can go in a single request.

How much does it cost?

Standard prices per image from the pricing page:

| Model | 1K | 2K | 4K | |---|---|---|---| | Nano Banana 2.1 | $0.0336 | $0.0504 | $0.113 | | Nano Banana 2 Lite | $0.0336 | Not offered | Not offered | | Nano Banana 2 | $0.067 (0.5K: $0.045) | $0.101 | $0.151 | | Nano Banana Pro | $0.134 | $0.134 | $0.24 |

Batch prices for Nano Banana 2.1 are half the standard rate, so the 1K batch price is $0.0168. Image models show "Not available" on the API free tier. Google AI Studio usage is free in available regions.

For the consumer app, Google's subscriptions page lists these plans:

Free: $0, and includes image creation and editing in the Gemini app
AI Plus: $4.99 per month
AI Pro: $19.99 per month
AI Ultra: from $99.99 per month, with a $199.99 tier also listed

The page gives no numeric image quota. It says usage is compute-based, with limits that reset every five hours up to a weekly cap.

How to use it

In the Gemini app: describe the image you want, or upload a photo and describe the change. Keep editing in the same conversation, since that is the recommended way to iterate.
In Google AI Studio: try the models without writing code. Usage is free in available regions.
In the Gemini API: call the model ID from the table above, and set the resolution with an uppercase K, such as 2K. The batch rate is half the standard rate for Nano Banana 2.1.
In Google products: the Nano Banana 2.1 model card lists Google Flow, Google Search AI Mode, Google Ads and Google Stitch as places the model is available.

The Gemini Help page we read still describes Nano Banana 2, Lite and Pro, not 2.1, so check the current help page for the features in your plan.

Watermarks and provenance

Google's API docs say all generated images include a SynthID watermark. Google also pairs SynthID with C2PA Content Credentials, according to its 26 February 2026 Nano Banana 2 post. The Nano Banana Pro launch post, from 20 November 2025, says you can upload an image to the Gemini app to check whether Google AI generated it.

At the Nano Banana Pro launch, images from free and Google AI Pro users also carried a visible Gemini sparkle watermark, which Ultra subscribers and AI Studio did not get. Check the current Gemini Help page before you rely on that detail. The Nano Banana 2.1 model card does not mention SynthID or C2PA, so the API docs are the source for the watermark statement.

Limits to know

Google's model card for Nano Banana 2.1 lists these limits:

Weak rendering of small text, long paragraphs and multi-page text
Imperfect character consistency
Partial instruction following in masked or doodle edits
Occasional left and right confusion
Hallucinations, and occasional slowness or timeouts

Google's own evaluations rank Nano Banana 2.1 above Nano Banana Pro. Its model card reports an overall text-to-image preference score of 1050 for 2.1 (Thinking), against 935 for Pro, and an infographic factuality score of 0.521, against 0.265. The independent blind-vote leaderboard also ranks 2.1 above Pro, at 1174 against 1115. Individual prompts vary, so compare the models on your own work. For a wider view of the market, see our guide to the best AI image generators in 2026.

Google's docs also list the imagen-4.0 models for shutdown on 17 August 2026, and name Nano Banana 2.1 as the replacement. Google's pages disagree on the shutdown date for gemini-2.5-flash-image, so check the deprecations page before you build on it.

Nano Banana in OndaLabs

OndaLabs is an AI voice studio, and its image studio offers three Google models under its own labels, alongside two GPT Image models. The Google models are Nano Banana Lite, Nano Banana and Nano Banana Pro. Their credit costs in the studio are:

Nano Banana Lite: 2,000 credits at 1K
Nano Banana: 2,000 credits at 1K, 3,000 at 2K and 6,000 at 4K
Nano Banana Pro: 7,500 credits at 1K or 2K, and 13,000 at 4K

The studio shows the credit cost before you generate, and you can generate or edit with reference images. Studio labels are OndaLabs' own. They do not name a Google API version, and OndaLabs is not a Google product or a Gemini API reseller. Image generation is included from the Starter plan, at $9 per month, and commercial use is included on Starter and above. Compare plans on the pricing page.

For the wider family of Google audio tools, see Gemini text to speech. For prompts that work across models, see how to write AI image prompts that work.

Frequently asked questions

What is Nano Banana?

Nano Banana is Google's nickname for its Gemini image models. The current API lineup is Nano Banana 2.1, Nano Banana 2, Nano Banana 2 Lite and Nano Banana Pro, and the original Nano Banana is now legacy.

What is the difference between Nano Banana, Nano Banana 2, 2.1 and Pro?

Nano Banana 2.1 is Google's primary high-efficiency model as of October 2026, released on 6 October 2026. Nano Banana 2 is the previous generation, and Google recommends 2.1 for new projects. Nano Banana Pro is the premium option for the most complex visual tasks, and it costs more per image at 1K.

Is Gemini image generation free?

The Gemini app's free plan includes image creation and editing, with no numeric quota on Google's page. Google AI Studio usage is free in available regions. The image models show "Not available" on the API free tier.

How much does the Gemini image API cost per image?

At standard rates, Nano Banana 2.1 costs $0.0336 at 1K, $0.0504 at 2K and $0.113 at 4K, and batch prices are half the standard rate. Nano Banana Pro costs $0.134 at 1K and 2K and $0.24 at 4K.

Does Gemini watermark its images?

Yes. Google's API docs say all generated images include a SynthID watermark. Google also pairs SynthID with C2PA Content Credentials, and you can upload an image to the Gemini app to check whether Google AI generated it.