Nano Banana Pro image generator and editor

What Nano Banana Pro is

Nano Banana Pro is Google's image generation and editing model, published on 20 November 2025 and named Gemini 3 Pro Image in Google's own material. Google built it on Gemini 3 Pro, which is why it reasons about a brief rather than only matching its words.

On this site it runs in the same image workspace as every other model. This page supplies the model, the controls this workspace exposes, the prompts Google published alongside its own samples, and a live quote before a task is submitted.

The distinction that matters when you choose a model is the one Google draws itself. Nano Banana Pro is positioned for high-fidelity work that needs maximum factual accuracy, legible text and fine control. The original Nano Banana, which is Gemini 2.5 Flash Image, remains the fast option for quick, fun edits.

Seven jobs Nano Banana Pro is suited to

The list follows what Google demonstrates with this model rather than what any reseller claims about it.

• Posters and packaging that have to carry readable text. Google calls this the model's headline strength and shows it working on flat headlines, long paragraphs, hand-drawn lettering and a phrase built out of physical objects.
• Infographics and diagrams assembled from real information. Google's own samples include a care guide for a houseplant and a step-by-step recipe, both built from the model's world knowledge rather than from an uploaded layout.
• Anything that depends on current facts. The model can reach Google Search, and Google's launch post shows a pop-art weather graphic generated from live weather data pulled through that route.
• Storyboards and image series, generated several frames at a time so a sequence keeps the same cast, style and palette.
• Compositions that blend many inputs: Google publishes a working limit of fourteen images and five people held consistent in a single workflow, which is what makes a group shot or a product line possible from reference material.
• Studio-style retouching decisions: changing the lighting direction, turning day into night, moving the focal point into the foreground, or adapting one asset to several aspect ratios without rebuilding the scene.
• Multilingual campaigns. The model can render text in several languages and translate the lettering already inside an image, so an existing design can be republished for another market.

Prompts you can copy

The prompts below are the ones Google publishes with the samples shown in the gallery above, reproduced verbatim. Treat them as starting points: keep the structure and replace the subject. Opening one loads it into the workspace.

Create an infographic about this plant focusing on interesting information.

Create an infographic that shows how to make elaichi chai

Create a storyboard for this scene

View of a cozy street in Berlin on a bright sunny day, stark shadows. the old houses are oddly shaped like letters that spell out "BERLIN" Colored in Blue, Red, White and black. The houses still look like houses and the resemblance to letters is subtle.

Specifications Google publishes

These figures come from Google's own model page and launch material for this model, not from a reseller's summary. Google repeats the caveat that input counts vary by surface, so read the workspace as the authority on what you can send here.

What this workspace exposes

The picker is the authority on what you can submit here. Another provider's page describes that provider's account, not this one, and price is resolved by the service rather than by this page.

Competing sites quote their own credit schedules for this model, and those numbers describe their accounts rather than this one. Your figure is the one the workspace prints before you submit.

Nano Banana Pro and the rest of the family compared

All three tiers are listed on this site. The positioning row is Google's own; the control rows come from this workspace's picker, which is why they can differ from what another site offers.

Because the three tiers share the same resolution ladder and the same reference-image limit in this workspace, the difference you actually feel is reasoning depth, text accuracy and speed rather than a ceiling on output size. Run the same brief through two of them when the choice matters.

Studio controls and the edit workflow

An edit starts with images you are allowed to use and an instruction that says what should change and what must survive it. There is no mask to draw: the model reasons about which parts the instruction refers to and leaves the rest alone.

Google groups the finer controls under studio-style editing and demonstrates four of them directly: adjusting the camera angle, changing the focus point, applying colour grading, and transforming the lighting of a scene. Two of those samples are on this page, one turning a snow scene from day into night and one moving the focal point onto flowers in the foreground.

Write the instruction in two halves. First the object and the change. Then the parts that must not move, such as the silhouette, the camera angle or the colour of everything else.

• Check faces and hands first. Resemblance drift is the most common unwanted change in a portrait edit.
• Re-read small text. Google lists small text and spelling as its own known weakness, so lettering and prices are the likeliest casualty.
• Compare reflections, shadows and contact points where an object was added or removed.
• Re-run with one instruction changed at a time, so you can attribute the difference to the instruction and not to the model's variance.

How to generate or edit here

Four steps, in the order the workspace asks for them.

1. Paste your prompt, or open a prompt from the gallery above with the copy button.
2. Choose a resolution tier and an aspect ratio. Add reference images when the job is an edit rather than a new image.
3. Read the quote, then sign in when the workspace asks for it.
4. Submit, review the result, and download the version you want to keep. If a task fails, read its status before retrying so you do not submit a duplicate.

Building a prompt that survives revision

Google publishes a formula rather than a keyword list. Name the deliverable, then the subject, the action and the scene, and build from there.

Google's own guidance splits the work in two. The first pass establishes the vision: subject, composition, action, location and style, plus a direct instruction when the job is an edit. The second pass refines it: aspect ratio, camera and lighting details, the exact wording and placement of any text, and a note on factual accuracy when the output is a diagram.

When several images are involved, Google recommends naming the role of each one, for example using one for the character's pose, another for the art style and a third for the background. Be specific rather than short. When you revise, change one decision at a time, or you cannot tell which change caused the difference.

What Google says the limits are

A model page should say where its own claims stop. Google publishes a limitation list for this model, and these are its five headings, kept in its own terms.

• Visual and text fidelity: not every image is perfect, and the model can still struggle with small faces, accurate spelling and fine details.
• Data and factual accuracy: world knowledge is extensive but not infallible, so infographics, annotated diagrams and data-driven visuals should be verified before they are published.
• Translation and localisation: text can be generated and translated in many languages, but grammar, spelling, cultural nuance and idiom can still slip.
• Complex edits and image blending: masked editing, major lighting changes and multi-image blends can produce unnatural results or visible artefacts.
• Character features: consistency is a strength but not a guarantee, and Google says it is still working on it.

Where the samples and specifications come from

Every figure and every image on this page traces back to one of the sources below. Where a value comes from a competing site rather than from Google, the row says so.