Google launches Nano Banana Pro for AI image creation

Google is starting to offer Nano Banana Pro, its new image generation and editing model based on Gemini 3 Pro. It is arriving as a paid preview for developers through the Gemini API, Google AI Studio and Vertex AI.

The model is designed to create images with more detail, more accurate text and controls suited to professional production. It is not limited to generating an image from a description: it can also modify specific elements, combine references and maintain a coherent composition.

What Nano Banana Pro can do

Nano Banana Pro is the name used for Gemini 3 Pro Image. Google presents it as an evolution of Nano Banana, which is based on Gemini 2.5 Flash Image and launched a few months earlier.

The new version lets you control aspects such as:

• Lighting, camera, focus and color grading.
• Composition and the position of elements.
• 2K and 4K resolutions.
• The integration of products, logos and other reference images.

It can also preserve the appearance of up to five people, combine up to six high-fidelity images or mix up to fourteen standard references in a single ad. For example, an application could receive a product photo, its logo and several visual references to create an ad ready for review.

Text is no longer the weak point

One of the most important changes is the generation of text inside images. The model can create logos, posters, menus, documents, comic panels and educational materials with clearer lettering that more closely follows the instruction.

This enables specific uses such as generating a multi-page comic in the chosen language, creating an infographic about bicycle maintenance or designing an advertising campaign with localized text.

Translation can also be applied to an existing image. If you change a menu from English to French, for example, the model tries to preserve the original design, visual style and layout instead of creating an entirely new image.

Images connected to current information

Nano Banana Pro incorporates Gemini 3 Pro's general knowledge and can use Google Search to look up information on the web when the grounding feature is enabled. Grounding means connecting the model's response to external sources to produce results based on more recent data.

This can be useful for creating biological diagrams, historical maps or infographics about a specific topic. Even so, the model's ability to look up information does not remove the need to review the result: a well-designed image can still contain incorrect data.

It is also coming to development tools

Google is expanding access beyond its own platforms. In Google Antigravity, its development environment with AI agents, agents can use the model to create visual prototypes of interfaces before writing the code.

Adobe and Figma are also integrating Gemini 3 Pro Image into their creative platforms. For developers, the company offers demo applications, documentation, prompting guides and examples for using the model from Google AI Studio or Vertex AI.

Each generated or edited image includes a SynthID digital watermark. Its purpose is to indicate that the content was created or modified with AI, although it does not replace a visible label and does not by itself guarantee that a person can detect the image's origin.

For you, the change will be most noticeable in tasks that previously required several tools: creating an ad with specific references, translating text inside an image, preparing an infographic or generating a visual prototype. Nano Banana Pro offers more quality and control than Nano Banana, but it also means higher cost and latency. For fast, inexpensive work, Google is keeping Gemini 2.5 Flash Image; for more detailed results, it is turning to this new model.