Fast and precise. Google introduces the Nano Banana 2 image generator, here's everything it brings

Image generation using artificial intelligence has become a common part of many applications over the past year, but quality results so far have required either a moment of patience or a premium subscription. Google is now trying to eliminate these compromises. The company yesterday announced the Nano Banana 2 model, officially designated Gemini 3.1 Flash Image, which is intended to bring the best of both worlds – the quality of the more expensive Pro version at the speed of the Flash model.

Text in images finally works

One of the biggest problems with generative models so far has been text rendering. Inscriptions on signs, captions in infographics, or text on greetings often looked like a random jumble of letters. Nano Banana 2 addresses this shortcoming – it can generate readable and accurate text, which opens the door to practical use for creating marketing materials, invitations, or visual notes.

Furthermore, the model handles text translation and localization directly within the image. So, if you need to convert a sign in English to Czech, you don’t have to regenerate the image from scratch. Just provide the instruction, and the model will replace the text.

Characters maintain their appearance

Another new feature is so-called subject consistency, which is the ability to maintain a consistent appearance of characters and objects across multiple images. Nano Banana 2 can handle up to five characters and fourteen objects within a single workflow. This will be especially appreciated by those who want to create comics, storyboards, or a series of illustrations with recurring characters.

Until now, with every new image, it was necessary to hope that the model would render the character similarly to before. Now, the result should match the original prompt much more reliably.

From notes to infographics

Nano Banana 2 draws from the Gemini knowledge base and can also utilize current information from web search. In practice, this means the model better understands the real world – it knows what specific places, products, or personalities look like and can project this knowledge into generated images.

Fast and precise. Google introduces the Nano Banana 2 image generator, here's everything it brings Adam Kurfürst News Adam Kurfürst News

Google highlights the ability to convert notes into diagrams or create data visualizations. Instead of tediously assembling an infographic in a graphic editor, you just need to describe what the image should contain, and the model will take care of the rest.

Resolution up to 4K

Regarding technical parameters, Nano Banana 2 supports resolutions from 512 pixels up to 4K and offers various aspect ratios. You can therefore generate both vertical images for social media and widescreen shots for presentations or web banners.

Google also mentions improved visual quality – richer colors, more realistic textures, and sharper details. Compared to the original Nano Banana model, the results should appear more natural.

Where to find Nano Banana 2

The new model is gradually expanding into various Google products. In the Gemini app, it replaces the existing Nano Banana Pro for Fast, Thinking, and Pro models. Google AI Pro and Ultra subscribers can still invoke the Pro version via the menu on the generated image.

Nano Banana 2 is also available in Google Search – specifically in AI mode and Google Lens. It has gained support in 141 new countries and eight additional languages. The model can also be found in the Flow service, where it functions as the default image generator without consuming credits. For developers, it is available in AI Studio and via the Gemini API.

Google also emphasizes that all generated images contain a SynthID watermark and C2PA metadata, which allow verification that the image was created by artificial intelligence.

Have you tried generating images in Gemini yet?

Source: Google