Google’s Nano Banana 2 brings faster, smarter AI image generation
Officially called Gemini 3.1 Flash Image, the new tool aims to combine high-quality visuals with substantially faster processing speed than previous versions.
Google has launched Nano Banana 2, its latest AI (artificial intelligence) image generation model, as part of its Gemini product lineup.
Officially called Gemini 3.1 Flash Image, the new tool aims to combine high-quality visuals with substantially faster processing speed than previous versions.
Nano Banana 2 is an evolution of Google’s existing image generation tools, building on the earlier models Nano Banana and Nano Banana Pro.
The original Nano Banana, formally called Gemini 2.5 Flash Image, became widely used within the Gemini ecosystem. Nano Banana Pro, or Gemini 3 Pro Image, introduced more accurate text rendering and enhanced world knowledge last year.
The key improvement in Nano Banana 2 is its balance of speed and fidelity. It generates and edits images quickly while producing high visual quality with clear details and rich textures, says the company.
Nano Banana 2 draws from the Gemini model’s real-world knowledge base as well as real-time information and images from web search to generate more accurate renderings of specific subjects, according to Google’s model overview.
Real-world knowledge refers to the use of live information from web searches to make images more contextually accurate.
One important upgrade is text rendering, which means the model can now place legible and accurate words directly into images. This has practical use for marketing mock-ups, posters,, and diagrams.
In addition, Nano Banana 2 supports in-image localisation, where text can be translated or adapted across languages within a generated picture without losing clarity.
Google brings Intrinsic under its fold to scale robotics and physical AI
Google has also highlighted subject consistency, a feature that allows the model to maintain the identity of characters or objects across multiple edits. For example, if a user generates a series of images featuring the same person, the model will strive to keep their appearance consistent from image to image.
Nano Banana 2 is being integrated across a wide range of Google products. It is now the default image model in the Gemini app for creating and editing pictures. It also powers AI Mode in Google Search and image features in Google Lens.
In addition, the model works within Flow by Google, a creative tool for stitching visuals together, and is available for developers through Google AI Studio and Gemini API. These developer tools let programmers use the model in their own apps and services, although a paid API key is required for access.
Alongside the performance upgrades, Google continues to focus on content provenance. Images generated by the model include an invisible digital watermark using its SynthID system. This helps signal to people and systems that the picture was produced by AI, and can be checked later using verification tools.
Edited by Swetha Kannan


