(Credit: Google)
Google has launched Nano Banana 2, an upgraded version of its viral AI image generator that promises stronger subject consistency, improved instruction following, and sharper visual details.
Nano Banana 2 (aka Gemini 3.1 Flash Image) can maintain resemblance for up to five characters and 14 objects in a single workflow, Google says. This means you don’t have to repeatedly remind the AI tool about the appearance of the characters or backgrounds in your project. The model also comes with higher prompt accuracy, meaning it adheres strictly to your prompts and captures their nuances, Google says.
The other big upgrade is output quality. Images now have improved lighting, richer textures, and sharper details. You can also get them delivered in various aspect ratios and in resolutions from 512 pixels to 4K. An image shared shows support for 1:1, 16:9, 9:16, and more.
Nano Banana 2 also borrows two features from the Nano Banana Pro model: advanced real-world knowledge and improved text legibility.
With the improved knowledge, Nano Banana 2 taps real-time information and images from the web to help you generate better infographics, diagrams, and data visualizations. And with better text legibility, the text in your outputs will have fewer typos and more clarity. The new image generator also lets you translate the text in your image, so you can share it with a wider audience.
(Credit: Google)I played around with the new Nana Banana 2 model for a few minutes, and it does seem to have improved text rendering. The Pro model, however, is still leaps ahead. (Paid Gemini plans start at $8 per month and go up to $250.)
One drawback of the new model is that accessing Nano Banana Pro now takes a few additional clicks. The “Create Image” tab in the Gemini app now pulls up Nano Banana 2 by default. If you need your images generated by Pro, you need to let the new model provide its output, and then click the three-dot menu below it to select “Redo with Pro.”
All users can try out Nano Banana 2 on the Gemini app, AI Mode, Google Lens, AI Studio, Google Cloud, Flow, and Google Ads right away. The image generator doesn’t require a subscription, but paid users are likely to have wider access as always.