Get ready to check the corners of images you see on social media for the Gemini watermark so you can figure out if it’s fake or not. On Thursday, Google announced that it is rolling out Nano Banana 2, the latest version of its image generation model, which the company claims comes imbued with “advanced world knowledge” and “precision text rendering and translation.”
Google described Nano Banana 2 as a “best of both worlds” mashup of the first version of its image generation tool and Nano Banana Pro, the advanced version of the model made available last November. It brings the “advanced intelligence” and “studio-quality creative control” of the Pro model to Nano Banana 2, which now runs on the company’s Gemini 3.1-Flash model, so the process of image generation should be faster than ever.
One of the primary calling cards of Nano Banana Pro was its ability to handle text—a task that most image generators have struggled with. That skill has been further honed for Nano Banana 2, which Google claims can now pull from the real-world knowledge base of Gemini, the company’s flagship LLM. The model will apparently be able to pull in real-time information and images directly from web search to “more accurately render specific subjects.” Google said the Nano Banana 2 should be ideal for infographics and data visualizations. (For the record, the company said the same about Nano Banana Pro.)
Google’s other point of focus for Nano Banana 2 is the new and improved creative control, including the ability to generate as many as five characters and up to 14 objects in a single workflow—though, again, that was kinda already part of the sell with Nano Banana Pro. This version of the model is supposed to be better at adhering to user prompts and is better at generating lighting effects and textures.
A lot of these features are rehashed from the company’s last model, though surely there are improvements to the overall outputs from Nano Banana 2. That’s because the real selling point with this model seems to be speed. Since it’s using the company’s Flash model, image generation takes less time. The company also says that the model can handle outputs that range in resolution from 512px to 4K.
According to Google, Nano Banana 2 will be available starting today and will basically push the old models out of their current spots across the company’s suite of apps. Nano Banana 2 will replace Nanon Banana Pro through the Gemini app (though Google AI Pro and Ultra subscribers can revert to the past model if they want), and v2 of the model will become the default in Google Search and Flow, the company’s AI creative studio app. It’ll also pop up in AI Studio + API for developers, Google Cloud, and Google Ads in the coming days.
Read the full article here
