Google couldn’t let Samsung have all the fun and attention this week. The Gemini-maker dropped Nano Banana 2 on Thursday, the second generation of its viral AI image editing tool.
You might remember Nano Banana from last fall, when it made many fans and shocked AI users with its photorealistic capabilities. It quickly leapfrogged AI image tools from competitors like Midjourney and OpenAI, particularly with its ability to adeptly handle photo edits on top of generating entirely new images. The Pro version came out a few months later and built on those capabilities with better text generation. Now Google is promising that Nano Banana 2 is, like Hannah Montana, the best of both worlds.
Nano Banana 2 levels up Google’s base AI image model. The company said it should have the speed of the original model with the accuracy and detailed work of the Pro version. It uses the world knowledge that was built into Gemini 3, previously restricted to the Pro model. That means AI images should be more rooted in reality and less likely to be hallucinations.
Text generation in images, like when you create greeting cards, should be clearer, too. Character consistency, something that creators have said Nano Banana excels at, is also getting a boost. Google said the new base model «delivers vibrant lighting, richer textures and sharper details,» along with the ability to generate in different aspect ratios and in resolutions up to 4K.
Read more: AI Slop Is Destroying the Internet. These Are the People Fighting to Save It
Since Nano Banana’s launch, and partly because of it, we’re all now anxiously wrestling with the role AI-generated content plays in our online ecosystems. Advanced models like the ones from Google create fabricated images and videos that are nearly indistinguishable from real ones. Low-quality AI content, sometimes called slop, has flooded social media. The vast majority of social media users believe they see AI-generated posts, but less than half (44%) are confident they can spot it, a CNET survey found. Photorealistic images like the kind Nano Banana can make are a big part of that problem.
The new model is available now everywhere you use Gemini. Nano Banana 2 will replace the original, so you’ll have to choose between it and the Pro alternative. If you create AI images with the new model, Google will attach invisible watermarks called content credentials to them. So if you want to know if an image was made with AI, you can upload it to Gemini, and it may be able to identify it. That only works if it was made with Google AI, though.
