Nano Banana 2 Will Appear on Google Earth.
The latest updates to Google Gemini, codenamed Nano Banana 2, have significantly improved the quality and accuracy of AI-powered imagery, and similar capabilities are now being rolled out to the Google Earth web app. This means you can use aerial, satellite, and 3D imagery stored in Google Earth as the basis for AI-powered imagery—for example, for historical reconstructions of the distant past or for sci-fi reimaginings of the future.
We’ve already seen how accurate and realistic Google Gemini features have become, powered by Nano Banana, and Google offers five ways to leverage these capabilities in Google Earth. First, you can learn as you explore, for example, by instructing the AI to create an infographic about the Statue of Liberty, highlighting key historical facts.
A text prompt, such as “imagine this empty lot in Tokyo as a vibrant shopping district with open spaces,” will transform a rather empty section of a Google Earth 3D map into something completely different. You can also visualize personal projects, such as a dream lakeside home or an addition to your current property. Google Earth will integrate the existing landscape with what you’ve specified, helping you brainstorm ideas during the planning stage.
You can also peer into the future and the past. “Sometimes it’s fun to let your imagination run wild,” Google says, suggesting asking AI to imagine what a particular place might look like in a hundred years. (Perhaps it’s worth shifting the emphasis from a dystopian vision of the future to a utopian one.)
Finally, Google invites you to delve into the past. The company demonstrated a demo with the prompt “create a hyper-realistic image of what the ruins of Pompeii looked like in 78 AD,” which could be used by, say, teachers. This would instantly transform today’s ruins “into a vibrant, colorful street scene from the Roman Empire,” according to Google.
However, these new features come with the usual AI image processing limitations. While the Nano Banana 2 can retrieve real-world information and images from the internet and process graphics with greater consistency and accuracy than previous models, AI still has its usual shortcomings: your lakeside home might not be entirely sturdy, and your Pompeii reconstruction might not be entirely authentic—so take these into account when creating your images.
When I asked Google about this, a company representative emphasized that Nano Banana 2 relies on a real-world knowledge base, which includes all the text and graphics found online, as well as a vast number of images fed to the AI during its training. Using Google Earth, for example, it will be able to trace the preserved ruins of Pompeii.
For greater clarity, you can try testing the Nano Banana 2 model in the main Gemini app: I asked it to render an imaginary scene from Pompeii just before the city was buried under Mount Vesuvius, then asked about the image’s accuracy. The AI responded that it combined a wealth of real-world historical data to render the scene, including details of Pompeii’s raised pavements and period clothing styles. Google hasn’t provided any specific usage restrictions, but the Nano Banana 2 update is already rolling out worldwide to all Google Earth web users.