If you’ve ever daydreamed about turning a vacant parking lot into a lush community park or wrapping the Statue of Liberty in wildflowers just to see how it looks, Google has quietly shipped the perfect toy for you: Nano Banana inside Google Earth. It’s playful, a little absurd, and surprisingly powerful – like handing a planet-sized sketchbook to anyone with a browser and an idea.

Google’s big new party trick is that you can now generate custom images directly on top of real satellite, aerial, and 3D imagery in Google Earth, powered by its Nano Banana 2 image generation model. In plain English, that means you pick any place on the globe, hit “create image,” type what you want to see – and Gemini’s Nano Banana goes to work reimagining that location while still respecting the underlying geography. So yes, it is finally time to start building gardens anywhere on Earth, from Times Square to your high school football field, without needing permits, budget approvals, or a single bag of fertilizer.

A garden in the clouds, grounded in reality

What makes this feel different from the usual “AI image generator” hype is the grounding in real-world data. Google Earth has spent years stitching together a photorealistic globe from satellite passes, aerial photography, and 3D city models, and Nano Banana is now layered directly on top of that foundation. Instead of creating fantasy landscapes from scratch, you’re remixing reality – keeping the streets, buildings, rivers, and coastline, but changing what sits on them.

In practice, the flow is almost suspiciously simple. You open Google Earth on the web, zoom into any spot – maybe that sad asphalt rectangle behind your local strip mall – and tap “create image.” From there, you type something like “transform this parking lot into a vibrant urban garden with raised beds, walking paths, and a small stage for community events,” and Nano Banana produces a rendered scene that matches your description while fitting neatly into the existing map. The outline of the lot stays the same; the surrounding roads and buildings are still where they should be, but your prompt turns the blank space into something alive.

That grounding is more than a technical detail – it’s the difference between daydreaming and prototyping. Teachers can ask Nano Banana to show students how Pompeii might have looked before the eruption, building a “hyper-realistic” reconstruction on top of the modern ruins. Urban planners can turn derelict plots into conceptual green spaces or test how a rooftop garden might change the feel of a block, all with a few lines of text. And regular people, the folks who will probably have the most fun with this, can drop whimsical gardens into their hometowns just to see what kind of vibrant, slightly surreal future they can imagine.

Nano Banana: from playful codename to serious tool

Nano Banana started life as one of those very Google-y codenames – the sort of thing you expect to see in an internal slide deck, not a product announcement. Underneath the banana jokes, though, is a genuinely capable image generation model, part of the broader Gemini family, tuned to produce detailed, playful, realistic, or downright whimsical scenes. It can even upscale outputs up to 4K, which matters once you stop treating these images as throwaway demos and start using them in presentations, proposals, or classrooms.

The Earth integration uses Nano Banana 2, the latest evolution of the model, and Google’s messaging around it is telling: this isn’t just a creative toy, it’s meant to be an everyday tool for people who need to visualize projects in context. In recent months, Google has been steadily threading generative AI through its geospatial stack – from Maps Imagery Grounding, which lets enterprise users conjure realistic Street View scenes for film sets or construction plans, to Earth AI imagery models that can automatically identify objects like bridges and power lines in satellite data. Nano Banana in Google Earth lands right at the intersection of those efforts, where “serious” geospatial analysis quietly meets “what if we turned this highway interchange into a pollinator-friendly park?”

This dual identity is where things get interesting. On one hand, Nano Banana is clearly pitched as a way to “visualize the past,” “reimagine an empty lot as a vibrant shopping district,” and generate professional real-estate plans, all language straight from Google’s own examples. On the other hand, there is nothing stopping you from typing “replace this industrial port with a massive botanical garden and floating greenhouses” and seeing what comes out. The underlying tech doesn’t care whether you’re a city planner or someone procrastinating on a Thursday night – it just sees coordinates, imagery, and a prompt.

The joy of building impossible gardens

It’s easy to talk about use cases in terms of “stakeholders” and “projects,” but Nano Banana’s Earth integration shines brightest when you approach it like a sandbox game. Imagine dropping into Manhattan and systematically turning every rooftop in a neighborhood into a pocket garden – herbs on one block, trees on the next, flower gardens near schools – just because you want to see what that version of the city might feel like. Or heading to a sunbaked suburb and transforming cul-de-sac lawns into shared vegetable plots and shaded community spaces.

There’s a particular thrill in zooming to a famous landmark and giving it an ecological makeover. Google uses the Statue of Liberty as an example for custom infographics, but with Nano Banana you can go further: render Liberty Island with dense native plantings instead of manicured lawns, add walking paths, bird habitats, or even solar-powered greenhouses along the shoreline. The underlying shape of the island stays true to the real world; the vegetation, structures, and textures respond to your imagination.

The “build a garden anywhere” idea works especially well because gardens are inherently contextual. A community garden under an elevated freeway looks different from a garden in a coastal village or a city square, and Nano Banana’s job is to blend your concept into the existing environment so it doesn’t feel totally disconnected from the setting. One prompt might produce wooden raised beds and murals under the viaduct; another might give you terraced plots climbing up a hillside, or floating garden pods on a lake surface. You’re not just editing a generic canvas – you’re remixing specific places, with all their quirks.

For educators and storytellers, that’s gold. A geography teacher can assign students different neighborhoods and have them imagine “climate-adaptive gardens” that respond to local conditions, then discuss what’s realistic and what’s aspirational. A journalist covering urban heat islands could use Nano Banana-generated visuals to show how tree cover or green roofs might transform a particularly vulnerable block. These aren’t engineering drawings, of course, but as conversation starters they’re more grounded than abstract diagrams and more engaging than spreadsheets.

Under the hood: Earth AI meets generative media

Behind the playful interface, Nano Banana in Google Earth sits on top of a serious geospatial AI stack that Google has been building out over several years. Google Earth AI, introduced as a family of “remote sensing foundation models,” is designed to handle large-scale satellite and aerial imagery analysis using vision-language models, open-vocabulary object detection, and adaptable vision backbones. Those tools help enterprises and nonprofits monitor everything from water levels and algae blooms to infrastructure conditions across huge regions.

In parallel, Google has been developing Maps Imagery Grounding and related features that let generative AI models “anchor” their creations to Street View and other mapping data, so that synthesized scenes respect the geometry and layout of real locations. When Google talks about Nano Banana “creating concepts grounded in the real world” inside Earth, it’s essentially describing that same idea applied to image generation over the globe.

According to Google, image generation with Nano Banana is available globally for Google Earth web users, and there’s no mention yet of strict usage caps, although real-world limits often show up quietly over time. For developers, there’s also an experimental Nano Banana integration documented for use with Google Earth imagery, hinting at future workflows where apps can automatically generate visual scenarios tied to coordinates or parcels. You can imagine real-estate sites using it to auto-render concept gardens on top of empty lots, or civic engagement platforms showing residents different green-space proposals contextually within their neighborhood.

That blend of generative media and geospatial grounding is part of a broader trend in mapping tech. At Google Cloud Next, the company showed off new AI features that shrink weeks of manual image review to minutes, and launched Earth AI imagery models capable of spotting specific objects across large datasets. Nano Banana feels like the more approachable, end-user-friendly face of that same stack: you don’t need to know what “Remote Sensing Foundations” are; you just need a rough idea and a location.

Reality, imagination, and the risk of pretty pictures

Of course, any time you give people a tool that can instantly beautify neglected spaces, there’s a tension between fantasy and reality. A Nano Banana garden doesn’t fix zoning rules, funding gaps, or political fights over land use, and it’s dangerously easy to fall in love with glossy renderings that never survive contact with actual budgets. Google, for its part, frames Nano Banana in Earth as a way to “visualize projects before any construction starts,” which is both pragmatic and a little optimistic.

The upside is that these images can make planning conversations more inclusive. It’s one thing to read that your city is considering a new park; it’s another to see a visual mockup of what that park might look like on the exact corner you walk past every day. Tools like Nano Banana lower the barrier to entry for that kind of visualization – you don’t need CAD skills, just a sentence or two describing what you’d like to see. Done thoughtfully, that can help communities sketch out possibilities, debate trade-offs, and push back on designs that look good in theory but feel wrong when you see them in situ.

There are also obvious ethical questions around historical visualization. Google highlights using Nano Banana to recreate scenes like “Pompeii in 78 AD,” which sounds great until you start wondering whose version of history is being encoded. AI-generated reconstructions are interpretations, not time machines, and teachers will need to emphasize that distinction when they drop these scenes into lessons. It’s the same caveat that already applies to AI in other domains, just dressed up in Roman architecture and satellite views.

Still, it’s hard not to feel a bit of optimism when you see people using generative tools to imagine more trees, more shade, more community spaces instead of more glass towers and parking decks. If nothing else, Nano Banana gives everyone a low-stakes way to explore those visions – to sketch out gardens on top of freeways, vacant malls, or even corporate campuses, and ask, “Why not?”

How to start tinkering with your own planet-scale garden

If you want to try this yourself, you don’t need a tutorial so much as a playful mindset. The controls are basically point, click, and type. Open Google Earth on the web, pick a place you care about – your neighborhood, your school, a park that feels underdeveloped – and zoom to the angle that gives you the best view of the area you want to change. Hit “create image,” and then treat the prompt box like a conversation with a very visual friend: “turn this empty field into a community food forest with winding trails and solar-powered lights,” or “replace these strip malls with mixed-use buildings and rooftop gardens.”

As you play, you’ll start to get a feel for how Nano Banana responds to different levels of detail. Specific adjectives and elements usually lead to richer, more coherent scenes – “native plants,” “accessible paths,” “rainwater collection,” “local art installations” – while vague prompts might give you something pretty but generic. And because the tool sits on top of real imagery, you can always re-prompt based on what you actually see: “keep the existing trees, but add raised beds along the south edge” or “turn the parking area into permeable pavers surrounded by flower beds.”

The outcome won’t be perfect, and that’s fine. The point here is less about engineering precision and more about sparking ideas. For a lot of people, it’s easier to react to a visual – “too many buildings, not enough shade,” “love the community stage, not sold on the fence” – than to blueprint-style diagrams or text-heavy proposals. By making those visuals cheaper and faster to create, Nano Banana effectively turns the entire planet into a sketchpad for public imagination.

Whether Nano Banana in Google Earth becomes a staple of serious planning workflows or settles into the role of “that fun thing people use to imagine gardens on highways” is still an open question. But right now, the barriers are low, the canvas is huge, and the prompts are limited mostly by how weird and hopeful you’re willing to get. If you’ve ever looked at a barren concrete lot and thought, “this could be a garden,” you finally have a way to see that version of reality – even if it starts as pixels before it becomes plants.