Google Earth adds Nano Banana 2 AI image generation on web

Google Earth now lets users generate AI‑crafted images directly in the web app using Gemini’s Nano Banana 2 model, a feature that rolls out globally today and is limited to a few generations per day per account [1].

How the feature works

The new “Create image” button taps Gemini’s Nano Banana 2 model, which blends satellite, aerial and 3D map data with web‑scale knowledge to produce visuals ranging from historic reconstructions to speculative future scenes [3]. Users zoom to a location, enter a text prompt, and the AI returns a rendered image that can be refined and saved. Unlike earlier workflows that required copying images into a Gemini chat, the process is now embedded in Earth, streamlining the creative loop [1].

Limits and rollout quirks

At launch, the capability is confined to the web version of Google Earth; the Android app, despite its 500 million+ downloads, does not receive Nano Banana 2 [1]. Daily usage caps apply, and each generation can take up to two minutes, reflecting the model’s compute intensity [1]. Google has not disclosed exact quota numbers, but the restriction ties the feature to a separate subscription tier distinct from Google One [1]. The rollout is global, but early adopters may encounter latency or need to restart prompts if they wish to adjust the view after generation [1].

Competitive context

Nano Banana 2 represents a step up from the original Nano Banana model that first showcased multimodal AI on Earth, offering higher fidelity and broader knowledge integration [2]. Competitors such as Microsoft’s Copilot and Adobe’s Firefly provide image generation, but they typically require a blank canvas rather than anchoring output to real‑world geography. By coupling AI output with actual map data, Google aims to differentiate its offering for use cases like urban planning, education and speculative design [3][4].

What to watch

• Expansion to mobile – Whether Google will extend Nano Banana 2 to the Earth Android app, addressing the current “desktop‑only” limitation.
• Quota adjustments – Changes to daily generation limits or pricing tiers could affect adoption among developers and educators.
• Competitor responses – Launch of similar geo‑anchored AI tools from rivals could pressure Google to accelerate feature enhancements.

The launch adds a creative layer to Google Earth, turning a static map into a sandbox for visual speculation. How the daily caps and web‑only availability shape user uptake will determine whether the feature becomes a niche novelty or a mainstream planning aid.