Google launches Nano Banana Pro image model built on Gemini 3 Pro

Google launched Nano Banana Pro, an image generation and editing model built on its new Gemini 3 Pro foundation. The model renders legible text inside images, accepts up to 14 image inputs at once, preserves the likeness of up to 5 people across blends, and outputs at 2K or 4K resolution.

The new model succeeds the original Nano Banana, which Google shipped earlier in 2025 as Gemini 2.5 Flash Image. Google described the original release as a tool that "empowered casual creators to express their creativity" through tasks like restoring old photos and generating miniature figurines. Nano Banana Pro targets a more demanding user: professionals, advertisers, filmmakers, and developers.

"Across our products and services, you now have a choice," Google said: the original Nano Banana for "fast, fun editing," or Nano Banana Pro for "complex compositions requiring the highest quality and visually sophisticated results."

What can Nano Banana Pro do?

Nano Banana Pro is Google's latest attempt to make generative imagery useful for actual knowledge work, not just novelty renders. The model can connect to Google Search to pull real-time information — recipes, weather forecasts, sports scores — directly into the visuals it generates. That makes it possible, Google said, to "create a quick snapshot for a recipe or visualize real-time information like weather or sports" inside a single image.

Google called Nano Banana Pro "the best model for creating images with correctly rendered and legible text directly in the image." The model supports short taglines and long paragraphs, multiple languages, and stylized typography including calligraphy. Google tied the upgrade to Gemini 3's ability to "understand depth and nuance," which the company said unlocks new image-editing possibilities around type. "With Gemini's enhanced multilingual reasoning," Google added, users "can generate text in multiple languages, or localize and translate your content so you can scale internationally."

The model also raises the ceiling on multi-image blending. It accepts up to 14 input images and maintains the consistency and resemblance of up to 5 people. That makes it possible to turn sketches into product mockups or blueprints into photorealistic 3D structures while keeping a brand's visual identity consistent across touchpoints. The feature targets marketers trying to keep a campaign's look uniform across formats.

Creative controls expanded alongside the blending limits. Users can adjust camera angles, change focus, apply color grading, and shift scene lighting — for instance, turning a daytime shot into a night scene or producing a bokeh effect. Output is available in multiple aspect ratios and at 2K or 4K resolution, which Google said prepares "creations ready for any platform, from social media to print."

Where can you access it?

Google is distributing Nano Banana Pro across its consumer, professional, and developer products at the same time, with access governed by subscription tier.

• Consumers and students: Nano Banana Pro is rolling out globally in the Gemini app under the "Create images" option when the "Thinking" model is selected. Free-tier users get limited quotas, after which the app reverts to the original Nano Banana. Google AI Plus, Pro, and Ultra subscribers receive higher quotas.
• AI Mode in Search: Available in the U.S. for Google AI Pro and Ultra subscribers.
• NotebookLM: Available globally to subscribers.
• Google Ads: Image generation is being upgraded to Nano Banana Pro globally, putting "cutting-edge creative and editing power directly into the hands of advertisers."
• Workspace: Rolling out in Google Slides and Vids.
• Developers and enterprise: Starting to roll out in the Gemini API, Google AI Studio, and Google Antigravity for UI layouts and mockups. Enterprises can build in Vertex AI now; Gemini Enterprise support is "coming soon."
• Creatives and filmmakers: Rolling out to Google AI Ultra subscribers in Flow, Google's AI filmmaking tool, to give "creatives, filmmakers and marketers even more precision and control over their frames and scenes."

How does Google mark AI-generated images?

Google continues to embed every image its models produce with its imperceptible SynthID digital watermark. The company is also exposing a new verification step to consumers: any user can upload an image into the Gemini app and ask whether Google AI generated it. The verifier starts with English-language image prompts; Google plans to extend it to more languages, audio, and video.

A visible watermark — the "Gemini sparkle" — will remain on images generated by free and Google AI Pro tier users. Google will strip the visible watermark from images generated by AI Ultra subscribers and from output produced inside Google AI Studio, citing "the need for a clean visual canvas for professional work."

Why does the release matter?

The launch puts Google in direct competition with the image generation capabilities embedded in OpenAI's ChatGPT and the dedicated tools shipped by Adobe and Midjourney. By building Nano Banana Pro on Gemini 3 Pro's reasoning layer, Google is betting that image models need grounding in real-world knowledge to serve professional workflows — recipe cards, sports explainers, localized marketing assets — rather than one-off creative renders.

The tiered access also clarifies where Google expects revenue to grow. AI Ultra subscribers get the watermark-free, higher-quota experience across Flow, NotebookLM, and the Gemini app. Free and Pro users see a visible Gemini sparkle on every image they generate. That pricing structure pushes professional image work toward Google's top subscription tier.

Google's broader SynthID rollout — including the new in-app verifier — lands as regulators in the European Union and United States press AI providers to label synthetic media. The in-app tool only confirms images made by Google's own models, leaving users with no way to verify content from OpenAI, Midjourney, or other vendors. Whether that scope will satisfy forthcoming provenance rules remains an open question the company has yet to address publicly.