Google rolls out Nano Banana 2 with sharper images, improved prompt controls

Google Nano Banana

• Google launches Nano Banana 2 image model for Gemini app
• Model improves visuals, prompt control, and text rendering.
• Nano Banana 2 adds AI Mode, Google Lens, and developer tools

Did our AI summary help?

Google has announced Nano Banana 2 as its latest image generation model, now rolling out to the Gemini app. Officially named Gemini 3.1 Flash Image, the new model replaces Nano Banana Pro across most Gemini experiences while promising sharper visuals, improved instruction following and production-ready controls.

Google says Nano Banana 2 builds on last year’s Gemini 2.5 Flash Image, originally codenamed Nano Banana, and November’s Gemini 3 Pro Image, also known as Nano Banana Pro.

Sharper images and stricter prompt control

Compared to its predecessor, Nano Banana 2 delivers richer textures, sharper details and more vibrant lighting. Google claims the model adheres more strictly to complex prompts, capturing specific nuances so that the generated image more closely matches the original request.

A key upgrade is improved subject consistency. The model can maintain character resemblance for up to five characters and preserve the fidelity of up to 14 objects within a single workflow. This is particularly relevant for storyboarding, marketing campaigns or visual narratives where continuity matters.

The Flash model now also includes features previously reserved for the Pro tier, such as accurate text rendering and translation, along with deeper world knowledge to render specific subjects more accurately. That enables use cases ranging from data visualisations and infographics to turning handwritten notes into diagrams or generating marketing mockups.

Production-ready specifications

Nano Banana 2 offers control over multiple aspect ratios and resolutions, ranging from 512 pixels up to 4K. That flexibility allows creators to generate assets suited for vertical social posts, widescreen presentations or larger display formats without sacrificing clarity.

Within the Gemini app, Nano Banana 2 now replaces Nano Banana Pro across the Fast, Thinking and Pro model options. However, Google AI Pro and Ultra subscribers will still retain access to Nano Banana Pro for specialised, high-fidelity tasks by regenerating images through the three-dot menu.

Google positions Nano Banana 2 for rapid generation, while Nano Banana Pro remains focused on maximum factual accuracy and higher-end detail requirements.

Expanding across Google’s AI ecosystem

Nano Banana 2 is also coming to AI Mode and Google Lens, and it becomes the default image generation model in Flow. The model is available in preview through AI Studio, the Gemini API, Google Antigravity, Vertex AI and Gemini CLI, signalling a broad rollout across Google’s developer and enterprise platforms.

Safety and verification upgrades

On the safety front, Google is combining its SynthID watermarking system with C2PA Content Credentials. The goal is to provide clearer transparency around not just whether AI was used, but how it was used in generating an image. C2PA verification support is also coming to the Gemini app.