Google launches Nano Banana 2 model with faster image generation

Summary

Google has announced Nano Banana 2, the latest version of its image generation model. The company says the new model — technically called Gemini 3.1 Flash Image — can produce more realistic images than the previous Nano Banana while also generating them faster.

With this launch, Nano Banana 2 becomes the default image generation model in the Gemini app across its Fast, Thinking, and Pro modes. Google positions the update as a speed-focused default option that still retains much of the visual fidelity users expect.

According to Google, Nano Banana 2 keeps some of the high-fidelity characteristics of Nano Banana Pro but improves generation speed. The company says it can create images at resolutions ranging from 512px up to 4K, supports different aspect ratios, and can handle complex, nuance-heavy image requests.

The rollout extends beyond Gemini. Google says Nano Banana 2 becomes the default image generator in Flow, its video editing tool, and in Search. In particular, the company states it will be the default for Google Search results via Google Lens and in AI Mode across 141 countries, available in the Google app and on the web across desktop and mobile.

For subscribers on Google’s higher-end plans — Google AI Pro and Ultra — Google says Nano Banana Pro remains available for specialized tasks, accessible by regenerating images from the three-dot menu.

In practice

In practice, Google is emphasizing a balance of speed and quality. The company says Nano Banana 2 supports outputs from 512px up to 4K and can generate images in different aspect ratios, which can matter for teams producing assets for multiple placements and formats.

Google also highlights consistency features aimed at storytelling workflows. The company says Nano Banana 2 can maintain character consistency for up to five characters and preserve the fidelity of up to 14 objects within a single workflow. That framing suggests fewer unwanted variations when iterating on scenes, characters, or complex compositions.

On prompt handling, Google says users can issue more complex requests with detailed nuances. Visually, the company claims improvements such as more vibrant lighting, richer textures, and sharper detail compared with prior versions.

Access is broad by default: Nano Banana 2 becomes the standard image generation model throughout the Gemini app and in Flow. In Search, Google says the model becomes the default behind experiences tied to Lens and AI Mode in 141 countries, spanning use in the Google app and on the web across desktop and mobile.

For developers, Google says Nano Banana 2 will be available in preview via the Gemini API, Gemini CLI, and the Vertex API. The company also says it will be available through AI Studio and its developer tool Antigravity, which Google notes it released last November.

Context

Google first released Nano Banana in August 2025. According to the report, that launch led people to generate millions of images in the Gemini app, particularly in countries such as India. In November, Google introduced Nano Banana Pro, positioned as a way to create more detailed, higher-quality images.

Nano Banana 2 now becomes the default model across Google’s consumer-facing entry points, while Nano Banana Pro is kept as an option for specialized use on higher-tier subscriptions (Google AI Pro and Ultra) via a regenerate action in the three-dot menu.

Google also ties the launch to content provenance. The company says all images generated through the new model will include a SynthID watermark, Google’s marker indicating AI-generated imagery. Additionally, Google says the images are interoperable with C2PA Content Credentials, a standard developed by an industry group that includes companies such as Adobe, Microsoft, Google, OpenAI, and Meta.

As an adoption datapoint, Google says that since launching SynthID verification in the Gemini app in November, people have used it more than 20 million times.

Why it matters

• Making a faster model the default across Gemini, Flow, and Search could rapidly expand usage and normalize AI image generation in everyday workflows.
• Google is claiming improvements that target real-world creative iteration: higher realism, better prompt nuance handling, and consistency across multiple characters and objects.
• Preview availability through Gemini/Vertex tooling and AI Studio/Antigravity signals an intent to drive developer adoption and integrations.
• SynthID watermarking and C2PA compatibility reinforce provenance tooling as AI-generated media becomes more widespread.