Google has launched Nano Banana 2, its latest image generation model, now the default for image creation across its Gemini applications and Google Search, the company announced today. The new model, technically Gemini 3.1 Flash Image, promises faster image generation and improved realism compared to its predecessors.
The rollout marks a rapid evolution in Google’s AI image capabilities. The original Nano Banana was released in August 2025, quickly gaining traction, particularly in India, with users generating millions of images within the Gemini app. In November, Google introduced Nano Banana Pro, offering higher fidelity and detail. Nano Banana 2 aims to bridge the gap, retaining the quality of the Pro version although significantly improving speed, according to Google.
Users of Nano Banana 2 can generate images ranging in resolution from 512px to 4K, with support for various aspect ratios. A key feature of the new model is its ability to maintain consistent character depictions across up to five individuals and track up to 14 objects within a single image, enhancing storytelling possibilities. Google states that the model also excels at interpreting complex prompts with nuanced details, producing images with vibrant lighting, richer textures, and sharper definition.
The integration of Nano Banana 2 extends beyond the core Gemini app. We see now the default image generation model within Google’s video editing tool, Flow. Google is deploying the model to power image results through Google Lens and AI Mode in Google Search, reaching users in 141 countries on both desktop and mobile platforms.
Subscribers to Google AI Pro and Ultra plans will continue to have access to Nano Banana Pro for specialized tasks, with the option to regenerate images using a three-dot menu within the interface. For developers, Nano Banana 2 is available in preview through the Gemini API, Gemini CLI, and the Vertex API, as well as through Google’s AI Studio and Antigravity development tool, released last November.
To address concerns about the authenticity of AI-generated content, Google is embedding all images created with Nano Banana 2 with a SynthID watermark. This watermark, initially launched in November, has already been used over 20 million times, according to the company. The images are also designed to be compatible with C2PA Content Credentials, a standard developed by an industry consortium including Adobe, Microsoft, Google, OpenAI, and Meta.