The video reviews Nano Banana 2, a new image generation model based on Gemini 3.1 Flash, highlighting its strong reasoning, accurate text rendering, and improved image quality that rivals the more expensive Nano Banana Pro while remaining smaller, faster, and cheaper. It introduces new features like expanded aspect ratios, better character consistency, and advanced tools, making it a cost-effective and versatile option for high-quality image generation in various Google apps.
The video introduces Nano Banana 2, a new image generation model based on the Gemini 3.1 Flash image. It is positioned as a smaller, faster, and cheaper alternative to the Nano Banana Pro, which uses the Gemini 3 Pro image. The main story is that Nano Banana 2 offers image quality that is significantly better than the previous Flash 2.5 model and often comes close to matching the Nano Banana Pro, making it a compelling option for users who want high-quality results at a lower cost.
One of the standout features of Nano Banana 2 is its strong reasoning ability over images, similar to Nano Banana Pro. The model excels at following precise instructions and rendering text accurately within images. For example, it can generate images of cats holding signs with specific messages, and it handles text in multiple languages, such as Italian and Thai, with impressive accuracy. However, the font style may change depending on the language, sometimes losing the original handwritten feel.
When compared directly to Nano Banana Pro, Nano Banana 2 performs well in many scenarios but occasionally falls short in terms of image perspective or detail. For instance, in a test involving a caveman taking a selfie with a T-Rex, Nano Banana 2 produced logical and coherent images, but the perspective was not always as accurate as the Pro version. The video suggests that providing more detailed prompts can help achieve better results with Nano Banana 2.
Nano Banana 2 also introduces several new features and improvements. It supports full aspect ratio options, resolutions down to 512 pixels, and two levels of “thinking” or reasoning. Notably, it includes access to tools like grounding with Google Search, which is also available in Nano Banana Pro, and a new image search tool exclusive to Nano Banana 2. The model has improved character consistency and can now use up to 14 reference images for generating complex scenes, making it ideal for product images or scenarios requiring multiple consistent elements.
In terms of pricing and availability, Nano Banana 2 is more affordable than Nano Banana Pro but slightly more expensive than the older version. It will be integrated into various Google apps, including the Gemini app, Anti-Gravity AI Studio, and Google Cloud’s Vertex AI. The video concludes by highlighting Nano Banana 2’s excellent text rendering and versatility, making it a strong choice for production use where cost and scalability are important. The presenter encourages viewers to try the model and share their experiences, noting that the release is promising for those awaiting the Gemini 3.1 Flash model.