Nano Banana 2: The new version of the image generator in Gemini – AI News – #1 March 2026

Last year, the original Nano Banana model revolutionized the way we approach generating and editing graphics using artificial intelligence. Shortly after, the Pro version provided users with studio-quality outputs and advanced control over their images. Today, Google is combining the best of both worlds.

Nano Banana 2 introduces the intelligence and visual quality previously reserved for Pro models, but it delivers them with the speed of Gemini Flash. As a result, the process of iterating, making corrections, and testing new prompts becomes incredibly smooth and efficient.

The tool has been heavily optimized for professional workflows. From the perspective of the creative and SEO industries, the most important improvements include:

How does Nano Banana 2 fare against user expectations? Based on initial general tests, there is a clear leap in quality, although it is not without a few stumbling blocks typical of the current stage of generative AI development.

First and foremost, the model’s ability to edit existing photos is noteworthy. When uploading a personal photograph and asking the AI to place the subject in a completely new, abstract, or specific environment, the tool can preserve an impressive amount of detail. In many cases, Nano Banana 2 flawlessly recreates clothing patterns or small jewelry, even in areas that were hidden in the original photo and required the algorithm to “draw” them in. The backgrounds themselves are highly photorealistic and convincing.

Generating infographics looks equally promising. The tool cleanly integrates generated text with graphical elements, avoiding the distorted letters or bizarre background artifacts that plagued older AI models. When we tested the model for Polish infographics, there were still a few distortions or illegible words, however the English results are very promising:

Despite its integration with the search engine, fetching real-time data still requires human verification. The model can occasionally pull outdated information (e.g., last week’s weather forecast instead of the upcoming weekend’s). Fortunately, a quick text correction directed at the chatbot is usually enough for it to instantly swap the data for the correct ones.

Another remaining challenge involves complex photomontages of human figures. Attempts to “paste” a face from a standard selfie onto the body of a professional athlete in a dynamic pose can yield comical results, resembling amateur work in a graphics program. The face is sometimes lit from a different angle than the rest of the body, which immediately gives away the image’s artificial origins.

Nano Banana 2 is already available and is gradually replacing the Pro version across many Google services. You will find it as the default tool in the Gemini app, in Search’s AI Mode, as well as in Google Ads (where it supports campaign creation) and AI Studio. For users on paid plans (AI Plus, Pro, and Ultra), the Nano Banana Pro model remains accessible for the most demanding tasks—simply generate an image with the new version, click the three-dot menu, and select the option to redo it with Pro.

Along with increasing photorealism, Google is putting a massive emphasis on content labeling. Every generated piece of media is secured with SynthID technology (an invisible watermark). Soon, C2PA Content Credentials will also be integrated into the Gemini app, allowing users to accurately track the history and modification methods of any given image.

Stay up to date with AI and SEO news! Subscribe to the Delante newsletter to regularly receive expert knowledge, analyses of the latest algorithms, and practical tips that will help you grow your online business. Do not miss any important updates – join our community today!

Source: https://blog.google/innovation-and-ai/technology/ai/nano-banana-2/