A mysterious project of “ nano-banana ” revealed to be the last image editor of Google

Google has just upgraded its AI image model, and it actually seems to be a fairly important step. The company has deployed Gemini 2.5 Flash Image today, a major refresh that promises a more intelligent and more flexible generation of images.

The improved model allows users to issue invites in natural language not only to generate images, but also to merge existing photos and make more precise modifications without creating strange distortions. He also draws from the “global knowledge” of Gemini to better understand what it generates.

This upgrade comes as Google tries to fill the gap with the industry leader, Openai. In the past, the generation of images has been a major engine for AI. The use of chatgpt soaped in March when the company launched its GPT-4O native image generator. The memes viral studio ghibli generated by the model resulted in the melting of the company’s GPU models, according to the CEO of Openai, Sam Altman. Chatgpt currently has more than 700 million weekly users. In comparison, the CEO of Google, Sundar Pichai, revealed during the call of the July profits of the company that Gemini had 450 million monthly users, still dragging behind Chatgpt.

https://www.youtube.com/watch?v=bi-wi2q_xog

With its last update, Google says that it has resolved one of the largest headaches in AI. Until now, keeping characters or coherent objects on several changes has been a major challenge for IA image generators.

“”You can now place the same character in different environments, present a single product from several angles in new parameters or generate coherent brand assets, while preserving the subject, ” The company wrote in a blog article.

Google says that users can now make very specific adjustments with just a prompt. For example, users can blur the background from an image, delete a stain from a t-shirt, modify the installation of a subject or even add color to a black and white photo.

Even before its official launch, the new model turned its head on the Crowdsourced Lmarena evaluation platform, where it appeared anonymously under the name of “nano-banana”. An X user shared how they used Nano-Banana to change the Altman shirt in a photo. The result was surprisingly good. Today, Google has advanced and claimed the property of the model, revealing that Nano-Banana was in fact the Flash Gemini 2.5 image.

In addition to being available on the Gemini application, the new model is now accessible to developers via the Gemini API, Google AI Studio and Vertex AI. Google has already built several model applications that use the new model on Google AI Studio, the company’s coding assistant, and have said that users can vibrate in addition.

The company also said that some developers had already experienced the application to see how it would be useful in the real world scenarios, such as the creation of real estate rating cards, uniform employees and product models.