Google Enhances AI Image Editing with Advanced Nano Banana Model

Recently, a notable development emerged in the realm of AI image editing with the introduction of a new model known as “nano banana.” This innovation, resulting from Google’s DeepMind, has demonstrated remarkable capabilities that have propelled it to the top position on LMArena’s image-editing leaderboard. Google has officially announced that the nano banana model will be rolled out to the Gemini app.

AI image editing allows users to modify images simply by providing textual prompts, eliminating the need for traditional software like Photoshop. Earlier this year, Google introduced editing features in the Gemini app, which showed promising performance right from the start. However, the unpredictable nature of generative systems often caused some elements of images to change in unexpected ways. With the new nano banana model, technically referred to as Gemini 2.5 Flash Image, Google claims to offer unparalleled consistency across edits. This means the model can remember details and maintain coherence through consecutive changes.

One of the significant advantages of the nano banana model is its ability to retain the appearance of subjects during edits. Users can upload photos of individuals and alter their style or attire; for instance, a person could be reimagined as a matador or a character from a classic ’90s sitcom. Thanks to the model’s consistent editing capabilities, the resulting images will still resemble the original individual, even after multiple modifications.

Moreover, Gemini’s advanced image editing features enable the merging of multiple images to create new compositions. For instance, separate images of a person and a pet can be integrated to produce a new heartwarming snapshot. This innovative application of generative AI might be considered one of its most fascinating uses. Additionally, the model can execute more abstract merges as per user prompts, generating a wide array of images while adhering to the model’s guidelines.

As is standard with Google’s AI image generation models, all outputs from the Gemini 2.5 Flash Image come with a visible “AI” watermark. Additionally, there is an invisible SynthID digital watermark embedded in the images that remains detectable, even upon moderate alterations.

You can experience the enhanced native image editing capabilities today by using the Gemini app. Google has also hinted that this new image model will soon be available in the Gemini API, AI Studio, and Vertex AI for developers, broadening its accessibility.