Recently, a remarkable development has taken place in the realm of AI image editing. A groundbreaking model named nano banana has emerged, captivating the attention of users and securing its position at the top of the LMArena leaderboard. Google has officially announced that this innovative model is a creation of Google DeepMind and is now being integrated into the Gemini app as of today.
Earlier this year, Google introduced editing capabilities within the Gemini app, allowing users to modify images simply by providing a prompt, eliminating the need for complex software like Photoshop. The initial model displayed impressive competence; however, like many generative systems, it exhibited a non-deterministic nature, resulting in unpredictable changes to certain elements of images. With the introduction of nano banana, or more technically, Gemini 2.5 Flash Image, Google promises an unparalleled level of consistency across edits, enabling the model to retain crucial details rather than producing random variations with each adjustment.
The enhanced capabilities of the nano banana model open up numerous exciting applications for AI image editing. For instance, users can upload a photo of an individual and transform their style or attire. Imagine reimagining someone as a matador or a character from a '90s sitcom. Thanks to the model's ability to maintain consistency through edits, the resulting images will still resemble the original person, even when multiple edits are made successively. Google assures that the final images will retain the likeness of the source material throughout the editing process.
The Gemini app's advanced image editing capabilities extend beyond simple transformations. Users can merge multiple images to produce entirely new creations tailored to their preferences. For example, Google showcased a scenario where separate images of a woman and a dog were combined, resulting in a heartwarming snapshot of the dog being cuddled—possibly one of the most delightful applications of generative AI to date. Furthermore, Gemini's functionality allows for abstract merging of elements, adhering to user prompts to craft a wide variety of images, all while complying with the model's protective guidelines.
As with previous Google AI image generation models, outputs from the Gemini 2.5 Flash Image will always feature a visible AI watermark in the corner. In addition, each image is embedded with an invisible SynthID digital watermark, which remains detectable even after modest modifications. This ensures transparency and accountability in the use of AI-generated content.
You can begin experimenting with the new native image editing features in the Gemini app starting today. Google also plans to roll out this innovative image model soon through the Gemini API, AI Studio, and Vertex AI for developers, expanding access to these cutting-edge tools and fostering creativity in AI image editing.
