Something unusual just shook up the AI image‑editing world. A quirky‑named model called “nano banana” has been making waves with skills so impressive it shot straight to the top of the LMArena leaderboard. Now, Google has confirmed the mystery model is actually an in‑house innovation from Google DeepMind — and it’s rolling out to the Gemini app starting today.
AI‑powered image editing lets you tweak photos by simply typing a prompt, skipping the hassle of manual Photoshop work. When Google first launched editing features in Gemini earlier this year, the results were strong. But, like most generative systems, edits could be unpredictable — small details might change in ways you didn’t ask for.
That’s where nano banana (officially Gemini 2.5 Flash Image) comes in. Google says it delivers unmatched consistency, remembering the image’s details instead of taking a fresh gamble with every edit.
What This Means for Your Photos
- Upload a picture of a person, then change their clothes, hairstyle, or even transform them into a matador or a ‘90s sitcom star — without losing their likeness.
- Make multiple edits in a row, and the subject will still look like the original photo.
- Merge two or more images to create something new — for example, Google’s demo combined a woman and a dog into a cozy cuddle scene (arguably, peak generative AI use).
- It even works for more abstract mash‑ups, as long as prompts stay within the model’s guidelines.
Built‑In Safeguards
Every image made with Gemini 2.5 Flash Image comes stamped with a visible “AI” watermark in the corner, plus an invisible SynthID digital watermark that survives moderate editing.
How to Try It
You can test the new editing powers in the Gemini app right now. Developers won’t be left out — Google says the model will soon hit the Gemini API, AI Studio, and Vertex AI.
