In August 2025, a mysterious image model called “nano-banana” started topping the anonymous leaderboard on LMArena, where people vote on which AI did better without knowing whose AI it is. Speculation ran wild. It turned out to be Google’s, and the name stuck so firmly that Google ended up using it in its own announcements. The official name, Gemini 2.5 Flash Image, never stood a chance.
Your face, still your face
AI image generators had long been great at making a picture and bad at editing one. Ask for a new haircut and you often got a new person. Nano Banana’s selling point was consistency: change the outfit, the background or the lighting, and your friend, your grandmother or your chihuahua in a tutu still look like themselves. You can also blend several photos into one, or apply the style of one image to another, just by describing what you want.
The figurine craze
The internet did what the internet does. Huge numbers of people turned their selfies, pets and holiday photos into shiny collectible figurines, complete with a box and a display stand, and the trend spread across social media in weeks. For developers, it cost about four cents an image through Google’s API.
The less cute part
A tool that can convincingly edit real people can also be used to put them where they never were. Google adds a visible watermark in the Gemini app and an invisible SynthID watermark to images it creates or edits, so they can be identified as AI-made. That helps, but only if platforms check for it, and only for tools that play along.
Tip: have fun with your figurine, but think twice before editing photos of people who did not ask to be edited.



