Google’s fresh AI image maker took my shirt off

I tried Google’s fresh Nano Banana Pro and immediately took off my clothes. I didn’t ask for it, but the AI ​​model apparently decided that my greeting card would look better with more skin.

Nano Banana Pro, as the name suggests, is aimed at professionals. Powered by Gemini 3, it is actually an improvement on the company’s popular image generation and editing tool, which has become popular on social media amid a trend that has turned selfies into hyper-realistic 3D figurines. Google says it allows you to create higher-quality images that can be printed, render readable text in photos, and combine multiple images into a single composition. It is also intended for “people who want to feel like a professional,” said Naina Raisinghani, product manager at Google DeepMind Edge. Sounds good because I’m not a professional by any means. For me, the results were shiny but silly. It looked good, but felt amateurish.

Using Nano Banana Pro is quite straightforward: you go to the Gemini app, select “create images” and turn on “thinking” mode. Just plug in the prompt (and image if you’re using one) and you’re ready to go. It’s also free, although there are limits, with quotas increasing for Google AI Plus, Pro and Ultra subscribers.

Google makes bold claims, promising “studio-quality designs,” “flawless text rendering,” and a host of clever and artistic changes. To test this, I uploaded a straightforward photo of myself nearby The Verge office in Fresh York with the Brooklyn Bridge in the background. I asked Gemini to change the lighting from daylight to nighttime and it did a pretty good job. The result looks credible. It even handles details that often confuse image generators, such as whether cars are going in the right direction. Adjusting the camera angle was just as simple. I asked Gemini to recreate the shot as if it had been taken from a higher angle on the right, and it did.

Google also claims that Nano Banana Pro can create infographics and diagrams that aid visualize real-time information such as weather and sports. Being British, I asked about the weather for the next four days in Washington and Fresh York, where I am currently staying. Visually, the infographic would fit perfectly into a basic forecast website. The text and numbers looked normal—a far cry from the garbled nonsense you often see in AI-generated images—and at the end, Gemini gave me a list of quotes that helped me confirm it was precise.

The model stumbled a bit with more complicated tasks. I asked for a summary of recent events Edge a story about how Europe is clamping down on its artificial intelligence and privacy regulations in a comic book-like format. The images and text were indeed reproduced flawlessly, using a cartoonish font, but the comic didn’t summarize the story at all, instead giving a vague overview of the bloc’s artificial intelligence bill. The problem may have been because I linked to the Gemini story instead of pasting the text.

Once I did this, I got a satisfactory comic-book-style summary. It conveyed the essence of the actual story, although I don’t think I would have been able to understand it easily if I hadn’t written the source material. It also included phrases that didn’t appear anywhere in my article.

To really feel like a professional designer, I tried my hand at creating greeting cards. Christmas is finally approaching. Considering I only uploaded three selfies, Gemini did a truly amazing job of creating three versions of my full body, each in different outfits and with different facial expressions. He also created a realistic snowy scenery with Christmas trees, as I requested, and decorated it with a “Merry Christmas!” upstairs, just like I asked.

Gemini took liberties when I asked to change the card’s snowy background to a summer beach for an Australian-style vacation. These freedoms were my deeply imitation clothes: two of my clones were topless. It was strange. There were also some significant AI-generated feet and a smiling sandman that replaced the snowman from the winter scene (built by my topless doppelgänger). However, there were some issues – the sandman lacked shadow, unlike the other rendered objects in the photo, and the Christmas lights on the palm trees magically glowed in the dazzling sunlight. I tested his precision editing skills by asking him to add some force to just one clone, which he did in a matter of seconds (if only it were that simple in the real world). Overall, the quality was excellent and the image would be somewhat believable (minus the abs) if you didn’t know that a immense tattoo was missing on my chest.

However, not everything was great. The model failed to keep the exact text on my card that I asked for. Instead of “Merry Christmas!” chose “Aussie Summer Christmas!” He also seems to have problems with animals: my sister’s cat sits in the exact same pose on stilts as the reference image I provided in each version of the card (he got a fancy Santa hat, though).

All in all I was impressed. The Nano Banana Pro is a clear improvement on the basic model. I was able to request more precise changes, which actually produced understandable text, removing a huge obstacle to the operate of such generative AI tools in the real world. But unfortunately these qualities were not enough to make me a good designer.