Google's Nano Banana AI image generator and editor has made steady progress since it was introduced in August. Its focus has always been on accuracy – not creating weird distortions or mangling text, as AI image generators have been prone to do. In my tests last month, I found it to be quite faithful, but conservative as a result.
For instance, it said repeatedly that it had colorized a black-and-white photo I gave it, although it actually hadn't. ChatGPT, for its part, handled that task with aplomb, but it also chose to change the composition and expression on the person's face. It looked good, but was also a fabrication.
Read more: Google's New AI Photo Editor Can't Match ChatGPT on Raw Power
Google has just announced a potentially huge upgrade of Nano Banana from version 2.5 to Nano Banana Pro (based on its Gemini 3 Pro AI model). Users of the free Gemini service get a limited number of credits or "quotas" – about enough for two projects, in my testing – then revert back to 2.5. Google AI Plus, Pro, and Ultra subscribers receive higher quotas.
Google makes big promises for the new model in its announcement, saying, "Nano Banana Pro can help you visualize any idea and design anything - from prototypes, to representing data as infographics, to turning handwritten notes into diagrams." But in a quick test of its advertised ability to create an infographic, I found it to be more of a work in progress – although it definitely has progressed from version 2.5.
As a test, I tried feeding Gemini a recent article I wrote about how to get and safely store the recovery key in case you get locked out of your encrypted Windows 10 or 11 hard drive. Then I asked it to create an infographic from the piece. The old version of Nano Banana (2.5) in my iPhone 16's Gemini app accepted a URL to the article and got to work. But it also went out to other websites in order to improve my work, which was not an improvement. When I told it to look only at my article, it produced an infographic full of made-up words.
Better text accuracy is one of Google's selling points for Nano Banana Pro, and better seemed to be the case when I fed a PDF of the same article to Pro. (It would not accept a URL.) The resulting infographic had all the elements right, but messed up the flow a bit. It basically sent you in endless circles. It did invent one word: "Enalled." I believe that was meant to be "Enabled."
I then asked it to create a more detailed, step-by-step version. It had perfect text and pretty images, although some were pretty weird, like a faceless man in a suit taking a photo with a smartphone. That was not one of the steps, and most of the real steps were missing.
In fairness, it was not a simple article. A detailed back-and-forth conversation with Gemini might yield something better.
Nano Banana Pro promises other advanced features I haven't yet had a chance to test, such as compositing. For instance, it can accept individual images and bring them into one image generated from your prompt. Google shows an example with thumbnail images of 14 bizarre, multicolored furry creatures (and a goat). A 105-word prompt puts them all together in a cozy cartoon living room.
Nano Banana Pro also offers advanced photo editing. For instance, you can adjust camera angles, change the focus, apply lens effects like shallow depth of field, and change the lighting, even switching from day to night or vice versa. I tried less-sophisticated versions of some of these capabilities in Nano Banana 2.5 in Google Photos on my iPhone a few weeks ago, including changing from day to night. It was entertaining, but I'm looking forward to trying the new and improved capabilities.
Read more: Google Brings Gemini AI to Mobile Photos App for Wild Edit Options
AI image generators/editors are moving fast and still breaking a few things. But the capabilities are steadily improving, and I look forward to the inevitable refinements that will keep coming.
[Image credit: Google, Sean Captain/Techlicious via Google Nana Banana Pro]
