It’s been three whole months since Google dropped its latest AI image model, so given the breakneck pace of AI development, we were overdue. Google delivered, with Nano Banana 2, released in late February.
At first glance, not much had changed from Google’s previous models. Google said the second-generation model would have the speed of the original model but the benefit of the pro’s (and Gemini 3’s) world knowledge for more realistic outputs. As CNET’s creative AI expert, I had to put the new model through its paces. Nano Banana Pro, after all, is our best overall pick among a crowded field of AI image tools — even if it did make me wonder if AI tools have gotten too good.
After my testing, I found that Google was mostly right. Nano Banana 2 was a speedy companion and reminded me why Google is still one of the biggest beasts in the AI game. But I did find cracks in its performance, proving there’s still room for improvement.
This is how Nano Banana 2 fared against the original and the pro model. A note about process: I limited my testing to photo editing, since that is one of the biggest use-cases for Nano Banana. You can use it to create entirely AI-generated images, though.
Nano Banana 2: Photo editing
I have had an ongoing, definitely one-sided feud with Gemini over its inability to fix this photo I shot last year of the Freakier Friday movie poster. The dying evening sunlight hit the glass, covering the sign at just the right angle to obscure a good chunk of the poster. Nano Banana 1 outright failed at this task. The vamped-up pro model didn’t fare much better. So, I had to see if Nano Banana 2 was up to the challenge. Spoiler: It wasn’t.
Look, clearly, this was just a bad photograph by me, human error all around. Reflections are notoriously hard to remove in editing. But one of the main selling points of creative AI is that it can take your bad photos and fix those human errors with technology. It clearly can’t do that in the worst-case scenario, which is disappointing.
In a not worst-case scenario, though, Nano Banana 2 does a decent job of removing smaller, less destructive reflections in CNET senior social media manager Allyza Umali’s glasses.
Nano Banana 2 fared similarly to the pro model for image editing. The most obvious advantage the second-generation model has is speed — the pro model takes 2-5 minutes to create an image, whereas creating images in fast mode with the base model creates results in less than a minute.
What you trade for speed is precision. I asked Nano Banana 2 to take a family photo and place us on the football field at Appalachian State University here in North Carolina. It did a decent job, but the pro model refined the details and included additional legible text like the stadium name without being asked.
For bigger edits, like changing the style of an image, Nano Banana 2 and the pro model performed similarly. I couldn’t notice big differences when I asked it to emulate the Pointillism style, though zooming in, you can see more fine dots in the pro’s output.
Nano Banana is known for its photorealism, so that makes it a top choice for those looking for a little AI-powered editing fuel. It’s certainly the best of the available options. But sometimes you just gotta accept that some errors are unfixable, and that maybe our photos are better off that way.
Should I use Nano Banana 2?
The Nano Banana models are chart-toppers for good reason. They’re adept at a variety of styles and excel where other tools fail, like with photorealism and including text. It’s not perfect by any means, but if you’re looking for an option, it’s certainly worth exploring.
Google pulled the original Nano Banana model from its options when it released Nano Banana 2, so you only have to pick between that and the pro model. If you need quick ideation and don’t care about fine details, stick with Nano Banana 2. The pro model has more power for more precise work, but be prepared to wait a bit longer for your generations.
The surge of AI image tools, with Nano Banana leading the pack, has amplified concerns from artists about the role AI plays in creative work. While some creators are embracing AI, others are more hesitant, calling out legal, ethical and environmental concerns about the tech.
Read the full article here