I tried Google’s new Nano Banana Pro, and it’s the AI Photoshop of my dreams

In its never-ending quest to defeat OpenAI, Google is trying all sorts of tactics to get people to switch from ChatGPT to Gemini. One of its best weapons in this fight is Nano Banana, Google’s AI image generation and editing model. It went insanely viral after launch, so Google has leaned hard on it, promoting it on as many platforms as possible. To maintain this momentum, Google has launched Nano Banana Pro, a significantly improved version of its AI image model. I had the opportunity to test this new model prior to launch, and I think it’s going to become a Photoshop replacement for image editing beginners like myself.

I’ve been using the original Nano Banana for a variety of things since it launched in late August. From modifying our existing stock images to seamlessly add new text, to generating eye-catching thumbnails for our Authority Insights Podcast, Nano Banana has been an immensely useful tool for my work. There’s no doubt that it does an incredible job at creating and modifying images — it wouldn’t have gone viral if it was terrible, after all.

But what you may not know without trying Nano Banana yourself is that it can sometimes be incredibly frustrating to work with. It often misses important details or misinterprets your prompt entirely, forcing you to start over. Asking it to make changes in a follow-up is often a waste of time because it tends to repeat the same edit, as if it’s banging its digital head against a wall. And if you need it to generate an image that isn’t a 1:1 square? Good luck.

Nano Banana Pro addresses nearly all these issues. It does a much better job at understanding your initial prompt, can actually handle requests for additional edits, and can finally create images in 16:9 and other aspect ratios. This is because it’s built on the new Gemini 3 Pro model, which is capable of reasoning. In contrast, the original Nano Banana is built on Gemini 2.5 Flash (which is why its official name is Gemini 2.5 Flash Image). Building it on top of a reasoning model makes a world of difference in how it understands prompts, transforming Nano Banana from a nifty but sometimes annoying tool into a genuine AI image assistant.

Here are several examples of images I created using Nano Banana Pro so you can get a feel for what it’s like and why I’m calling it an AI Photoshop for beginners.

Making memes: How Nano Banana Pro lets beginners join in on the fun

Now, I know what some of you are thinking: AI can’t make memes! You’re right; memes are made by people. But part of what defines a meme is the act of sharing it widely and, often, remixing it to fit a specific subculture or reference. While meme generator templates make simple edits easy, other memes require more effort to adapt. You need a certain level of Photoshop proficiency to mold images to fit specific formats — skills that many people, myself included, lack. So, we sit on the sidelines, admiring the high-effort edits others create, wishing we could participate without sinking dozens of hours into learning Photoshop just to make a joke.

Thanks to Nano Banana Pro, I can now create these kinds of memes in moments. Here are two examples showing the Android Bot mascot adapted to the “Nah, I’d win” and “Domain Expansion” scenes from the anime series Jujutsu Kaisen.

I can’t imagine how many hours it would take me to make either of these manually, yet Nano Banana Pro did them in seconds. What impressed me most, however, was the ability to go back and forth with Gemini to refine the first image. This would have been impossible with the original Nano Banana, where I would’ve had to start a dozen new chats, tweaking the initial prompt each time, just to achieve the same result.

Nano Banana Pro can finally edit images without butchering faces

One of the major issues with the original Nano Banana, at least in my experience, was how it sometimes distorted faces during edits. I once asked it to remove some objects and people from the background of a group photo I took with friends. It cleared the background well enough, but the result was inexplicable: it changed all our faces, making us look like entirely different people!

I was worried that Nano Banana Pro would suffer from the same flaw and that I would have to be extra careful with my prompting. Fortunately, I’m happy to report that it can handle image edits without mangling faces.

Both of these edits are impressive. Gemini did a fantastic job rendering “Nano Banana” onto the gingerbread cookie, making the text look as if it were truly piped in frosting. It also seamlessly swapped the Galaxy XR headset for an Apple Vision Pro, even going so far as to recreate part of my nose and blend a reflection of my friend’s apartment onto the visor. Granted, if you look closely, the reflection doesn’t exactly match the real room, but you likely wouldn’t notice without seeing the original photo.

Why Nano Banana Pro is more than an AI Photoshop

While I previously compared Nano Banana Pro to Photoshop, I believe it goes beyond that comparison. Photoshop can already use AI to add, remove, or expand content, and it can generate images from basic text prompts or references, just like Nano Banana.

However, unlike Photoshop, Nano Banana Pro can reason. It isn’t limited to simple text prompts or reference images. Because it is built on Gemini 3 Pro, Nano Banana Pro can analyze complex data before generating an image. For example, I shared my current workout routine and asked it to create a graphic visualizing the muscles I’m training.

I also provided a list of Android betas and their release dates, asking for a timeline. In response, it wrote and executed a Python script to plot the data.

Are these graphics 100% perfect? No; they require some additional edits to be fully accurate. However, the fact that Nano Banana Pro can interpret and visualize raw data opens up exciting possibilities.

Nano Banana finally supports more aspect ratios

The original Nano Banana was notoriously stubborn, often refusing to generate anything other than 1:1 square images. I had to trick it by uploading a blank white image with my desired dimensions, forcing it to fill that canvas. Nano Banana Pro solves this by natively supporting aspect ratios like 16:9, 2:1, and more, giving you far more flexibility in where you can use your generated images.

AI is still AI: What Nano Banana Pro struggles with

Despite its many improvements, Nano Banana Pro isn’t flawless. The resolution of generated images is still below 1080p, for one. And while text rendering is significantly better, the model still trips up occasionally, particularly when dealing with background text or words not explicitly defined in the prompt.

It also frequently fails to display the correct time on analog clocks — a persistent hurdle for AI image models. Additionally, getting the perfect image still requires a fair bit of finesse; if your prompting skills aren’t up to par, you might find the results horrifying. Finally, safety guardrails remain in place to prevent edits involving public figures. So, you can’t, for instance, plaster Gabe Newell’s face on a Steam Machine to create the “Valve GabeCube.” For that, you’ll still need to learn Photoshop.

Overall, I’m incredibly impressed with Nano Banana Pro. Thanks to its reasoning capabilities, it’s a substantial improvement over the original model. It’s better in nearly every regard and significantly less frustrating to use. As a result, I can now see myself using it for more than just work. While Nano Banana Pro still struggles with things like background text rendering, it is more than capable enough to replace Photoshop for many beginners. It certainly has for me!

What do you think of Nano Banana Pro? Let us know in the comments below!