Nano Banana Is Fantastic, But Don't Bury Photoshop Just Yet
As great as Nano Banana is at editing images, it won't replace professionals.
It’s time for another Thursday “Hot Take.”
TL;DR
Gemini 2.5 Flash Image (aka “Nano Banana”) is now the world’s best image model that also excels at detail-preserving edits, but let’s not equate “good enough for the average Joe” with “Photoshop extinction event.”
What is it?
Gemini 2.5 Flash Image is Google's latest image model, which made quite a splash last week under its pre-release nickname, "Nano Banana."1 Two days ago, Google officially claimed ownership of the model.
While Nano Banana is topping leaderboards for text-to-image generation, what truly makes it special is its ability to make precise edits to existing images while keeping characters consistent and preserving details.
Here's a quick taste:
It can blend separate characters and insert them into novel settings:
And it can handle multi-turn edits without losing key details of the original image:
In short, Nano Banana is pretty damn neat.
How do you use it?
If you want to try Nano Banana for yourself, you've got three options...and all of them are free!
Option 1: Gemini app
Go to gemini.google.com or open the Gemini app on your phone
Start a new chat.
Upload the image(s) you want to edit.
Request your changes by simply describing them.
That's it!
If you just want to create new images, select "Create images" in the Tools menu, then describe the image you want:
Like so:
Option 2: Google AI Studio
Go to aistudio.google.com and sign in with your Google account
Select Chat in the left-hand column
Select "Gemini 2.5 Flash Image Preview" from the right-hand model picker
Now you can upload images to edit or request new ones as above
Like so:
Option 3: LM Arena
Head to lmarena.ai
Select "Direct Chat" in the top dropdown
Click on "Generate Images" under the prompt box
Select "gemini-2.5-flash-image-preview" in the top dropdown
Now you can upload images to edit or request new ones as above.
Important note: Unlike Gemini and Google AI Studio, LM Arena reserves the right to make your prompts public for research purposes.
Why should you care?
Character consistency has long been the holy grail of AI image and video generation. It allows for long-form storytelling, precise image remixing, different shots of the same setting, and more.
Right now, no other AI model is as good at keeping characters and image details consistent as Nano Banana.
The previous best image model, GPT-4o image generation, struggled with this. In fact, it was so bad at preserving details and characters over multiple turns that it spawned a short-lived-but-fun trend: People would upload an image and repeatedly ask ChatGPT to "Create a replica of this image. Don’t change anything." The results were disturbing:
So yes, Nano Banana is the closest the average user can get to making nuanced image edits in natural language.
But in its typical fashion, the Internet jumped from that to "Photoshop is dead" in a split second.
Seriously, go type in “nano banana photoshop” into your search bar.
I’ll wait.
No? Fine, I’ll do it for you:
Why must we constantly pair the arrival of one awesome tool with the death of something else? Is it the drama?
We already killed Google back in early 2023:
Now, Google is back from the grave to murder Adobe?
Will this cycle of violence ever end?!
Jokes aside, there's quite a gap between "This AI tool lets people make quick and precise image edits" and "Professional Photoshop workflows are now obsolete.”
For one, Nano Banana still has many limitations. Google itself highlights that the model struggles “with small faces, accurate spelling, and fine details in images.”
When I threw my “long text” prompt from this image model showdown at Nano Banana, it couldn’t nail the show program text after multiple tries—something that GPT-4o image generation did quite consistently:
This AI Search video, while very positive about Nano Banana, has a whole section dedicated to tests where other models do better:
But also—and I know this is a very controversial statement—there’s probably a tiny difference between making quick AI-powered edits and doing professional-grade visual design work.
How about we enjoy playing with this new, super cool tool without instantly proclaiming the death of all professional design?
Have fun out there!
🫵 Over to you…
Have you tried editing images with Nano Banana? What’s been your experience? Have you run into other obvious limitations? I’d love to hear your thoughts!
Leave a comment or drop me a line at whytryai@substack.com.
Thanks for reading!
If you enjoy my writing, here’s how you can help:
❤️Like this post if it resonates with you.
🔄Share it to help others discover this newsletter.
🗣️Comment below—I love hearing your opinions.
Why Try AI is a passion project, and I’m grateful to those who help keep it going. If you’d like to support my work and unlock cool perks, consider a paid subscription:
I’ll mostly stick to the reader-friendly “Nano Banana” name in this post.
New thing to play with YAAAAY! Must be the drama; we need AI Deathmatch pitting software battles and AI oblivion