Comparing Nano Banana (aka Gemini) with ChatGPT images

This page has 0 threads | Add post

Blogs

Many people around the world enjoy and benefit from our blogs. If you're one of them - and can afford to - please make a small donation to help keep this and future blogs, exercises, skills assessment tests, tips, tutorials and videos freely available to everyone!

Some other pages relevant to this blog include:

Artificial Intelligence training classes

You can also book hourly online consultancy for your time zone with one of our 7 expert trainers!

Comparing Nano Banana (aka Gemini) with ChatGPT images
There's been a lot of talk recently about how Nano Banana (the image generation component of Google Gemini) is racing ahead of the competition. Is it?

Posted by Andy Brown on 29 January 2026

In this blog

The tests
The first test - still life with potato
The second test - travel bingo card
Test 3 - creativity
Conclusions and recommendation

I've read a lot recently about Nano Banana, which it turns out is just the image generation component of Google Gemini (you can see why it's called Nano Banana here, if you're interested):

What you see when you enter a prompt in Gemini to create an image.

I've compared the main image generation tools recently, but things are changing so fast with AI that I thought it was time to revisit the subject.

To make sure that I'm comparing like for like, and seeing results which are as good as they can be, I'm using the paid version of both AI tools (ChatGPT Plus 5.2 and Gemini 3 Pro). The cost for both is suspiciously simiilar at about $20 a month.

The tests

I'm writing this before doing the tests, because I don't want to influence what the results will be. I'm going to try 3 prompts:

Testing	Prompt
Quality of picture	"Create a photo-realistic image of a bowl of fruit also including one potato, with subtle lighting and shadows. The bowl of fruit should be the only thing in view."
Displaying text / accuracy	"Create a picture for children to use to play picture bingo during a long car journey in the UK. The picture should be titled "Wise Owl Bingo" and have 12 pictures, with each having a text description of the picture underneath. Choose 2 things which are ridiculously easy to see, 10 normal ones of different levels of difficulty and 2 which are humorously impossible (one because it's so unlikely you would see it and one because it's made up)."
Following instructions / creativity	Please create a picture of a cute creature browsing through books in a book shop, while the customers look at it in amusement and surprise.

In each case I'll show the results, but also time them (to compare image generation speeds). I also want to see how easy it is to tweak the images generated.

The first test - still life with potato

Here's what each AI tool produced:

Tool	Seconds	Image
Gemini	17
ChatGPT	38

I'm no photo expert, but I think ChatGPT has the edge here just - its photo is more ambitious, and has more subtlety of shading and colour.

The second test - travel bingo card

This was designed to test two things: how well the tools can cope with text, and how good they are at following instructions. Here are the results:

Tool	Seconds	Image
Gemini	13
ChatGPT	80

Gemini's effort isn't bad at all, especially given how quickly it appeared, although I did ask for 12 pictures not 9 (and the police car has a unicorn on top of it). But the pictures and concepts on the ChatGPT bingo card are better, even if there are 4 impossible things to see rather than the 2 I asked for. I thought I'd give each of the tools another chance, and got this:

Tool	Additional prompt	Results
Gemini	I asked for 12 pictures not 9, and the police car has a unicorn on top of it. Please try again.
ChatGPT	I asked for only 2 which are humorously impossible, but you've given me 4. Please try again.

Neither tool has covered itself in glory! Gemini still hasn't given me 12 pictures, and ChatGPT has still given me 4 impossible things to see (with the Loch Ness Monster now split across two frames).

Test 3 - creativity

The prompt for my final test was:

Please create a picture of a cute creature browsing through books in a book shop, while the customers look at it in amusement and surprise.

Here's what this produced:

AI tool	Seconds	Image
Gemini	15
ChatGPT	38

This probably comes down to personal preference. I much prefer the Gemini image, which is more interesting and appealing (and I think closer to what I requested). If you look closely the ChatGPT image also has a boy's head on the bookshelf (a typical AI anomaly).

Conclusions and recommendation

I draw 3 clear conclusions - 2 from the tests above, and one from my own use of the two tools:

Conclusion	Notes
Gemini is much quicker	ChatGPT typically takes more than twice as long as Gemini to create the same image. Personally this doesn't bother me as I set images going in different tabs of my web browser, and then do something else while they're created.
Both produce superb images	Both from the tests above and from my own experience, there's not much to choose between Gemini and ChatGPT for image creation; they both produce excellent pictures.
Gemini doesn't make changes well	I've continually found when using the two tools for my own purposes that it's hard to get Gemini to make changes to any picture that it's generated (it's usually easier just to start again). All AI image generation tools suffer from this problem, but Gemini more than most, I've found.

Because I don't care too much about generation speed, the 3rd conclusion means that I'll be sticking with ChatGPT as my image generation tool for now.

I used a combination of the two tools to create the images in our latest newsletter picture quiz, and continually found myself getting frustrated with Gemini's refusal to make the changes I requested to an image. In the end I used ChatGPT to generate the majority of the 20+ images in this quiz.

Some other pages relevant to this blog include:

Training courses in AI (Artificial Intelligence)

You can also book hourly online consultancy for your time zone with one of our 7 expert trainers!

This blog has 0 threads

Add a new post

ALL BLOGS

ARTIFICIAL INTELLIGENCE BLOGS

ARTIFICIAL INTELLIGENCE (AI) BLOGS

The tests

The first test - still life with potato

The second test - travel bingo card

Test 3 - creativity

Conclusions and recommendation

Head office

London

Manchester