Image AI: Make ChatGPT Look Obsolete with this Free App
One of the most popular applications of AI is image creation. Testing what you look like as a muppet or an anime character, or presenting yourself as an astronaut — the possibilities are endless! Now, a mysterious tool is preparing to perhaps even unseat ChatGPT from its throne. We'll tell you the name of this mysterious AI, who's behind it, and most importantly, how you can use it for free!
When I look online to see what people are doing with generative AI, it's mostly pictures. They either come up with their own scenes or want to have their own photos reworked. One problem that many of them have: They don't look like their own template, or the scene doesn't remain consistent. However, if you are thinking up a little picture story, for example, it would of course be helpful if the hero always looked the same.
Is Google the Company behind this Mystery Tool?
With Nano Banana, a new, unofficially released AI tool is now preparing to solve these problems. For a few days now, impressive images have been circulating online showing what the tool is capable of. So far, no company is willing to say who is really responsible for the app. However, it is rumored to be Google's new powerful image weapon.
Now I’m testing Nano Banana with product replacement. even with product photos that have complex patterns nano banana can still match them perfectly. On average it only takes me 2–3 tries to get a solid result. Well played
— DStudioproject (@D_studioproject) August 20, 2025
base image : @letz_ai pic.twitter.com/UpEdhMOQkF
Google itself has not commented on this, but there are a few clues: Logan Kilpatrick, head of product for Google's AI Studio, posted a banana emoji, and Naina Raisinghani, product manager at Google DeepMind, posted a picture of a banana stuck to a wall. None of this is proof, but it is at least circumstantial.
Regardless of this, Nano Banana is not yet publicly accessible. But there is a trick you can use to experiment with the tool right now. The coolest thing is, it doesn't cost you a single penny, and you can try it out indefinitely!
- Also read: Hey, Google — make the picture better!
You can already use the Brand New Image Tool here
To see for yourself what Nano Banana can do, you need to visit the LMArena page. You can experiment with numerous AI models there, including working on images. You can either select a model or let two models compete against each other. The catch? You can select an incredible number of models here, whether GPT, Flux, Grok, Gemini, or others — but Nano Banana is not yet listed.
There is, however, a "Battle" option. You can't choose which models to use there. It's a total blind flight: you name your prompt, upload one or more pictures, and receive two pictures. You can then decide which one is better, or whether they are both equally good or equally bad. After your vote, you will then be shown which models were used. And Nano Banana appears right there, time and again.
What does this mean for you? You may have to be a little patient and try more often. But that doesn't matter, as you have no limit. This is how it works:
- Go to the LMArena page.
- Select "Battle" at the top.
- Formulate your prompt (in English).
- Click on "image".
- Attach one or more images using the + icon on the left.
- Click on the button with the up arrow on the right to generate the image.
- You will now be shown two images — select your favorite, followed by the models used, which will be displayed at the top.
How Good is Nano Banana, Really?
Here is my initial assessment after a few attempts and also after having seen the images generated elsewhere. So yes, Nano Banana can really work miracles! But that doesn't mean that you won't still get nonsensical results from time to time. Sometimes, I preferred the images of Qwen (a Chinese model) or Flux better, but as a rule of thumb, Nano Banana delivers sensational results.
The highlight for me, however, is the consistency in the images. Before I go on at length, I'd better show you the results. Take a look at the following three examples:
The image is selected from the tweet embedded above and shows how the model's handbag was replaced. Simply upload both images and use the prompt "Change the bag with the second image", and you will get this result.
What I found exciting in this example is that the background - i.e., the bar — remains consistent. Incidentally, the top left image is from Flux, the other three are from Nano Banana. You can see minimal differences in the people, but I think they are consistent enough to tell a story with pictures, for example.
Here, I wanted a candy landscape with Toblerone mountains alongside popsicle and lollipop trees and wine gum fish. Then I wished for a candy cane ship on the river. Then I uploaded a portrait of myself, and wanted me to be happily jumping around with a chocolate bar in my hand. And on the last picture, I simply asked for the same picture, but at night.
In this picture, you can see, on the one hand, that I really didn't use anything except "The same picture, but now it's night" — and, on the other hand, how much worse the result of the AI looked from the likes of ChatGPT 4o. And yes, overall, what Nano Banana delivered was very impressive. What do you think?
- Also exciting to know: Party photo or Mona Lisa: This is how you can soon enter every picture
PS: The article picture above is, of course, also from Nano Banana. Let me know if you want me to share the prompt for this picture ;)
