Gemini’s ‘Nano Bananas’ AI Image Editing Is Fun, but I Ran Into Too Many Slipups

I tried image editing capacities in the latest Google generative AI model, and you can also. Named Gemini 2.5 Flash Image, it improves the possibility of modifying your photos natively in Gemini.

AI lovers have called the “Nano Bananas” model, stimulated by a series of teasers from Google frames. Google has invested massively in its generative media models this year, abandoning the updated versions of its image and video generator at its annual conference on E / S developers. The AI ​​video generator of Google Veo 3 was amazed with a synchronized audio, a first among the Giants of the AI. And the creators have produced more than 100 million AI videos with the AI ​​filmmaker tool from Google, Flow.

I spent a lot of time testing the creative software of AI, and I was delighted to see the last drop in Google in the middle of the whole banana fanfare. But my 2.5 Flash Image test has shown that it is not because something has a flashy entry that it will always be up to its overhaul. Here’s how my experience with Gemini Nano Bananas went: good, evil and frustrating.

What worked

The Gemini Bananas model is frightening to add elements with existing images, mixing elements generated by the right AI in any image you have taken. It also maintains a decently stable level of coherence – which means that people in my photos were not too distorted or banal after having followed the processing of AI. These are both important distinctions for AI image programs, and something Google said it had worked to improve.

You can see these two characteristics on this photo of my sister and me. Our general appearances are unchanged in the published version (right), showing this consistency of the characters. I asked Gemini to add a third sister who looked like both of us, which rarely did well by adding a third woman between the two of us.

I was also quite impressed by the speed with which the Gemini could spit finished images. Everywhere in less than a minute gets a gold star on my part, and Gemini regularly manages requests in less than 15 seconds. I also appreciated the way he added a watermark to all the images he created and edited – even if I don’t like the way technological companies have corrupted Emoji Sparkles for AI, it is extremely important to have content markers generated by AI. The synthetic and behind the scenes of Google also helps to differentiate the content of the AI ​​from the images created by man.

Gemini is also good in the creation of Images AI roughly, but I recommend using its image generator Imagen 4 or another IME – they have more controls and practical parameters that bring you closer to what you want with less work.

What didn’t really work

There are serious limitations to gemini bananas. He automatically generated square images and follow -up prompts asking that images be adapted to other dimensions have been ignored or stranded.

I also noticed that Gemini reduced the resolution of several of my photos. I mainly take photos with my iPhone 16, which has stellar cameras, but after having traveled the gemini banana model, these good details were often blurred. It’s boring and will not win any photographer.

I tried several times to ensure that Gemini managed the photo changes that would have been difficult to make manually. It is an area of ​​photo retouching where AI is supposed to excel – automate banal modifications but with a high intensity of details. Unfortunately, the Gemini really had trouble with rapid membership here, which means that it did not do what I asked.

I tried several times so that Gemini withdraws the reflections from a smaller film snap in the Friday, but they remained obstinately. And the more I tried to have it removed the reflections, the more the image quality has become poor with each prompt. The once clear text was finally illegible after finally abandoning, not to mention accidental and frightening damage caused to the faces of Lindsey Lohan and Jamie Lee Curtis.

Gemini Nano Bananas had trouble generating images in different dimensions. Rejection and cropping images are a basic photo editing process, but Gemini do not – or could not – manage simple dimensioning guidelines in my invites.

I contacted Google about the problems of resolution and dimension and a spokesman said that the technology company was “aware and actively worked on the two questions. It was a big update of our previous model, but we will continue to improve the model”.

Overall, Gemini Nano Bananas proved to me that Google is serious to continue to dominate in the generative media. But it has significant traps, with too much concentration on the generation of new elements rather than the use of AI to improve and modify current photo problems. For the moment, the Nano Bananas model is best suited to Gemini fans who want to make major changes quickly. For those of us who are looking for more specific tools, we will have to wait for the next big update of Google or find another program.

Gemini nano bananas availability, pricing and privacy

You have nothing to do to access the new model; It is automatically added to the basic gemini 2.5 Flash 2.5 model. GEMINI is available for free, with more models and higher user ceilings available in Google AI plans from $ 20 per month.

If you are a paid subscriber, you can also access the model via Google AI Studio. From there, all you have to do is download an image and type your prompt. Each invite uses one to two thousand tokens, depending on the level of detail required. Adobe Express and Firefly users can also access the new model now.

Google Gemini Privacy Policy indicates that it can use the information you download to improve its AI products, which is why the company recommends that it is to download sensitive or private information. The policy of prohibitive use of the company of the company also prohibits the creation of illegal or abusive equipment.

To find out more, see the best IA images generators and everything announced during the Made by Google Pixel 10 event.