Google Gemini Just Got a Lot Better at ‘Photoshopping,’ and I’m Worried

Did you know that you can customize Google to filter the garbage? Take these steps For best search results, including the addition of Lifehacker as a favorite source for new technological.

Google has updated its Gemini application (and its website) to make the generation of images a little more intuitive, and for once, which I previously announced as novelty could now be a viable photoshop alternative. There is still typical AI waste, but the new model, tested under the name “Nano banana“And now live for all Gemini users as a Flash Gemini 2.5 image, does a lot to allow you to adapt an image to your liking. Everything always has a watermark and” AI “warnings in metadata, but prepare to be much more exciting about the question of whether a photo is real or not – the new Gememini blurs.

Google Gemini is now better to edit real photos

What makes the model updated so special is the focus on maintaining details on several photos. Now, instead of generating essentially from zero each time you ask the Gemini application a photo, it can take back parts of a source photo or an image previously generated and change only what you ask for. There are two major reasons why it counts, and ironically, one of them really means using fewer AI.

For example, let’s say you have a photo of you wearing a red shirt, but you want it to be blue. Previously, you had two options: you had to either take the image in Photoshop yourself and modify it manually, or use it as an IA prompt and continue to generate until you look close to the original photo, but now with the blue shirt. With the changes in Nano Banana, Google has refined its model so that it leaves most of your image alone and only changes the shirt.

Credit: Michelle Ehrhardt, Google

For example, here is this exact situation, with some photos of me. Note how the model keeps beautiful details like the frizz of my hair or my specific facial expression and my pose. It is not perfect, and you will notice that my skin looks a little more fluid in the published version, but with the new updates, Gemini is now able to determine what I mean by “shirt” and to concentrate most of its modifications on this subject. I would say that the shirt also looks a little unnatural, especially around my right shoulder, but I have not given many Gemini to work in my invite. This is where the next big change comes into play.

Use gemini to modify the same result several times

This is where the real thing is. Whether it is entirely generated by AI or not, you can now use images generated previously as a basis for future generations. In other words, if the Gemini did not have something good the first time, you can ask him to try again until this is the case.

To give you an idea of ​​what it looks like, here is the same photo of me in the blue shirt, but now with added peas, to better match the red shirt of the original photo.

Credit: Michelle Ehrhardt, Google

And here is an image entirely generated by the AI ​​of a cat, which I had gemini in orange.

Credit: Google

It’s huge for the generation of IA images. Previously, when you ask Gemini to make small adjustments to the content it is already generated, you will essentially get new photos each time, as is the case with these dogs wearing hats.

Credit: Google

Now, however, you can ensure that the itere application in the same photo several times, which means that if the initial result does not seem convincing, you have the possibility of repairing it. For me, it takes this to be a novelty – where you must essentially run a wheel with each generation and hope that it would land on something useful – to a real Photoshop threat.

Google suggests, for example, that you could use it to see what you would look like if you lived in a different decade, or if you had a different career. I admit that the results seem convincing enough to work for occasional messages, especially if you download a real photo as a context. I am next to real life Mona Lisa, but reinvented as an artist.

Credit: Michelle Ehrhardt, Google

It is not strictly realistic (why is there a second Mona Lisa next to me?), But I could see a certain type of person doing enough of it to flood social media with messages like this. Spend a little time to iterate on it, and you could probably even give the impression that I just went to the Louvre.

But if you are an AI skeptic like me, there is always a saving grace that shows that the model has a little room to grow.

The combination of photos is not yet well

Although the new Gemini updates make the iteration of existing photos much more viable, asking him to generate new content, where he cannot count too much on a source photo, always gives you a notable radiance. One of the additional features that Google announced with this update was the possibility of using Gemini to combine several source photos in one. But while other changes mainly involve making small adjustments to existing photos, it always requires that AI consists a lot to assemble the photos, and this is where you are most likely to run the same old problems.

Credit: Michelle Ehrhardt, Google

For example, by following one of Google’s suggested examples, I downloaded a photo of myself and my cat to Gemini, and I asked him to take a photo of us cuddling together. But while the other tests I did with this update looked a lot like the source photos, the result here gave me a version of myself in a shirt too tight, with too bright hair, huddling a cat that is too chunky. The large lines were right – my face always looks like me, my cat’s fur motif is almost intact, and the sofa even has the right color and good general shape. But in addition to a few small inconsistencies with it, let’s say, the folds on the sofa, or my dimples, or the lamp in the background (which seems to have two posts), whoever met my cat knows that it is not so large. The photo also has this vaseline look and too treated which is endemic to AI.

To a certain extent, it is to be expected. I have not downloaded too many photos, and certainly none of me or my cat in the poses presented in the Image AI. The AI ​​had no way of knowing what we would look like from different angles, especially since my selfie was only a whim. But what I obtained means that when AI lacks useful source information and must intudate what a scene should look like, it always behaves in familiar problems that facilitate the distinction of photos made without AI. I could probably make the photo of the AI ​​more realistic if I downloaded pictures of source closer to what Gemini wanted to generate, of course, but I must then wonder what would be the interest of AI in the publishing process?

In any case, I can say with confidence that the manufacture of advanced modifications of IA looks convincing will always take a lot of human intervention.

Prepare for a mixture of AI and reality

The new Gemini updates are, for me, the most impressive when used for smaller adjustments, where I think the threat to Photoshop enters.

What does that mean? Well, for its part, this means that free AI tools are finally to the point where you could be able to use them to do with a natural language prompt which could have taken a few minutes in hand before. Adobe has already said that she was planning to Incorporate Nano Banana In Photoshop, but prepare for new changes in traditionally untouchable applications as IA progresses. This is to the point where, at least for little things, it can really threaten your traditional workflow.

For people who are not content creators, expect to develop an even more demanding eye on what is and is not real online. Although the images entirely in terms of AI are often quite easy to spot, and more realistic changes can be mainly harmless (nobody cares about the color of my shirt), gemini updates now facilitate that never mix reality with a little lie. Here is an image that I made the new Gemini of Taylor Swift in a red baseball cap, if you catch my drift.

Credit: Google

While we are waiting to see how it takes place, it’s the right time to remember that if an image advances your alarms, the Gemini put the AI ​​filigranes in the lower left corner of all its results, and will mark the photos generated using in their metadata, which you can see on iPhone and Android while swinging on a downloaded photo. There are ways to clean metadata, but as a withdrawal, because the most convincing changes are likely to use real photos like their sources (I made for the Taylor Swift above), as a last resort, you can also use a Google reverse image search To try to find the original unchanged. Be careful there.