Here’s how Zoom Ai Pixel compares to a real 100x lens

If you missed it last week among other huge messages, Google sent a telephone camera with an enlargement function that uses generative artificial intelligence. That’s right: Pixel 10 Pro is equipped with artificial intelligence in the camera application that cleans the shit images of the digital zoom up to 100x. This is a nightmare of co-a-photo, but it is also quite good-at least it seems to be. But it is challenging to be completely sure what you should look like when you photograph when it is much mile. So I brought a bell to comparisons side by side: Nikon Coolpix P1100.

For those unknown, the P1100 is a massive ultra-Tomni camera with an equivalent range of 24-3000 mm. When you have such optics, you don’t have to do any height, just like Pixel 10 Pro. The camera uses some reduction of noise, sharpening and color adjustments. But it doesn’t have to be completely guess What each individual pixel should look like, because he had some information at the beginning.

The digital zoom, like Pixel 10 Pro, is a different story. Upscaling image 10 or 20 or 100 times without the benefit of optical enlargement leaves many gaps to fill. Algorithms can spend quite good guessing, but they are simply: guess. Pixel 10 Pro Pro Zoom makes it guessing with generative artificial intelligence. And if we take photos of the AI ​​zoom, where is it better to start with the moon?

He asks a lot of camera for smartphones to take a photo of the moon, and Google is not the first manufacturer of the phone that led AI to fight. The Pro Res Zoom version certainly looks similar to the moon, but AI gives it a strange spongy texture, which does not look quite good-especially comparing it with the P1100 version.

Pictures above Lumen fieldThe outer part was downloaded with a view in the Seattle center near Pike Place Market around Mili. It was a foggy, shadowy day, so I apologize for gloomy images, but they give a better idea of ​​where the pro res zoom stands out and where it falls. The AI ​​model makes the numbers on the signs readable and pristine really well, but basically removes metal cladding on the side of the building, such as too aggressive noise reduction. And once again, AI does not know what to do with writing.

These photos of the headquarters of the main Starbucks, Mili south of Lumen, were taken from the same point of view. On the tiny screen, the AI ​​version seems fine, but if you look carefully, you can see where it turned the lamps into windows and gave the clock on the tower some Salvador Dalí treatment.

On a more bright day, I pointed out both cameras at a different landmark in Seattle. I was about three miles from the space needle and met another enemy of distant -range photographs: Haze Haze. And she didn’t quite know what to do with distorted lines and created Tim Burton’s the Space Needle Instead. But you can see that the P1100 could not do much better, which with the entire heated atmosphere between the lens and the object.

Haze heat is clearly a problem in this situation. I did not stand too far from the planes in the Boeing field in the above pictures, but between me and the planes there were a lot of heated asphalt that I photographed. But apparently AI is shining. In fact, it can be yours Just Option if you try to improve something as challenging as Haze Haze.

Everything will get complicated here. Generative AI has been in the tools for editing photos for years and is extremely useful when removing noise from a photo taken with an venerable digital SLR. Haze toasty is an even more unpleasant problem; Random distortions and waves are almost impossible to correct with conventional digital tools for editing photos. Landscape and wildlife photographers are It already includes tools for editing artificial intelligence This can do things that your ordinary Lightroom sliders can dream about.

Is it different when AI is in the camera application, not only in a professional image editor that you would operate after the fact? Absolutely. Will the pro res zoom understand it very much? Also yes. But it was an illuminating exercise and I don’t think we will hear it recently about the generative artificial intelligence used in the image capture tool itself.