If a company releases new phone models but doesn’t change the cameras, would anyone pay attention? Fortunately that’s not the case with Google’s new Pixel 10, Pixel 10 Pro and Pixel 10 Pro Fold phones, which make a few advancements in the hardware — hello, telephoto camera on the base-level Pixel for the first time — and also in the software that runs it all, with generative AI playing an even bigger role than it has before.

«This is the first year where not only are we able to achieve some image quality superlatives,» Isaac Reynolds, group product manager for the Pixel cameras, told CNET, «but we’re actually able to make you a better photographer, because generative AI and large models can do things and understand levels of context that no technology before could achieve.»

Modern smartphone cameras must be more than glass and sensors, because they have to compensate for the physical limitations of those same glass and sensors. You can’t expect a tiny phone camera to perform as well as a large glass lens on a traditional camera, and yet the photos coming out of the Pixel 10 models surpass their optical abilities. In a call that covered a lot of photographic ground, Reynolds shared with me details about new features as well as issues of how we can trust images when AI — in Google’s own tools, even — is so prevalent.

Pro Res Zoom adds generative AI to reach 100x

The new Pro Res Zoom feature is likely to get the most attention because it strives for something exceptionally difficult in smartphones: long-range zoom that isn’t a fuzzy mess of pixels.

You see this all the time: Someone on their phone spreads two fingers against the screen to make a distant object larger in the frame. Photographers die a little each time that happens because, by not sticking to the main zoom levels — 1x, 2x, 5x and so on — the person is relying on digital zoom; the camera app is making pixels larger and then using software to try to clean up the result. Digital zoom is certainly better than it once was, but each time it’s used, the person sacrifices image quality for more zoom in the moment.

Google’s Super Res Zoom feature, introduced with the Pixel 3, interpolates and sharpens the image up to 30x zoom level on the Pixel 10 Pros (and up to 20x zoom on the Pixel 10 and Pixel 10 Pro Fold). The new Pro Res Zoom on the Pixel 10 Pro pushes way beyond that to 100x zoom — with a significant lift from AI.

Past 30x, Pro Res Zoom uses generative AI to refine and rebuild areas of the image based on the underlying pixels captured by the camera sensor. It’s similar to the technology that Magic Editor uses when you move an object to another area in the image, or type a prompt to add things that weren’t there in the first place. Only in this case, the Pixel Camera app creates a generative AI version of what you captured to give the image crisp lines and features. All the processing is performed on-device.

Reynolds explained that one of the factors driving the creation of Pro Res Zoom was the environments where people are taking photos. «They’re taking pictures in the same levels of low light — dinners did not get darker since we launched Night Sight,» he said. «But what is changing is how much people zoom, [and] because the tech is getting so much better, we took this opportunity to reset and refocus the program on incredible zoom quality.»

Pro Res Zoom works best on static scenes such as buildings, skylines, foliage and the like — things that don’t move. It won’t try to reconstruct faces or people, since generative AI can often make them stand out more as being artificially manipulated. The generated image is saved alongside the image captured by the camera sensor so you can choose which one looks best.

What about consistency and accuracy of the AI processing? Generative AI images are built out of pixel noise that is quickly refined based on the input driving them. Visual artifacts have often gone hand-in-six-fingered-hand with generated imagery.

But that’s a different kind of generative AI, says Reynolds. «When I think of Gen AI in this application, I think of something where the team has spent a couple of years getting it really tuned for exactly our use case, which is image enhancement, image to image.»

Initially, people inside Google were worried about artifacts, but the result is that «every image you see should be truly authentic to the real photo,» he said.

Auto Best Take

This new feature seems like a natural evolution — and by «natural,» I mean «processor speeds have improved enough to make it happen.» The Best Take feature was introduced with the Pixel 8, letting you capture several shots of a person or group of people, and have the phone merge them into one photo where everyone’s expressions look good. CNET’s Patrick Holland wrote in his review of the Pixel 8, «It’s the start of a path where our photography can be even more curated and polished, even if the photos we take don’t start out that way.»

That path has led to Auto Best Take, which does it automatically — and not just grabbing a handful of images to work with. Says Reynolds, «[It] can analyze… I think we’re up to 150 individual frames within just a few seconds, and pick the right five or six that are most likely to yield you the perfect photo. And then it runs Best Take.»

From the photographer’s point of view, the phone is doing all the work, though, as with Pro Res Zoom, you can also view the handful of shots that went into the final merged image if you’re not happy with the result. The shots are full-resolution and fully processed as if you’d snapped them individually.

«What’s interesting about this is you might actually find in your testing that Auto Best Take doesn’t trigger very often, and there’s a very particular reason for that,» said Reynolds. «Once the camera gets to look at 150 items, it’s probably going to find one where everybody was looking at the camera, because if there’s even one, it’ll pick it up.»

Improved Portrait mode and Real Tone

Another improvement enabled by the Pixel 10 Pro’s Tensor G5 processor is a new high-resolution Portrait mode. To take advantage of the wide camera’s 50-megapixel resolution, Reynolds said the Pixel team rebuilt the Portrait mode model so it creates a higher quality soft-background depth effect, particularly around a subject’s hair.

Real Tone, the technology for more accurately representing skin tones, is also incrementally better. As Reynolds explained, Real Tone has progressed from establishing color balances for people versus the other areas of a frame to individual color balances for each person in the image.

«That’s not just going to mean better consistency shot to shot, it means better consistency scene to scene,» he said, «because your color, your [skin] tone, won’t depend so strongly on the other things that happened in the image.»

He also mentioned that a core component of Real Tone has been the ability to scale up image quality testing methods and data collection in the process of bringing the feature’s algorithms to market.

«What standards are we setting for diversity and equity, inclusion across the entire feature set?» he said. «Real Tone is primarily a mission and a process.»

Instant View feature in the Pixel 10 Fold

One other significant photo hardware improvement has nothing to do with the cameras. On the Pixel 10 Pro Fold, the Pixel Camera app takes advantage of the large internal screen by showing the previous photo you captured on the left side of the display. Instead of straining to see details in a tiny thumbnail in the corner of the app, Instant View gives a full-size shot, which is especially helpful when you’re taking multiple photos of a person or subject.

Camera Coach

So far, these new Pixel 10 camera features are incorporated into the moment you capture a photo, but Reynolds also wants to use the phones’ cameras to encourage people to become better photographers. Camera Coach is an assistant that you can invoke when you’re stuck or looking for new ideas while photographing a scene.

It can look at the picture you’re trying to take and help you improve it using suggestions such as getting closer to a subject for better framing or moving the camera lower for a more dramatic angle. When you tap a Get Inspired button, the Pixel Camera app looks at the scene and makes suggestions.

«Whether you’re a beginner and you just need step-by-step instructions to learn how to do it,» said Reynolds, «or you’re someone like me who needs a little more push on the creativity when sometimes I’m busy or stressed, it helps me think creatively.»

CP2A content credentials

All of this AI being worked into the photographic process, from Pro Res Zoom to Auto Best Take, invariably brings up the unresolved question of whether the images we’re creating are genuine. And in a world that is now awash in AI-generated images that look real enough, people are naturally guarded about the provenance of digital images.

For Google, one answer is to label everything. Each image captured by the Pixel 10 cameras or touches Google Photos is tagged with C2PA Content Credentials (Coalition for Content Provenance and Authenticity), even if it’s untouched by AI. It’s the first smartphone with C2PA built in.

«We really wanted to make a big difference in transparency and credibility and teaching people what to expect from AI,» said Reynolds. «The reason we are so committed to saving this metadata in every Pixel camera picture is so people can start to be suspicious of pictures without any information.»

Marking images that have no AI editing is meant to instill trust in them. «The image with an AI label is less malicious than an image without one,» said Reynolds. «When you send a picture of someone, they can look at the C2PA in that picture. So we’re trying to build this whole network that customers can start to expect to have this information about where a photo came from.»

What’s new in the Pixel 10 camera hardware

Scanning the specs of the Pixel 10 cameras, listed below, you’d rightly notice that they match those found on last year’s Pixel 9 models, but a couple of details stand out.

For one, having a dedicated telephoto camera is no longer one of the features that separates the entry-level Pixel from the pro models. The Pixel 10 now has its own 10.8 megapixel, f/3.1 telephoto camera with optical image stabilization that offers a 5x optical zoom and up to 20x Super Res Zoom.

It’s not as good as the 48-megapixel f/2.8 telephoto camera used in the Pixel 10 Pro and Pixel 10 Pro XL (the same one used in the Pixel 9 Pros), but that’s not the point. You don’t need to give up extra zoom just to buy a more affordable phone.

Another difference you’ll encounter, particularly when recording video, is improved image stabilization. The optical image stabilization is upgraded in all three phones, but the stabilization in the Pixel 10 Pros is significantly improved. Although the sensor and lens share the same specs as the Pixel 9 Pro, the wide-angle camera in the Pixel 10 Pro models necessitated a new design to accommodate new OIS components inside the module enclosure. Google says it doubled the range of motion so the lens physically moves through a wider arc to compensate for motion. Alongside that, the stabilization software has been tuned to make it smoother.