Half of the Pixel 11's 4K60 Video Is Generated

30 Real Frames, 30 Generated

Choose 4K60 HDR with Video Boost on the Pixel 11 Pro XL, and the phone is in reality filming 30 frames per second. Google has confirmed to Forbes that the missing frames are generated in the company's cloud. The flagship video from Google's most expensive phone was already partly synthetic — before anyone pressed a single AI button.

That context is what makes the latest news more than a feature launch: In July, Google introduced Video Remix in Google Photos, a tool that, according to the company, creates videos "in seconds" from material already sitting in your archive. A photo library that for roughly fifteen years has been a passive record of what actually happened becomes, in one move, a tool where new clips can be produced from old ones.

The timing isn't conveniently awkward. It is directly strained. That same summer, on August 1, Google had to switch off a brand-new AI image feature in Google Earth after it managed to fabricate a nuclear power plant in Iran — an episode Forbes documented. And on August 29, four weeks after the Earth feature was pulled, Forbes laid out the confirmed frame generation in Video Boost. In other words: the same company is now offering generative video editing of personal memories, in a period when its generated pixels have already had to be recalled once and been revealed to be replacing real capture data once.

What Video Remix Actually Does

It is worth sticking to what Google itself says. Video Remix is a feature in Google Photos that produces videos quickly from the user's own material. According to Google's blog post, it runs on Gemini and offers template-based effects and style changes — altering the lighting in a clip, swapping the background, or rendering footage in styles like watercolor, sketch, or oil painting.

That means Video Remix is not, for now, a tool that invents arbitrary new events out of thin air. It is a styling and remixing machine for clips you already own. But the line is thinner than it looks, and the mechanism is the same as in Video Boost: the compute-heavy video processing is moved to Google's cloud, and once pixels can be fetched from a model instead of a sensor, the technical boundary between "processing" a clip and "generating" one disappears.

The Pixel 11 launch hard-codes this strategy into the product line. AP describes the new series' slimmer camera modules, more powerful zoom, and more AI features. Google's devices chief Rick Osterloh told CNBC that Pixel and iPhone are now heading in "very different directions," and that Google is leaning on Gemini features the company believes are unique to Android. TechRepublic calls it Google's biggest Gemini push on mobile yet. In other words: generated content is not a side branch next to the camera. It is the camera strategy.

The Archive Stops Being Evidence

The consequence doesn't land on tech reviewers. It lands on anyone with a Google Photos archive.

For a home video clip, "as filmed" has long been a silent assumption. When remixing is one tap away, the untouched clip becomes one version among several generated alternatives — and no one who views the result later can readily tell the versions apart. The assumption flips: that something "really happened" becomes a statement that must be documented, not something an audience can take for granted.

This is not hypothetical. Google took twice as long as usual to answer on this point: the company had to remove an AI feature because it invented a nuclear power plant, and at the same time it stands established that a flagship camera delivers footage in which half the frames are model output. The failure mode is documented, at both levels — in the editing tool and in the capture pipeline itself.

The Counterargument That Actually Holds

There is a real defense for all of this, and it should be taken seriously.

Video Boost shows why: frame generation in the cloud is the price of video quality the phone doesn't have the hardware to film. For the user, it can be a reasonable trade — a smoother memory in exchange for frames no one will miss. Osterloh's point about "very different directions" is not just marketing; Apple and Google have chosen different answers to the question of what a phone is for, and Google's answer is that the cloud and the model should do more of the work.

And Google has shown the mechanism can correct itself: the Earth feature was switched off when it failed. That feedback loop is worth something.

The counterargument has a blind spot all the same. The New York Times's review of the Pixel 11, published August 19, asks whether users want this phone at all — an AI that orders groceries, books tables, and takes photos on the owner's behalf. The skepticism is not about whether the models can; it's about whether anyone asked. Transferred to the photo archive, the question sharpens: the archive's value lay precisely in the fact that no one chose its contents. It was what happened. When generation becomes the default's next-door neighbor, it is no longer only the person filming who decides what the memory contains — it is also the template and the model.

The Burden Shifts

Labels and watermarks can help, but they don't solve the core problem: when the cloud can supply whatever frames are missing, the responsibility for remembering what actually happened moves from the camera to the person holding it. The photo archive was the one place where you didn't have to testify to anything.

Now it's a format.

Just two weeks separated Google Photos' Video Remix launch in July from the nuclear power plant Google had to remove in August. The company likes to call these directions. For the user, it's a choice that has already been made — in the archive, in the camera settings, in the next clip no one will ever ask about.