Apple just shipped its first set of serious AI photo editing tools to the iPhone, and they land in the iOS 27 developer beta right now. According to The Verge AI, the update adds three new generative editing features to the native Photos app, marking what the publication calls a tipping point for what Apple lets you do to your own pictures. The Verge AI got hands-on time with the beta, so keep in mind Apple may still tweak these before the public release.
What stands out here is the shift in approach. Apple is finally tapping cloud models instead of relying only on what runs on the phone, which is exactly the move that made Google’s tools so much stronger.
What Apple launched
There are three new tools, or as The Verge AI puts it, maybe two and a half:
- Clean Up 2.0: The object-removal tool that pulls photobombers and distractions out of the background. It existed before but was so weak it barely counted. The big change: it can now route to more powerful cloud models instead of staying fully on-device, so it fills in convincing detail instead of leaving weird artifacts. The Verge AI says it now does the job and expects it to be popular.
- Extend: Think cropping in reverse. It expands the edges of your frame and paints in plausible filler when your composition was too tight. It only adds a little padding, tends to avoid editing people, and sometimes refuses to extend in certain directions.
- Spatial Reframing: The most ambitious and, per the review, the most problematic. It mimics moving the camera around a scene so you can recompose a shot you already took, building on Apple’s existing 3D-ish photo effect.
How it compares
Apple is playing catch-up, and the report says so plainly. By iPhone standards these features are a leap. By the wider market’s standards they’re tame next to what Google’s Pixel phones already do. The Verge AI notes that Google’s Magic Editor has been cloud-powered for years, which is why it ran circles around the on-device version Apple introduced last year. At the same time, Extend looks more restrained than Samsung’s early efforts, which were eager to invent objects out of thin air.
Where it gets weird
Clean Up is the easy win. The Verge AI describes it as the generative tool it feels least queasy about, good for pulling a stranger out of the background without drama.
Extend mostly behaves, leaning on symmetry to fill gaps. It matched a missing side mirror on a rally car convincingly. But it also invented a potted plant on a side table that looked real but wasn’t, the kind of thing that feels off to post on Instagram.
Spatial Reframing is where things get strange. Predicting detail in three dimensions is harder than in two, and it shows. The farther you are from your subject, the more realistic the result but the smaller the actual change, which raises the question of why bother. Get close, and the AI works harder and stumbles. In one test it invented an entire person sitting next to Apple’s Craig Federighi. On selfies, faces skew into uncanny valley territory fast. The reviewer’s verdict on that one: nice in theory, but in practice, no thanks.
Availability and labels
The features are live in the iOS 27 developer beta now, ahead of a public release. Photos edited with these tools carry SynthID labels that flag AI modification. Instagram picked up that signal on upload, but only shows it if you tap into an “AI Info” menu, which means the disclosure is easy to miss.
Why it matters
Apple putting cloud-grade generative editing in the world’s most popular camera is a big deal simply because of reach. Clean Up will get used constantly and mostly help. The harder question is the one The Verge AI leaves hanging: as these tools get easier and labels stay buried, the shared assumption that a photo shows something that actually happened keeps eroding. The capability is here. The norms haven’t caught up.
For the full hands-on breakdown, check the original report at The Verge AI.