Apple's Spatial Reframing tool in Photos for iOS 27 is an interesting use of Apple Intelligence, but don't push it too far just yet.
WWDC 2025 included a neat feature for Spatial Photos that lets users convert flat images into 3D scenes. By moving the iPhone around, you could temporarily re-angle your shot and explore the scene.
It was a neat trick, but it wasn't that much of a major feature for photographers. It did, at least, give you an idea of how a Spatial Photo would look on something like an Apple Vision Pro.
One year later, Apple has decided to try and get the feature to be more practical to iPhone users. That comes in the form of Spatial Reframing.
Fixing the imperfect shot
One of the problems with photography is regretting not lining up the shot perfectly. Frequently, you'll look back at what you've just taken, and the background isn't in quite the position you wanted it to be.
If you were to edit it traditionally, you would try to cut around the subject and move it over, then fill in the new empty pixels with a clone tool or something else. An expert image editor can do this, and most people won't tell that an edit has taken place at all.
That requires time and skill, which the average photo-taker doesn't really have or wish to invest.
Spatially reframing an image in iOS 27
Spatial Reframing is a blend of the previous Spatial Photos feature and generative AI smarts. The idea is that you can select an image, the iPhone will analyze it, and then you can alter the angle of the camera's view to a new one.
In theory, that would be a quick and relatively painless process, and with no issues at all. Depending on the photo you throw at it, you may just get that, but with some massive caveats.
Background filling
You can find the Reframe feature under the editing section of an image, under Tools.
Once you tap it, the screen fills up with a multi-colored filter, as it scans the shot. Once scanned, you're instructed to touch and drag to adjust the perspective.
You can also use a two-finger pinch to pan, zoom, and rotate the image.
Dragging the picture around gives you a similar effect to the Spatial Photos, but to a more extreme degree. You can move the angle far enough to one side or the other that you can uncover sections of the background that simply aren't visible in the original shot.
Reframing a portrait of a kitten works well. Original on the left.
In the preview, this is filled in with a minimal generative graphic that won't be used in the final picture. This is especially the case for the edges, which appear blurred because it's a lot to generate on the fly, and that's not necessary in a preview.
Once you have set your new angle, hitting the Reframe button sets the processing in motion. After a few seconds, you have your reframed image.
In the short time we have played with the feature, we tried it out with a pair of images, to see how it works with a close-up portrait and with a wider scene.
The portrait, which we used an old image of a kitten, was handled pretty well. It was a minor shift of the camera to the right, with a small amount of deformity to the cat's image.
More apparent is the background, as the left-hand side of the image was completely generated by the tool. It did pretty well in its blurry state, and if I showed anyone who didn't know the room's layout, they wouldn't tell.
Our second test was with a wide touristy photograph of the Colosseum in Rome. It's a difficult structure with many archways, and the subjects are a distance away from the camera in the middle of the frame.
When lining up the shot, we could tell that it was trying to make a vague-but-acceptable background for the preview, which is fine.
The final image has some plusses, but also some minuses.
Reframing a tourist scene initially looks OK. Original on the left.
On the plus side, it handed generating the background really well. Arches and road that were not visible in the original shot were created and put into place in the back quite well.
Less well done are the subject faces. You can tell that, as part of the reframing, the bodies and heads are taken into account, and are similarly adjusted to match the rest of the image.
This can sometimes work well, but the resulting warp to the faces is unflattering, to put it mildly.
Caveat Emptor
Spatial Reframing, as a concept, makes perfect sense. If you have a camera system and processing that can take apart a scene, move elements around, and smartly generate missing bits, there's no reason not to do it.
This would be a massive task for a human to undertake, so what it's coming up with is pretty phenomenal for a first try.
Close-ups of the original [left] and the warped faces of the reframe [right]
That said, we are talking about a feature that is in a developer beta, that is months from release, and the first real attempt too. It's expected that there will be hiccups and foibles here.
Expect more improvements in the future.
As it stands, it's a nice feature that could make for some fun shot changes. Content altered by the feature is not going to make the cover of Vogue anytime soon, so professional editors can breathe a sigh of relief.
If you don't push it too far, it's decent enough to make your Instagram cat photos a bit better.
