Apple tests AI photo edits in iOS 27 Photos app
The iOS 27 developer beta adds upgraded Clean Up, image expansion and perspective-shifting tools that can also invent details.
By Hana Yoshida · Markets Reporter
3 min read
Apple is testing a broader set of AI photo-editing tools inside the iPhone’s native Photos app, according to The Verge’s Allison Johnson. The changes matter because they bring generative image editing to a default app used by iPhone owners, raising fresh questions about what a phone photo represents.
The tools are available now in the iOS 27 developer beta, The Verge reported, and Apple may still change them before a public release. Johnson described three additions: an improved Clean Up tool, a new Extend option and a more ambitious Spatial Reframing feature.
Clean Up gets cloud help
Clean Up is Apple’s object-removal feature, used for tasks such as taking a person or unwanted detail out of the background. The Verge said the version introduced last year relied on on-device models and often left visible glitches after removing objects.
In iOS 27, Clean Up can use more powerful cloud-based models, according to The Verge. Johnson reported that the updated tool was far more convincing in her testing and compared the approach to Google’s Magic Editor, which has used cloud processing for years.
The Verge said the improved Clean Up worked well for routine edits, including removing small blemishes or background distractions. Johnson predicted the feature would likely be popular with iPhone users because it handles common cleanup jobs with little friction.
Extend adds pixels around the frame
The new Extend feature lets users widen the borders of an existing photo by generating content beyond the original edges, The Verge reported. Johnson described it as useful when a subject was framed too tightly and the photographer wants more space around the composition.
The feature appears to have limits. According to The Verge, Extend adds only a modest amount of extra image area, sometimes restricts expansion to certain directions and appears reluctant to alter people in the frame.
Johnson found the results generally credible, including one case where the tool generated a missing portion of a rally car and added a matching side mirror. The Verge also reported a case where Extend created a potted plant on a side table, a detail that looked believable but was not present in the original scene.
Spatial Reframing changes perspective
Spatial Reframing goes further by letting users adjust a photo as if the camera had shifted position, The Verge reported. The feature builds on Apple’s existing 3D-style photo effect and allows only limited movement, roughly within the range of how far the photographer might have moved an arm while taking the picture, according to Johnson.
In practice, The Verge found the results less predictable than with Extend. Johnson reported that a reframed photo from an Apple event appeared to generate a person next to Craig Federighi, and that close-up subjects could look distorted when the AI filled in missing facial detail.
The Verge said images edited with these tools receive SynthID labels indicating AI modification. Johnson reported that Instagram recognized the information on uploaded images, though users had to open the image’s “AI Info” menu to see it.
Johnson wrote that labels alone do not resolve concerns about trust in phone photos. The Verge’s testing suggests Apple’s tools are more restrained than some rival efforts, but even limited edits can add uncertainty about whether a photographed object, person or camera position was real.
This story draws on original reporting from The Verge.