Apple's iOS 27 Photos AI Editing Features: Extend, Spatial Reframing, and Upgraded Cleanup

Apple announced this week at WWDC 2026 that iOS 27 will bring three AI-powered editing tools to the built-in Photos app: Extend, Spatial Reframing, and a substantially upgraded Cleanup. The update brings iOS 27 Photos AI editing features that previously required specialized editing software or third-party apps into the default app that ships across iOS 27, iPadOS 27, and macOS 27, AppleInsider reported this week. Apple is presenting all three as practical fixes for common photography mistakes, not creative AI generation tools.

Craig Federighi said at WWDC that Apple is "concerned" that AI could impact how "people view photographic content as something they can rely on as indicative of reality," The Verge reported. That concern, voiced during a product keynote, sits directly alongside the announcement of tools designed to make it easier to change what a photograph shows.

What Extend, Spatial Reframing, and Cleanup actually do

Extend works in reverse of cropping. Instead of trimming edges, it synthesizes new image content beyond the original frame. Pinch to zoom out, and the app fills the exposed canvas with generated scenery matched to the existing shot. Practical uses include giving a subject more breathing room, fitting a portrait to a widescreen display, or straightening a tilted horizon without losing anything already in frame, TechCrunch reported this week. The tool can also rework older photos to fit wallpapers and social media formats, AppleInsider noted. The closest existing equivalent is Adobe Photoshop's Generative Expand, The Verge observed, a feature that has lived behind a paid Creative Cloud subscription.

Spatial Reframing changes the apparent viewpoint of a photo after it was taken. The user touches and drags the image to shift perspective; edges blur while Apple's generative models synthesize what would have been visible had the camera been repositioned, with a real-time preview of the effect as it's being applied, per TechCrunch this week. Apple says the tool builds on spatial modeling developed for Vision Pro and generates new pixels only where the perspective change requires them, leaving the original subject and untouched areas intact, per The Verge and AppleInsider (both this week). Extend changes the frame size; Spatial Reframing changes the simulated camera position within an unchanged frame.

Cleanup removes objects from within a photo and reconstructs the scene behind them. Users can tap, brush, or circle what they want removed, TechCrunch reported. Earlier versions worked best on small, simple removals. The iOS 27 version is a "major upgrade," according to Apple, with better reconstruction quality even on complex backgrounds and the ability to handle larger removals without the result falling apart, per The Verge and AppleInsider (both this week). Users can now choose which underlying model handles the removal: Fast for quick results, High Quality for detailed reconstruction, or Auto to let the system decide.

Output quality for all three tools will vary depending on image complexity and how much new content needs to be generated, AppleInsider noted. WWDC demos use controlled examples. Real-world performance across varied lighting, intricate backgrounds, and large removals has not been independently tested.

Apple Intelligence Photos editing: the "fix, don't fabricate" framing and its limits

Apple designed all three tools to repair common photography mistakes rather than generate new images from scratch, AppleInsider observed this week. The portrait where someone got cut off. The family photo with a stranger in the background. The tilted horizon that would require cropping out half the scene to fix. By anchoring the tools to scenarios most users have encountered, Apple positions this as photography correction rather than AI fabrication.

That framing also does pragmatic work. It places these edits closer to the long-accepted tradition of darkroom correction, separating them from text-prompt image generators that produce content with no photographic source material as an anchor. A mainstream audience that has accepted red-eye removal for decades may be more comfortable with a tool framed as "fix the frame" than one framed as "generate a scene."

The comparison to existing professional tools is worth noting. Cleanup, Extend, and Spatial Reframing bring capabilities into a free, default app that photographers have previously needed dedicated software, significant technical skill, or considerable time to execute. That shift in access is the real scope of what Apple announced, AppleInsider reported. The average user gains editing capabilities they didn't have before. So does anyone who wants to quietly alter what a photo records.

Any image edited through Cleanup, Extend, or Spatial Reframing will carry a SynthID watermark, Google's framework for embedding AI provenance signals in image files, The Verge reported this week. Apple also continues to refer to the outputs as "photos," a labeling choice The Verge identified as significant given how substantially the tools can change what those photos depict. Federighi's "concerned" quote reads differently in that context. Apple is aware it is entering contested ground, not just shipping a convenience update.

What watermarking does not resolve is more specific. Whether SynthID markers survive common downstream paths screenshots, social media compression, messaging apps, third-party exports is not confirmed in any publicly available documentation. The research available at launch does not establish durability outside the Apple ecosystem. Watermarking addresses detection in principle; it does not restrict which edits can be made, to which images, or by whom. Apple has not publicly described any additional safeguards beyond the watermark itself.

Three open questions before the fall release

Developer betas of iOS 27, iPadOS 27, and macOS 27 are available now, with public betas expected later this summer and the full release this fall, AppleInsider reported this week. Three questions will determine whether this amounts to a genuine mass-market editing shift or a narrowly scoped demo win.

Device eligibility is unconfirmed. Apple has not said which iPhone or iPad models will support Extend, Spatial Reframing, and the upgraded Cleanup. These features depend on Apple Intelligence, which has historically required recent silicon. The answer to this question will determine whether the update reaches the majority of active iPhones or lands primarily for users who have upgraded hardware in the past year or two. A capability available only to recent buyers is a different kind of announcement than one that reaches the full installed base.

Real-world output quality on complex scenes is unverified. WWDC demos are curated by design. The research notes explicitly that results will vary with image complexity and the volume of content that needs to be synthesized, AppleInsider noted this week. Independent testing on genuinely difficult images, complex backgrounds, large-scale removals, and mixed lighting will be the actual benchmark. The "major upgrade" claim for Cleanup is Apple's own characterization; outside reviewers haven't had access to confirm it.

Watermark durability after sharing is an open question. Whether SynthID markers survive a screenshot-to-iMessage round trip, a social media upload, or a third-party export is not established in any documentation published to date. That gap matters for the accountability argument Apple is making. A provenance signal that disappears the moment an image leaves the Photos app offers weaker guarantees than one that travels with the file. Apple has not addressed this publicly.

The betas will start producing answers on all three fronts, and independent testing during the summer will give a clearer picture of where the WWDC pitch holds up and where it doesn't.