Inside iOS 27: Apple’s New AI Photo Tools—Extend, Reframe, and the Enhance Question Mark

Spatial Reframing: Changing Perspective Post-Shoot

This tool uses AI and spatial modeling to effectively simulate a different camera angle after a photo has been captured. If you missed eye contact by a split second, accidentally caught a distracting street sign above someone’s head, or simply prefer the composition a few inches to the right, you can touch and drag the image to shift perspective . The feature works on standard 2D photos, not just Apple’s spatial videos—the "spatial" label refers to the computational modeling behind the scenes .

The Enhanced Clean Up Tool

Apple didn’t just add new tools; it overhauled the existing Clean Up object remover. The upgraded version can now target and remove "even larger objects" against busy, complex backgrounds—a significant step up from its previous limitations .

The Enhance Question Mark

Before WWDC, Bloomberg’s Mark Gurman reported that three tools were in development: Extend, Reframe, and Enhance. Enhance was described as a one-tap auto-improvement button akin to Adobe Lightroom’s Auto tone tool, designed to intelligently adjust lighting, color, and exposure . During the June 8th keynote, Apple did not prominently demo or even mention Enhance, showcasing only Extend, Reframe, and the new Clean Up . Earlier insider reports noted that while Enhance was the least technically ambitious of the three, Extend and Reframe had been struggling in internal testing with coherent generative output and perspective stability . It’s unclear whether Enhance was pulled from the final release, merged into another feature, or simply omitted from the presentation; its status remains unconfirmed for the public fall release.

Apple’s Privacy Architecture for Photo AI

The tension with generative AI is that complex tasks like outpainting or perspective rendering require serious computational power. Apple’s solution is a three-tier privacy hierarchy that only escalates to the cloud when absolutely necessary.

• On-Device (Tier 1): Simple adjustments and the initial processing of an image occur entirely on the iPhone’s Neural Engine using local, distilled models. No image data leaves the device.
• Private Cloud Compute, or PCC (Tier 2): Complex generative work—like Extend’s outpainting—routes to Apple’s custom Apple Silicon servers housed in sealed data centers. Apple’s architecture is stateless: data is processed ephemerally and is never stored, logged, or accessible to anyone including Apple employees .
• The Google Gemini Question (Tier 3): Although Apple partnered with Google to use a custom 1.2-trillion-parameter Gemini model for advanced Siri intelligence , the photo editing tools operate on a separate stack. Apple’s official architecture documents indicate that generation models for tools like Extend and Spatial Reframing run on Apple’s own foundation models within the PCC layer, not on Google’s infrastructure . Raw user photos are not passed to Google.

Daily Usage Caps and iCloud+

Since server-side image generation is computationally expensive, Apple imposes daily usage limits on features that require cloud processing . These limits are enforced on the device itself. iCloud+ subscribers will receive measurably higher daily caps as a new added benefit, positioning privacy not just as a feature, but as a managed resource .

Hardware Requirements: A Three-Tier Compatibility Reality

Apple’s hardware landscape for iOS 27 is a patchwork of capabilities.

Base OS Compatibility: iOS 27 supports iPhone 11 and newer, including the iPhone SE (2nd generation). This keeps the same support window as iOS 26 .

Apple Intelligence Cutoff (Basic): The AI photo tools (including Clean Up, Extend, and Reframe) are part of the Apple Intelligence suite, which requires at least an iPhone 15 Pro (A17 Pro chip) or newer .

The 12GB RAM Cutoff (Advanced): The most powerful on-device Apple Foundation Model demands 12GB of unified memory. This hardware requirement means it will only function on the iPhone 17 Pro, iPhone 17 Pro Max, and the ultra-thin iPhone 17 Air. The standard iPhone 17, which ships with 8GB of RAM, is not supported for the highest fidelity model processing .

Bottom line: The flashy tools are not a one-size-fits-all update. Your experience will vary significantly based on whether you have an iPhone 15 Pro, a standard iPhone 17, or the top-tier Pro models.

Broader Context: Conservative by Design

The photo editing launch must be viewed alongside Apple’s larger WWDC strategy. Apple deliberately positioned the iOS 27 photo tools as “subtle” editors, focusing on on-device processing wherever possible . This contrasts starkly with Google’s more aggressive, cloud-reliant Pixel AI photography. The internal development difficulties, and the conspicuous soft-pedaling of the Enhance feature, demonstrate that Apple is willing to delay or downscale features rather than ship unreliably or compromise its privacy marketing.

The photo updates arrive alongside the Gemini-powered Siri overhaul, new AI Safari tab management, and new parental controls , cementing iOS 27 as a massive AI update—but one where execution is clearly more challenging than the polished keynote lets on.