Apple RAW Engine 9: The Biggest Camera Software Shift Since 2017

Apple’s Core Image RAW engine has processed nearly a decade of iPhone and Mac photos without a major architectural update. At WWDC 2026, that changed: Core Image RAW Engine 9 ships with iOS 27, iPadOS 27, and macOS 27, and for the first time since 2017, Apple is rebuilding how the software handles the fundamental tension between sharpness and noise.

What Happened

Apple Core Image Engineer David Hayward described the update as “the biggest update yet” to Apple’s RAW handling stack. The architectural change is significant: RAW 8 ran demosaicing (converting the sensor’s Bayer or X-Trans grid into RGB pixels) and denoising as two separate sequential operations. RAW Engine 9 collapses both into a single CoreML inference pass on the Neural Engine in A- and M-series chips.

The technical principle mirrors what DxO has built with DeepPrime 3: combining the two steps lets the model operate on the full sensor pattern while removing noise, rather than denoising pixels that have already been interpolated and thus lost some spatial information. Apple’s expanded dataset covers “784 different cameras,” with explicit support for non-Bayer X-Trans sensors—the sensors used in Fujifilm bodies, which have historically required specialized RAW processing.

Developer testing described the engine as a “new noise reduction king” in high-ISO comparisons, with visible improvements in sharpness, color accuracy, and noise handling up to ISO 51200.

Why It Matters for Builders and Shooters

The delivery mechanism matters as much as the quality improvement. Core Image RAW is an API, not just a first-party feature—every third-party app built on Apple’s Core Image RAW stack gets the quality upgrade automatically when users update to iOS 27 or macOS 27. That includes Pixelmator Pro, and any developer using Apple’s RAW processing APIs. No SDK changes required on the developer side.

For photographers who process through an Apple-ecosystem workflow, this is a meaningful floor raise. The Neural Engine acceleration means the heavier compute doesn’t noticeably slow processing on current hardware—the work happens on dedicated silicon rather than the main CPU.

The competitive context is worth noting: DxO has commanded a premium specifically because its DeepPrime denoising produced visible quality advantages over Apple’s engine. That gap was the primary technical justification for $200+ annual subscriptions to dedicated RAW processors. RAW Engine 9 narrows—though may not fully close—that gap for photographers who don’t need DxO’s lens correction or calibration profiles.

Source: PetaPixel

Content created with AI assistance and reviewed for accuracy.