Apple has announced a dramatic overhaul of its RAW photo processing pipeline with the release of iOS 27, iPadOS 27, and macOS 27. The company is introducing RAW 9, the ninth major iteration of its system-level RAW image processing engine. The update leverages machine learning to extract more detail, reduce noise, and improve color accuracy—even when reprocessing older RAW images captured years ago.

What is RAW and Why It Matters

RAW is an image file format that preserves all data captured by a camera's sensor without any in-camera processing. Unlike JPEG or HEIF, which apply sharpening, noise reduction, and white balance adjustments during capture, RAW files give photographers full control over exposure, color temperature, highlights, and shadows during post-production. This flexibility comes at the cost of larger file sizes and the need for specialized software to interpret the sensor data.

Apple has maintained its own RAW processing stack within Core Image, providing support for nearly 800 camera models with individual calibration profiles. Photographers using apps that leverage Core Image—such as Apple Photos, Lightroom, or third-party editors—benefit from consistent, high-quality demosaicing and noise handling across all supported cameras. With each new version of iOS, Apple has refined the algorithms responsible for converting raw sensor data into viewable images. RAW 9 represents the most significant leap in quality yet.

RAW 9: A Machine Learning Breakthrough

The key innovation in RAW 9 is the use of a tiled CoreML model that simultaneously performs demosaicing and denoising. Demosaicing is the process of reconstructing a full-color image from the sensor's color filter array, while denoising removes the luminance and chroma noise that inevitably appear at higher ISOs. Traditionally, these two steps are performed sequentially, which can lead to artifacts or loss of fine detail. By coupling them in a single neural network, RAW 9 can better preserve texture and color fidelity.

David Hayward, a Core Image engineer at Apple, explained during a WWDC session that the model runs on the Apple Neural Engine (ANE), ensuring optimal performance without draining battery life. The ANE's parallel architecture is well-suited for image processing tasks, allowing RAW 9 to apply the model to each tile of the image independently before stitching the results together. This tiled approach also makes it possible to process very large RAW files—like those from 50MP or 100MP sensors—without exceeding memory limits.

Striking Visual Improvements

During the session, Hayward demonstrated several side-by-side comparisons between RAW 8 and RAW 9. In a low-noise image from a Sony Alpha 7 II, RAW 9 produced sharper text and finer details on a vintage dial indicator. The difference was more pronounced in high-ISO scenarios. One example used a Canon 5D Mark III shot at ISO 51,200—an extremely noisy setting. The raw sensor data appeared as a chaotic mix of luminance and chroma noise, with individual crayons in a box nearly indistinguishable. RAW 8 managed to recover some color and structure, but the results were still muddy. RAW 9, however, rendered the crayons with clear, accurate colors and visible specular highlights on the wax surfaces.

Another test involved a Fujifilm X-T5 image captured at ISO 12,800. Fujifilm cameras use a unique X-Trans sensor pattern, which is notoriously difficult to demosaic. RAW 8 introduced color artifacts and lost detail in the texture of embroidery yarn. RAW 9 eliminated those artifacts and made small text on the yarn tags legible. The weave of the thread appeared far more natural, with minimal noise.

Backward Compatibility and Developer Tools

One of the most appealing aspects of RAW 9 is its ability to reprocess existing RAW files. Photographers who have libraries of older images can revisit them in iOS 27 and see significantly improved results. The machine learning model is applied retrospectively, meaning that even photos taken years ago can benefit from the new processing. Developers can enable RAW 9 in their apps by adopting the updated Core Image APIs, and the session provided guidance on how to balance quality and performance for real-time editing versus batch export.

Apple's commitment to RAW processing has made its ecosystem a strong choice for professional and enthusiast photographers. With RAW 9, the company is closing the gap with dedicated RAW converters like Adobe Camera Raw and Capture One, especially in terms of noise reduction and detail preservation. The neural engine ensures that even complex computations happen quickly, making the editing workflow smooth on devices like the iPhone and iPad Pro.

The update also reinforces Apple's broader strategy of integrating machine learning into core system features. From computational photography on the iPhone to RAW processing on Mac, the same underlying technology is being used to deliver tangible benefits. RAW 9 is yet another example of how on-device AI can enhance creative tools without compromising privacy, since all processing occurs locally.

For photographers who want to see the full technical details, Apple's WWDC session 'Enhance RAW image processing with Core Image' is available online. The session covers the CoreML model architecture, performance tuning, and advanced options for custom rendering pipelines. Early adopters of the iOS 27 and macOS 27 betas can already test RAW 9 with their own camera files, and the final release is expected later this year alongside the public launch of the new operating systems.

Source:9to5Mac News