How Photonic Engine Improves Low-Light Photos in iOS 16

Apple introduced the Photonic Engine in iOS 16 to significantly enhance computational photography in mid-to-low-light environments across the iPhone 14 lineup. By shifting the multi-frame image processing system known as Deep Fusion earlier into the imaging pipeline, the Photonic Engine preserves dramatically more data, resulting in brighter exposures, true-to-life colors, and finer texture preservation in challenging lighting conditions.

Moving Deep Fusion Earlier in the Pipeline

In earlier camera architectures, Apple applied Deep Fusion—a pixel-by-pixel machine learning process—later in the imaging process after images were demosaiced and compressed. The Photonic Engine fundamentally alters this workflow by executing Deep Fusion directly on uncompressed, linear RAW data straight from the sensor.

Because it operates on raw sensor data, the algorithm works with the maximum amount of light information captured by the camera hardware. By processing uncompressed pixels before any data loss occurs, the system retains subtle details, subtle dynamic range gradients, and color accuracy that would otherwise be discarded during image processing.

Multi-Exposure Frame Fusion

The Photonic Engine relies on advanced multi-frame image stacking. When a user presses the shutter in low light, the camera captures a rapid burst of frames at various exposure values, including short exposures to freeze motion and preserve highlights, alongside longer exposures to gather shadow details.

Using the A-series Neural Engine and Image Signal Processor (ISP), the Photonic Engine analyzes these multiple exposures at a sub-pixel level. It selects the best elements from each frame—taking sharp edges from the underexposed shots and color saturation from the brighter ones—and merges them into a single, cohesive file. This significantly reduces digital noise without the aggressive smearing or over-smoothing often seen with traditional software noise reduction.

Real-World Improvements Across All Lenses

By executing processing earlier and leveraging machine learning, the Photonic Engine yields noticeable gains across all cameras on the device:

  • Greater Detail in Textures: Intricate patterns, fabrics, hair, and foliage in dim environments maintain their natural structure rather than turning into soft, muddy patches.
  • Realistic Color Reproduction: Low-light scenes often suffer from color cast or desaturation. The Photonic Engine preserves natural skin tones and accurate background hues by calculating color values directly from the RAW data.
  • Wider Dynamic Range in Dim Light: Shadows are brightened naturally without blowing out subtle ambient light sources, such as candles, neon signs, or streetlamps.
  • Universal Camera Support: The enhancement applies to the Main camera, the Ultra Wide, the Telephoto, and the front-facing TrueDepth camera, delivering up to 2x to 3x improvements in low-light capture compared to previous generations.