Early reports indicate that the iPhone 11 sensor has a higher ISO range and faster possible shutter speeds. But Apple told me that the real improvements are due to a bump from an 8-bit rendering pipeline to 10 bits, and something it calls “semantic rendering,” which is basically an update to Smart HDR that recognizes individual elements of an image and adjusts them appropriately.
From my conversations with Apple, semantic rendering basically goes like this:
The iPhone starts taking photos to a buffer the instant you open the camera app. So by the time you actually press the shutter button, it’s captured four underexposed frames and the photo you want. Then it grabs one overexposed frame. (This is all basically the same as the iPhone XS and the Pixel 3, except the Pixel doesn’t grab that overexposed frame.)
Smart HDR looks for things in the photos it understands: the sky, faces, hair, facial hair, things like that.
Then it uses the additional detail from the underexposed and overexposed frames to selectively process those areas of the image: hair gets sharpened, the sky gets de-noised but not sharpened, faces get relighted to make them look more even, and facial hair gets sharpened up.
Smart HDR is also now less aggressive with highlights and shadows. Highlights on faces aren’t corrected as aggressively as before because those highlights make photos look more natural, but other highlights and shadows are corrected to regain detail.
The whole image gets saved and shows up in your camera roll.
This all happens instantly every time you take a photo.