The iPhone takes a referens photo that's meant to stop motion blur as much as possible. Next, it combines three standard exposures and one long exposure into a single "synthetic long" photo. Deep Fusion then breaks down the reference image and synthetic long photo into multiple regions identifying skies, walls, textures and fine details (like hair). Next, the software does a pixel-by-pixel analysis of the two photos -- that's 24 million pixels in total. The results of that analysis are used to determine which pixels to use and optimize in building a final image.
Apple says that the entire process takes a second or so to happen. But to allow to you to continue snapping shots, all of the information is captured and processed when your iPhone's A13 processor has a chance. The idea is that you won't be waiting on Deep Fusion before taking the next photo.
https://www.cnet.com/google-amp/news/iphone-1...