In an Oct. 24, 2026, post about Portrait mode's blur, @johnnie marks what he calls Portrait mode's 10th anniversary with an account of the idea behind its simulated blur. He invokes a principle from oil painting: “if you get the eyes right, nothing else matters.”
By his account, disparity calculations from two sensors produced a depth map, but areas without texture made those calculations impossible. Fine detail could produce bad transitions with a lower-resolution map, while a full-resolution map took too long to compute, he writes. He thought the initial result looked “fake.” Reducing the blur would have made those problems less obvious, he says, but would have lost the wide-open-lens look the feature was meant to create.
Instead, @johnnie writes, the team used machine learning to find the eyes, drew a mask to protect them and paired it with the depth map. That let them increase blur to hide depth-calculation errors while allowing ears and noses to soften. He says Portrait mode shipped a month after the iPhone because working out the painting-inspired approach took time.
In a quote-post about Portrait mode's hardware and software, @joshelman calls it one of his few true iPhone “wow” moments and praises that mix. In his quote-post on Portrait mode, @panzer admires its combination of history, art and computational science.
Comments