AVIF for Next-Generation Image Coding

An interesting takeaway from the comparisons shown is that AVIF seems to be much better about getting rid of detail without creating artifacts. It's particularly evident in the second example photo, where the original has a lot of detail in the texture of the paint on the door. AVIF drops most of that detail and gives it a smooth, almost airbrushed look. It's obviously missing a lot when you compare it to the original side-by-side, but on its own it doesn't look immediately repulsive.

To me, this absolutely screams "mobile," as on a small screen you probably wouldn't be able to see much of that detail even if it were technically present. I wonder if in the future we'll see some sort of mechanism for progressive image loading that allow the client device to choose its own level of compression, possibly based on server-provided hinting or something.