you are viewing a single comment's thread
view the rest of the comments
[–] 10 points 1 month ago* (last edited 1 month ago) (3 children)

The return of the king

I really hope adoption catches on. It was mostly killed because Google was trying to push webp, I think hoping that services would embrace its DRM features. That gave AVIF breathingroom, but it was more of an afterthought from AV1 and not really designed as a proper image format. I'm excited to see the browsers start to slowly come back around to JXL

  • source
  • hideshow 3 child comments
  • [–] 3 points 1 month ago (1 child)

    In terms of performance and compression, AVIF is almost considered better than JXL (different images vary) - why push for JXL rather than AVIF? (I love JXL and use it myself when I can, but I wish to have the argument ready for when I'm asked this.)

  • source
  • parent
  • hideshow 1 child comment
  • [–] 4 points 1 month ago* (last edited 1 month ago)

    In order of importance imo

    Great lossless, progressive rendering (big difference for web browsing esp on slow net or when browsing many imgs), Backward compatibility (lossless transcoding of jpg), better performance for non-photo content, and faster encoding in some circumstances.

    Subjectively, I've compared AVIF and JXL a fair bit and found JXL to be better at a given size (roughly, hard to get exact) except for very extreme compression where AVIF seems to pull ahead but no one should be crushing their shit that hard I think.

    The main argument for AVIF is that you mostly get it for free if you implement AV1.

    https://speedvitals.com/blog/avif-vs-jpeg-xl/

    Edit personal testimony: I've been using JXL for a while now for all my internal stuff. e.g. I do a lot of av1 anime transcoding I've settled on 'perceptively lossless' 99% JXL for my comparison images because I and everyone I tested couldn't reliably pick the compressed image in blind AB tests. AVIF had a slight skew, idk if statistically significant; but it's filesizes were also bigger by that point (I know it's taboo to not use lossless for frame comparisons, but I wouldn't be transcoding if I didn't care about file size on disk and keeping a bunch of lossless images per season really adds up at scale, and it seemed to preserve enough high frequency detail that comparisons were meaningful and representative)

  • source
  • parent