Hacker News Reader

Comment by kps | Hacker News Reader

Excellent. JPEG XL is a usable general-purpose image format, not just a passive delivery format like WebP and AVIF.

Could you explain how so? JPEG XL seems cool enough but what about e.g. AVIF makes it "just a passive delivery format" instead of a general-purpose image format? It handles color spaces, lossless, lossy, up to 12 bit, up to 4:4:4/RGB, transparency, and animation at a good quality/size ratio so what's so drastically different about JPEG XL to move them to separate categories?

AVIF does not support progressive encoding.

Assuming you meant "de"coding: it does, albeit I like JPEG XL's more. To me, that point seems an argument for things being the other way around anyways (i.e. general-purpose but not a good delivery format).

I stand (partially) corrected: https://github.com/AOMediaCodec/libavif/pull/640

I am not sure how exactly the "progressive avif" works, but it looks like basically something optional that comes at a cost in the final file size, whereas in jxl some basic progressive encoding is always available (at least for lossy images) and doesn't come at a cost in file size.

> Assuming you meant "de"coding

Is there something wrong with the other phrasing? You need to have a progressively encoded image before you can decode it progressively.

I think 12-bit colour stops it being adopted for ‘master’ images intended for editing. The current top comment says that 16-bit depth is a requirement for radiology. Also, I've read that its lossless encoding is often very poor — larger than PNG.

(I didn't downvote, BTW; it's a reasonable question.)

the tldr is that video formats are relationship mediocre for images because the tradeoffs you make to compress an image that will be seen for 1/60th of a second are different than those for a static image.

Your point is not wrong in general but I want to nitpick that webp and avif are based on video compression techniques for I-frames which while themselves only shown for a single 1/FPS duration will impact a longer time slice of the video as other frames will reference them.

I'm not very familiar with media encoding formats. What makes webp and avif well suited for delivery but not general purpose?

They're suited for delivery because they're based on video codecs (WebP on VP8, AVIF on AV1) for which consumer hardware typically has hardware decoding support, but were not designed for still images. AVIF is newer and substantially better than WebP, but still has limited colour depth and channels, so it's comparatively poor for editing. Lossless mode is not good for non-photo content. Encoding is relatively slow without hardware support.

> They're suited for delivery because they're based on video codecs (WebP on VP8, AVIF on AV1) for which consumer hardware typically has hardware decoding support,

But no one is doing HW decode for web images. (other than maybe Apple with JPEG? I've seen conflicting reports). AVIF images that aren't actually videos are always software decoded.

Reply on news.ycombinator.com