Post

The CDN is used for content teams who use non transparent images at large(r) sizes. From what I know we found a cutoff at certain sizes where avif wins, mostly in compute times. It is probably using libjxl but i can't be certain. Will hook you up with the people if you want more details 👌

Like I said in the article, they don't add anything that videos don't (except transparency in AVIF, but that's because AV1 was dumb not to add support). They play like shit, and with the exception of AVIF, their file sizes are baaaaad.

GIF culture is still popular and I find it weird to use video for a 2 frame infinite loop. I trust you when you say current formats are bad for this, but it seemed like your opinion was less empirical, that you didnt like the idea of an animated image format no matter how good it is

Performance is nice but I find regrettable these new image formats do not aim to redefine what a digital image is. For instance, I would really like to see an image format that supports depth information & parallax layers ; would be great in VR or on any other setup with eye-tracking

Didn't use a stock encoder (I've been working on one based on libjxl). On illustrations like this, a well ordered global palette tends to do better instead of a ton of local palettes. What I meant when I said current encoders kinda suck.

It takes my breath away how good these encoders are. I had to zoom in very far to notice any difference whatsoever while moving this slider.

I'm so confused this is the opposite of everything we were hearing about JPEG-XL on recent years, like at the time when Google canned it over a patent spat with Microsoft is everything here outdated? jpegxl.info what happened? did AVIF suddenly go way beyond it's initial benchmarks? 🤷‍♂️

jpegxl.infoJPEG XL: Superior Image CompressionExplore JPEG XL, the next-generation image format delivering superior compression, fast encoding/decoding, and seamless JPEG transcoding. The future of image compression is here.

I've been calling out misinformation on JPEG XL for years. Google didn't can it. They've been actively funding it. Many of the most active contributors to JPEG XL work at Google. The "Google is trying to silence JPEG XL" thing was pure conspiracy theory.

really? but I saw the issue where Chromium maintainers removed it and closed it I don't remember the reason that was given, but I distinctly remember it being obviously bs and then the patent that both Google and Microsoft tried to snatch up you think that was all bs? 🤔

It was never released in Chrome. It was behind a flag, and getting stuff behind a flag in Chrome is a low bar. They looked at it, and decided it wasn't really worth it, so they didn't take the experiment further. Is that bs?

I love the 2010s YouTube thumbnail style og:image. Great post, as always. I'm going to have to reevaluate a lot of my JPEG XL understanding, as a lot has changed in recent years.

Some of it is down to change, but a lot of it is down to misinformation. A lot of the claims around JPEG XL just simply haven't been true, and folks got pretty unpleasant if you tried to challenge it.

The JPEG XL images mostly seem better to me at 60-80% quality levels. It just seems smarter about what detail it is preserving like inside the race car cockpit, the glasses outline and the eyes, or the red panda's eyes and nose texture.