I compare PNG and the four modern formats, AVIF, HEIF, WebP, JPEG XL, on tasks/images that PNG was designed for. (Not on photographs or lossy compression.)
It seems like the natural categories are (1) photographs of real things, (2) line art, (3) illustrator images, (4) text content (eg, from a scanned document).
Is there a reason you used only synthetic images, ie, nothing from group 1?
The motivation behind the benchmarks was to understand what are the options today for optimizing the types of image we use PNG for, so I used the same set of images I had used previously in a comparison of PNG optimizers.
The reason the set does not have photographs: PNG is not good at photographs. It was not designed for that type of image.
Even so, the set could do with a bit more variety, so I want to add a few more images.
Wonderful. Allow an "unmonitored" extension from a random stranger on the Internet have access to "all data for all websites" just to support an image format for which Mozilla should have long built in native support...
That's not the reason, but the excuse. The reason Firefox doesn't have jxl is that it is funded by Google, and someone at Google decided that it has to die.
Also the parent comment was about that you really shouldn't just let a random Russian guy run any javascript on any website you visit, that's stupid.
Also also, am I missing something, or Firefox extensions are broken, there is no way to limit an extension to websites (allow or disallow), or even just to check the source code of an extension?
> That's not the reason, but the excuse. The reason firefox doesn't have jxl is that it is funded by Google, and someone at Google decided that it has to die.
So what, you think they were just lying when they said that they'll ship JXL when it has a Rust implementation? You think Mozilla devs were just bluffing when they were working directly with the JXL devs over the last year to make sure everything would work right?
Firefox Nightly v149 has added experimental support via Settings > Firefox Labs:
Webpage Display
Media: JPEG XL
With this feature enabled, Nightly supports the JPEG XL (JXL) format. This is an enhanced image file format that supports lossless transition from traditional JPEG files. See bug 1539075 for more details.
One thing I like about JPEG-XL is that it supports all kinds of weird image formats.
For example, I used to work with depth data a lot, which is best expressed as monochrome 16-bit floating point images. Previously, TIFF was the only format that supported this. Many shops would instead save depth images as UINT16 .PNG files, where the raw pixel intensity maps to the camera distance in mm. The problem with this is that pixels more than 65.535 meters away aren't representable. (Hot take: I personally think this is one reason why nobody studies depth estimation for outdoor scenes.)
JPEG-XL supports more weird combinations here, e.g. storing greyscale float32 images (with alpha even! you can store sparse depth maps without needing a separate mask!)
It's like, uniquely suited to these sorts of 3D scene understanding challenges and I really hope people adopt the format for more scientific applications.
> One thing I like about JPEG-XL is that it supports all kinds of weird image formats.
And it is probably the reason why browser vendors disliked it. Lots of complexity, it means a big library, which is high maintenance with a big attack surface. By comparison, webp is "free" if you have webm, as webp is essentially a single frame video.
AFAIK browsers do not reuse any VP8 codepath for WebP, they just use libwebp, which decodes lossy images in software. WebP has a non-VP8 lossless mode too. The concern about image format attack surface is also probably because of the recent exploit in libwebp.
On the subject of tiff, why is it not used more? I mean, it is more or less really a container format right. Why are we not using it all over the place but with modern compression methods?
As just one of innumerable examples, it's the basis for Adobe's DNG raw photo format and many proprietary raw formats used by camera manufacturers (Nikon NEF, Canon CRW and CR2, etc.).
Speaking as an outside observer, the ISO Base Media File Format seems to have more mindshare for newer applications, presumably on account of its broader scope and cleaner design.
Yup, Gnome Web loads it just fine! Man, it really is a great browser. I try to switch to it every 6 months, but then I remember that it doesn't support extensions at all. I could give up everything, but not 1Password. Nothing is worth copy/pasting credentials and losing passkeys entirely.
For what purpose? While it's a perfectly good password manager, when used with Gnome Web it also means copy/pasting passwords and losing passkeys. Doesn't it?
Designers might also be hesitant to use an untested file format for print, too.
If there’s a large amount of paper that’s been purchased for a job, I definitely wouldn’t want to be the one who’s responsible for using JPEG XL and – for whatever reason – something going wrong.
Pixels are cheaper than paper or other physical media :)
They request formats that their equipment handles. They're not in the business of converting a user's file type from one to another. That would be inconsistent from what the user sent.
Here's who I order from, you can see the particulars of what they request.
> They're not in the business of converting a user's file type from one to another.
Their job is getting an image file into reality, not to be the absent owner of a big machine.
> That would be inconsistent from what the user sent.
If the machine accepts some type of normal image file, then they can losslessly convert other file formats to that type. There is nothing inconsistent about that.
My first statement is an opinion/judgement, not an assumption.
I'm confident my second statement is true. Note that any argument that says niche formats are a problem because color space might be ambiguous also applies to the formats they do accept.
Checking the Firefox bugs on this, it seems they decided to replace the C++ libjxl with a rust version which is a WIP, to address security concerns with the implementation. All this started a few months ago.
Maybe the zen fork is a bit older and still using the C++ one?
... update. after reading the comments in the rust migration security bug, I saw they mentioned "only building in nightly for now"
I grabbed the nightly firefox, flipped the jxl switch, and it does indeed render fine, so I guess the rust implementation is functioning, just not enabled in stable.
... also, I see no evidence that it was ever enabled in the stable builds, even for the C++ version, so I'm guessing Zen just turned it on. Which... is fine, but maybe not very cautious.
Google Chrome is using a Rust implementation. The existence and sufficient maturity of it is the reason they were willing to merge support in the first place.
Flipping `image.jxl.enabled` made it work for me after refreshing the page. I'm using Librewolf 146.0.1-1, but I guess it works just fine in firefox 146
Jake Archibald has an excellent post about progressive image rendering, including some metrics on JPEG XL compared to AVIF[0].
> "I was also surprised to see that, in Safari, JPEG XL takes 150% longer (as in 2.5x) to decode vs an equivalent AVIF. That's 17ms longer on my M4 Pro. Apple hardware tends to be high-end, but this could still be significant. This isn't related to progressive rendering; the decoder is just slow. There's some suggestion that the Apple implementation is running on a single core, so maybe there's room for improvement.
> JPEG XL support in Safari actually comes from the underlying OS rather than the browser. My guess is that Apple is considering using JPEG XL for iPhone photo storage rather than HEIC, and JPEG XL's inclusion in the browser is a bit of an afterthought. I'm just guessing though.
> The implementation that was in Chromium behind a flag did support progressive rendering to some degree, but it didn't render anything until ~60 kB (39% of the file). The rendering is similar to the initial JPEG rendering above, but takes much more image data to get there. This is a weakness in the decoder rather than the format itself. I'll dive into what JPEG XL is capable of shortly.
> I also tested the performance of the old behind-a-flag Chromium JPEG XL decoder, and it's over 500% slower (6x) to decode than AVIF. The old behind-a-flag Firefox JPEG XL decoder is about as slow as the Safari decoder. It's not fair to judge the performance of experimental unreleased things, but I was kinda hoping one of these would suggest that the Safari implementation was an outlier.
> I thought that "fast decoding" was one of the selling points of JPEG XL over AVIF, but now I'm not so sure.
> We have a Rust implementation of JPEG XL underway in Firefox, but performance needs to get a lot better before we can land it."
Strange, as Cloudinary's test had the opposite conclusion -- jpegxl was significantly faster to decode than avif. Did the decoders change rapidly in a year, or was it a switch to new ones (the rust reimplementation)?
If decode speed is an issue, it's notable that avif varied a lot depending on encode settings in their test:
> Interestingly, the decode speed of AVIF depends on how the image was encoded: it is faster when using the faster-but-slightly-worse multi-tile encoding, slower when using the default single-tile encoding.
>> My guess is that Apple is considering using JPEG XL for iPhone photo storage rather than HEIC, and JPEG XL's inclusion in the browser is a bit of an afterthought.
Because JPEG XL is the first format to actually bring significant improvements across the board. In some aspects AVIF comes close, in others it falls far behind, and in some it can’t even compete. There’s just nothing else like JPEG XL and I think it deserves to be supported everywhere as a truly universal image codec.
Are there any up-to-date WebKit browsers for Android? The best I could find was Lightning, but it hasn't been updated in years.
Edit: I found A Lightning fork called Fulguris. It didn't work with the JPEG XL test image, but I really like the features and customizability. It's now my default browser on Android.
The closest thing I know of is Igalia has a project trying to port https://wpewebkit.org/ to Android https://github.com/Igalia/wpe-android and they have a minibrowser example apk in the releases of the current state (but I wouldn't call it a Chrome drop in replacement or anything at the moment - just the closest thing I know on Android).
There was a constraint - since 2009, the Joint Photographic Experts Group had published JPEG XR, JPEG XT and JPEG XS, and they were probably reluctant to break that naming scheme.
They're running out of good options, but I hope they stick with it long enough to release "JPEG XP" :-)
In the photography world it's shorthand for "photo unedited straight from the camera". Popular with Fujifilm cameras especially due to their 'film simulation' modes which apply basically a filter to the image.
What makes jpeg compression bad isn’t low bandwidth. It’s really good at compressing an image for that.
What makes jpeg bad is that the compression artifacts multiply when a jpeg gets screen captured and then re-encoded as a jpeg, or automatically resized and recompressed by a social media platform. And that definitely isn’t a problem that has gone away since dialup, people do that more than ever.
I'm not saying it's true, I obviously understand that not all jpegs are low quality and over compressed. That's just how the word is generally used by people, especially those outside of tech who aren't well versed in different image formats.
It seems to me this point of discussion always tends to get way too much focus. Should it really raise concern?
Of all the people who interact with image formats in some way, how many do even know what an image format is? How many even notice they’ve got different names? How many even give them any consideration? And out of those, how many are immediately going to think JPEG XL must be big, heavy and inefficient? And out of those, how many are going to stop there without considering that maybe the new image format could actually be pretty good? Sure, there might be some, but I really don’t think it’s a fraction of a significant size.
Moreover, how many people in said fraction are going to remember the name (and thus perhaps the format) far more easily by remembering it’s got such a stupid name?
Actually, I remember when JPEG XL came out, and I just thought: cool, file that one away for when I have a really big image I need to display. Which turned out to be never.
I regularly work with images larger than 65,535px per side.
WEBP can only do 16,383px per side and the AVIF spec can technically do 65,535, but encoders tap out far before then. Even TIFF uses 32-bit file offsets so can't go above 4GB without custom extensions.
Guess which format, true to its name, happens to support 1,073,741,823px per side? :-)
Honestly, that's exactly what it sounds like to me too. I know it's not, but it's still what it sounds like. And it's just way too many letters total. When we have "giff" and "ping" as one-syllable names, "jay-peg-ex-ell" is unfortunate.
Really should have been an entirely new name, rather than extending what is already an ugly acronym.
Unrelated but I read "it did not saw" and immediately thought, this person is Dutch. Then I saw the .nl domain. Not sure if this double-conjugation mistake is common in other ESL speakers but I hear it a lot living in the Netherlands.
While I get why, it bugs me that they have comparison images between jxl and other formats, yet it doesn't actually use jxl, as evidenced by all images displaying correctly on my chrome browser.
This is standard practice. They need to use current lossless formats to display examples to people who don't have the format yet. They are still showing accurate examples of compression artifacts. I'm not sure what else you'd expect them to do.
According to CanIUse, no browser implementation currently supports progressive decoding [1]. This is unfortunate, since progressive decoding theoretically is a major advantage of JPEG XL over AVIF, which doesn't allow it in principle, even though ordinary JPEG allows it. But apparently even a default (non-progressive) JPEG XL allows some limited form of progressive decoding [2]. It's unclear whether browsers support it though.
That’s an interesting speculation, but I’m inclined to believe their official reasoning. (That being they just didn’t really care about the format and/or went with whatever Chrome said at first. A year or so later they changed their mind and said they wanted an implementation in a memory-safe language, which prompted the JXL team to work on it.)
https://op111.net/posts/2025/10/png-and-modern-formats-lossl...
I compare PNG and the four modern formats, AVIF, HEIF, WebP, JPEG XL, on tasks/images that PNG was designed for. (Not on photographs or lossy compression.)
reply