JPEG XL faces fierce competition from AVIF. And Google. A bit of both0:00 - What is JPEG?1:20 - JPEG XL2:07 - AVIF3:08 - JPEG XL's Strengths4:58 - AVIF Suppo...
Jpeg XL isn’t backwards compatible with existing JPEG renderers. If it was, it’d be a winner. We already have PNG and JPG and now we’ve got people using the annoying webP. Adding another format that requires new decoder support isn’t going to help.
Jpeg XL isn’t backwards compatible with existing JPEG renderers. If it was, it’d be a winner.
According to the video, and this article, JPEG XL is backwards compatible with JPEG.
But I’m not sure if that’s all that necessary. JPEG XL was designed to be a full, long term replacement to JPEG. Old JPEG’s compression is very lossy, while JPEG XL, with the same amount of computational power, speed, and size, outclasses it entirely. PNG is lossless, and thus is not comparable since the file size is so much larger.
JPEG XL, at least from what I’m seeing, does appear to be the best full replacement for JPEG (and it’s not like they can’t co-exist).
It’s only backwards compatible in that it can re-encode existing jpeg content into the newer format without any image loss. Existing browsers and apps can’t render jpegXL without adding a new decoder.
Legacy client support. Old devices running old browser code can’t support a new format without software updates, and that’s not always possible. Decoding jxl on a 15yo device that’s not upgradable isn’t good UX. Sure, you probably can work around that with slow JavaScript decoding for many but it’ll be slow and processor intensive. Imagine decoding jxl on a low power arm device or something like a Celeron from the early 2010s and you’ll get the idea, it will not be anywhere near as fast as good old jpeg.
Google rammed webp through because it saved them money on bandwidth (and time during page loading) and because they controlled the standard. They’re doing the same thing with jpeg now that they control jpegli. Jpegli directly lifts the majority of features from jpegxl and google controls that standard.
That’s a good argument, and as a fan of permacomputing and reducing e-waste, I must admit I’m fairly swayed by it.
However, are you sure JPEG XL decode/encode is more computationally heavy than JPEG to where it would struggle on older hardware? This measurement seems to show that it’s quite comparable to standard JPEG, unless I’m misunderstanding something (and I very well might be).
That wouldn’t help the people stuck on an outdated browser (older, unsupported phones?), but for those who can change their OS, like older PC’s, a modern Linux distro with an updated browser would still allow that old hardware to decode JPEG XL’s fairly well, I would hope.
Optimized jpegxl decoding can be as fast as jpeg but only if the browser supports the format natively. If you’re trying to bolt jxl decoding onto a legacy browser your options become JavaScript and WASM decoding. WASM can be as fast but browsers released before like 2020 won’t support it and need to use JavaScript to do the job. Decoding jxl in JavaScript is, let’s just say it’s not fast and it’s not guaranteed to work on legacy browsers and older machines. Additionally any of these bolt on mechanisms require sending the decoder package on page load so unless you’re able to load that from the user’s cache you pay the bandwidth/time price of downloading and initializing the decoder code before images even start to render on the page. Ultimately bolting on support for the new format just isn’t worth the cost of the implementation in many cases so sites usually implement fallback to the older format as well.
Webp succeeded because Google rammed the format through and they did that because they controlled the standard. You’ll see the same thing happen with the jpegli format next, it lifts the majority of its featureset from jpegxl and Google controls the standard.
The video actually references that comic at the end.
But I don’t see how that applies in your example, since both JPEG and JPEG XL existing in parallel doesn’t really have any downsides, it’d just be nice to have the newer option available. The thrust of the video is that Google is kneecapping JPEG XL in favor of their own format, which is not backwards compatible with JPEG in any capacity. So we’re getting a brand new format either way, but a monopoly is forcing a worse format.
Adding more decoders means more overheads in code size, projects dependencies, maintanance, developer bandwidth and higher potential for security vulnerabilities.
A balance has to be struck. The alternative isn’t not getting anything better, it’s being sure the benefits are worth the costs. The comment was “Why is [adding another decoder] a negative?” There is a cost to it, and while most people don’t think about this stuff, someone does.
The floppy code was destined to be removed from Linux because no one wanted to maintain it and it had such a small user base. Fortunately I think some people stepped up to look after it but that could have made preserving old software significantly harder.
If image formats get abandoned, browsers are going to face hard decisions as to whether to drop support. There has to be some push-back to over-proliferation of formats or we could be in a worse position than now, where there are only two or three viable browser alternatives that can keep up with the churn of web technologies.
I mean, the comic is even in the OP. The whole point is that AVIF is already out there, like it or not. I’m not happy about Google setting the standards but that has to be supported. Does JPEGXL cross the line where it’s really worth adding in addition to AVIF? It’s easy to yes when you’re not the one supporting it.
They’re confusing backwards and forwards compatible. The new file format is backwards compatible but the old renderers are not forward compatible with the new format.
My understanding is that webp isn’t actually all that bad from a technical perspective, it was just annoying because it started getting used widely on the web before all the various tools caught up and implemented support for it.
I just wish more software would support webp files. I remember Reddit converting every image to webp to save on space and bandwidth (smart, imo) but not allowing you to directly upload webp files in posts because it wasn’t a supported file format.
If webp was just more standardized, I’d love to use it more. It would certainly save me a ton of storage space.
I think this might sound like a weird thing to say, but technical superiority isn’t enough to make a convincing argument for adoption. There are plenty of things that are undeniably superior but yet the case for adoption is weak, mostly because (but not solely because) it would be difficult to adopt.
As an example, the French Republican Calendar (and the reformed calendar with 13 months) are both evidently superior to the Gregorian Calendar in terms of regularity but there is no case to argue for their adoption when the Gregorian calendar works well enough.
Another example—metric time. Also proposed as part of the metric system around the same time as it was just gaining ground, 100 seconds in a minute and 100 minutes in an hour definitely makes more sense than 60, but it would be ridiculous to say that we should devote resources into switching to it.
Final example—arithmetic in a dozenal (base-twelve) system is undeniably better than in decimal, but it would definitely not be worth the hassle to switch.
For similar reasons, I don’t find the case for JPEG XL compelling. Yes, it’s better in every metric, but when the difference comes down to a measly one or two megabytes compared to PNG and WEBP, most people really just don’t care enough. That isn’t to say that I think it’s worthless, and I do think there are valid use cases, but I doubt it will unseat PNG on the Internet.
I’m not under the impression it would unseat PNG anytime soon, but “we have a current standard” isn’t a good argument against it. As images get higher and higher quality, it’s going to increase the total size of images. And we are going to hit a point where it matters.
This sounds so much like the misquoted “640K ought to be enough for anybody” that I honestly can’t take it seriously. There’s a reason new algorithms, formats and hardware are developed and released, because they improve upon the previous and generally improve things.
People don’t need to give a shit, you just need websites and servers and applications to produce and convert images to the new format and the rest will happen "by itself’
It should be pretty much invisible to the users themselvea
Soo they added webp and AV1, which aren’t that much better then old jpeg, especially with the modern jpeg encoder JpegLi. But JpegXL is out of the question.
Those examples all have a good reason that does not apply here. Browsers already support multiple formats and added a few in the last decade.
I haven’t dug into the test data or methodology myself but I read a discussion thread recently (on Reddit - /r/jpegxl/comments/l9ta2u/how_does_lossless_jpegxl_compared_to_png) - across a 200+ image test suite, the lossless compression of PNG generates files that are 162% the size of those losslessly compressed with JPEG XL.
However I also know that some tools have bad performance compressing PNG, and no certainty that those weren’t used
It has a higher bit depth at orders of magnitude less file size. Admittedly it has a smaller max dimension, though the max for PNG is (I believe) purely theoretical.
Honest question, does JPEG XL support lossless compression? If so, then it’s probably objectively better than PNG. My understanding with JPEG is that there was no way to actually have lossless compression, it always compressed the image at least a little.
Compared to something like JPEG XL? [PNG] is hands down worse in virtually all metrics.
Until we circle back to “Jpeg XL isn’t backwards compatible with existing JPEG renderers. If it was, it’d be a winner.”
APNG, as an example, is backwards compatible with PNG.
If JPEG-XL rendered a tiny fallback JPEG (think quality 0 or even more compression) in browsers that don’t support JPEG-XL, then sites could use it without having to include a fallback option themselves.
Why are you using PNG when it’s not backwards compatible with gif? They don’t even render a small low quality gif when a browser which doesn’t support it tries to load it.
When png was released, it was unsupported by the majority of browsers (and is still not supported by everything mind you) but didn’t have a fallback to a more widely adopted format. It was finalized 9 years after gif, which admittedly is a third of the gap between now and png finalization.
Fallback support isn’t needed. It never has been before, why would it suddenly be needed now? Servers are more than capable of checking the browser on request and serving a different format based on that. They’ve been capable of doing that for decades, and the effort that goes into it is virtually non existent.
It’s a different culture between PCs and consoles. Consoles are standardized computers - they all have the same* hardware. Game developers can be confident in what functionality their games have access to, and so use the best they can.
PCs in comparison are wildly different from user to user due to being modular: you can pick from many parts to create a computer. As such, devs tend to focus on what most PC’s can do and make them optionally better if the PC has access to supporting hardware (e.g. RTX ray-tracing cores).
Besides, video games are drastically complex in comparison to static images 😛
You can’t add new and better stuff while staying compatible with the old stuff. Especially not when your goal is compact files (or you’d just embed the old format).
You have to offer something compelling for everyone. Just coming out with yet another new standard™ isn’t enough. As pointed out earlier, we already have:
jpeg
Png
Webp
HEIC
What’s the point of adding another encoder/decoder to the table when PNG and JPEG are still “good enough”?
PNG and JPEG aren’t good enough, to be honest. If you run a content heavy site, you can see something like a 30-70% decrease in bandwidth usage by using WebP.
Jpeg XL isn’t backwards compatible with existing JPEG renderers. If it was, it’d be a winner. We already have PNG and JPG and now we’ve got people using the annoying webP. Adding another format that requires new decoder support isn’t going to help.
“the annoying webp” AFAIK is the same problem as JPEG XL, apps just didn’t implement it.
It is supported in browsers, which is good, but not in third party apps. AVIF or whatever is going to have the same problem.
According to the video, and this article, JPEG XL is backwards compatible with JPEG.
But I’m not sure if that’s all that necessary. JPEG XL was designed to be a full, long term replacement to JPEG. Old JPEG’s compression is very lossy, while JPEG XL, with the same amount of computational power, speed, and size, outclasses it entirely. PNG is lossless, and thus is not comparable since the file size is so much larger.
JPEG XL, at least from what I’m seeing, does appear to be the best full replacement for JPEG (and it’s not like they can’t co-exist).
It’s only backwards compatible in that it can re-encode existing jpeg content into the newer format without any image loss. Existing browsers and apps can’t render jpegXL without adding a new decoder.
Why is that a negative?
Legacy client support. Old devices running old browser code can’t support a new format without software updates, and that’s not always possible. Decoding jxl on a 15yo device that’s not upgradable isn’t good UX. Sure, you probably can work around that with slow JavaScript decoding for many but it’ll be slow and processor intensive. Imagine decoding jxl on a low power arm device or something like a Celeron from the early 2010s and you’ll get the idea, it will not be anywhere near as fast as good old jpeg.
But how is that different to any other new format? Webp was no different?
Google rammed webp through because it saved them money on bandwidth (and time during page loading) and because they controlled the standard. They’re doing the same thing with jpeg now that they control jpegli. Jpegli directly lifts the majority of features from jpegxl and google controls that standard.
That’s a good argument, and as a fan of permacomputing and reducing e-waste, I must admit I’m fairly swayed by it.
However, are you sure JPEG XL decode/encode is more computationally heavy than JPEG to where it would struggle on older hardware? This measurement seems to show that it’s quite comparable to standard JPEG, unless I’m misunderstanding something (and I very well might be).
That wouldn’t help the people stuck on an outdated browser (older, unsupported phones?), but for those who can change their OS, like older PC’s, a modern Linux distro with an updated browser would still allow that old hardware to decode JPEG XL’s fairly well, I would hope.
Optimized jpegxl decoding can be as fast as jpeg but only if the browser supports the format natively. If you’re trying to bolt jxl decoding onto a legacy browser your options become JavaScript and WASM decoding. WASM can be as fast but browsers released before like 2020 won’t support it and need to use JavaScript to do the job. Decoding jxl in JavaScript is, let’s just say it’s not fast and it’s not guaranteed to work on legacy browsers and older machines. Additionally any of these bolt on mechanisms require sending the decoder package on page load so unless you’re able to load that from the user’s cache you pay the bandwidth/time price of downloading and initializing the decoder code before images even start to render on the page. Ultimately bolting on support for the new format just isn’t worth the cost of the implementation in many cases so sites usually implement fallback to the older format as well.
Webp succeeded because Google rammed the format through and they did that because they controlled the standard. You’ll see the same thing happen with the jpegli format next, it lifts the majority of its featureset from jpegxl and Google controls the standard.
https://xkcd.com/927/
The video actually references that comic at the end.
But I don’t see how that applies in your example, since both JPEG and JPEG XL existing in parallel doesn’t really have any downsides, it’d just be nice to have the newer option available. The thrust of the video is that Google is kneecapping JPEG XL in favor of their own format, which is not backwards compatible with JPEG in any capacity. So we’re getting a brand new format either way, but a monopoly is forcing a worse format.
The other thing is, disk space and internet speeds just keep getting cheaper so… Why change just for disk space?
The bottom line.
https://xkcd.com/927/
Adding more decoders means more overheads in code size, projects dependencies, maintanance, developer bandwidth and higher potential for security vulnerabilities.
The alternative is to never have anything better, which is not realistic
Yes, it means more code, but that’s an inevitability. We already have lots of legacy stuff, like, say, floppy disk drivers
A balance has to be struck. The alternative isn’t not getting anything better, it’s being sure the benefits are worth the costs. The comment was “Why is [adding another decoder] a negative?” There is a cost to it, and while most people don’t think about this stuff, someone does.
The floppy code was destined to be removed from Linux because no one wanted to maintain it and it had such a small user base. Fortunately I think some people stepped up to look after it but that could have made preserving old software significantly harder.
If image formats get abandoned, browsers are going to face hard decisions as to whether to drop support. There has to be some push-back to over-proliferation of formats or we could be in a worse position than now, where there are only two or three viable browser alternatives that can keep up with the churn of web technologies.
I mean, the comic is even in the OP. The whole point is that AVIF is already out there, like it or not. I’m not happy about Google setting the standards but that has to be supported. Does JPEGXL cross the line where it’s really worth adding in addition to AVIF? It’s easy to yes when you’re not the one supporting it.
They’re confusing backwards and forwards compatible. The new file format is backwards compatible but the old renderers are not forward compatible with the new format.
JPEG XL in lossless mode actually gives around 50% smaller file sizes than PNG
Oh damn, even better than the estimates I found.
My understanding is that webp isn’t actually all that bad from a technical perspective, it was just annoying because it started getting used widely on the web before all the various tools caught up and implemented support for it.
It’s certainly not bad. It’s just not quite as good.
I just wish more software would support webp files. I remember Reddit converting every image to webp to save on space and bandwidth (smart, imo) but not allowing you to directly upload webp files in posts because it wasn’t a supported file format.
If webp was just more standardized, I’d love to use it more. It would certainly save me a ton of storage space.
So… your solution is to stick with extremely dated and objectively bad file formats? You using Windows 95?
What’s wrong with PNG?
For what it is? Nothing.
Compared to something like JPEG XL? It is hands down worse in virtually all metrics.
I think this might sound like a weird thing to say, but technical superiority isn’t enough to make a convincing argument for adoption. There are plenty of things that are undeniably superior but yet the case for adoption is weak, mostly because (but not solely because) it would be difficult to adopt.
As an example, the French Republican Calendar (and the reformed calendar with 13 months) are both evidently superior to the Gregorian Calendar in terms of regularity but there is no case to argue for their adoption when the Gregorian calendar works well enough.
Another example—metric time. Also proposed as part of the metric system around the same time as it was just gaining ground, 100 seconds in a minute and 100 minutes in an hour definitely makes more sense than 60, but it would be ridiculous to say that we should devote resources into switching to it.
Final example—arithmetic in a dozenal (base-twelve) system is undeniably better than in decimal, but it would definitely not be worth the hassle to switch.
For similar reasons, I don’t find the case for JPEG XL compelling. Yes, it’s better in every metric, but when the difference comes down to a measly one or two megabytes compared to PNG and WEBP, most people really just don’t care enough. That isn’t to say that I think it’s worthless, and I do think there are valid use cases, but I doubt it will unseat PNG on the Internet.
I’m not under the impression it would unseat PNG anytime soon, but “we have a current standard” isn’t a good argument against it. As images get higher and higher quality, it’s going to increase the total size of images. And we are going to hit a point where it matters.
This sounds so much like the misquoted “640K ought to be enough for anybody” that I honestly can’t take it seriously. There’s a reason new algorithms, formats and hardware are developed and released, because they improve upon the previous and generally improve things.
My argument is not “we have a current standard”, it’s “people don’t give enough of a shit to change”.
People don’t need to give a shit, you just need websites and servers and applications to produce and convert images to the new format and the rest will happen "by itself’
It should be pretty much invisible to the users themselvea
And I suppose sysadmins and application developers are not people?
You’re thinking in terms of the individual user with a handful of files.
When you look at it from a server point of view with tens of terabytes of images, or as a data center, the picture is very different.
Shaving 5 or 10% off of files is a huge deal. And that’s not even taking into account the huge leap in quality.
jpeg xl lossless is around 50% smaller than pngs on average, which is a huge difference
https://siipo.la/blog/whats-the-best-lossless-image-format-comparing-png-webp-avif-and-jpeg-xl
Soo they added webp and AV1, which aren’t that much better then old jpeg, especially with the modern jpeg encoder JpegLi. But JpegXL is out of the question.
Those examples all have a good reason that does not apply here. Browsers already support multiple formats and added a few in the last decade.
I use arch btw
Only thing I can think of is that PNG is inherently lossless. Whereas JPEG XL can be lossless or lossy.
I haven’t dug into the test data or methodology myself but I read a discussion thread recently (on Reddit - /r/jpegxl/comments/l9ta2u/how_does_lossless_jpegxl_compared_to_png) - across a 200+ image test suite, the lossless compression of PNG generates files that are 162% the size of those losslessly compressed with JPEG XL.
However I also know that some tools have bad performance compressing PNG, and no certainty that those weren’t used
It has a higher bit depth at orders of magnitude less file size. Admittedly it has a smaller max dimension, though the max for PNG is (I believe) purely theoretical.
Honest question, does JPEG XL support lossless compression? If so, then it’s probably objectively better than PNG. My understanding with JPEG is that there was no way to actually have lossless compression, it always compressed the image at least a little.
JPEG XL supports lossless compression with a roughly 35% reduction in file size compared to PNG.
Lovely! 🤌
Until we circle back to “Jpeg XL isn’t backwards compatible with existing JPEG renderers. If it was, it’d be a winner.”
APNG, as an example, is backwards compatible with PNG.
If JPEG-XL rendered a tiny fallback JPEG (think quality 0 or even more compression) in browsers that don’t support JPEG-XL, then sites could use it without having to include a fallback option themselves.
Why are you using PNG when it’s not backwards compatible with gif? They don’t even render a small low quality gif when a browser which doesn’t support it tries to load it.
Are you seriously asking why a commonly supported 27 year old format doesn’t need a fallback, but a 2 year old format does?
When png was released, it was unsupported by the majority of browsers (and is still not supported by everything mind you) but didn’t have a fallback to a more widely adopted format. It was finalized 9 years after gif, which admittedly is a third of the gap between now and png finalization.
Fallback support isn’t needed. It never has been before, why would it suddenly be needed now? Servers are more than capable of checking the browser on request and serving a different format based on that. They’ve been capable of doing that for decades, and the effort that goes into it is virtually non existent.
So what your saying is that I should save everything as a BMP every time. Why compress images when I can be the anchor that holds us in place.
By us I guess I mean the loading bar
That’s why all my files are in TGA.
Forgive my ignorance, but isn’t this like complaining that a PlayStation 2 can’t play PS5 games?
It’s a different culture between PCs and consoles. Consoles are standardized computers - they all have the same* hardware. Game developers can be confident in what functionality their games have access to, and so use the best they can.
PCs in comparison are wildly different from user to user due to being modular: you can pick from many parts to create a computer. As such, devs tend to focus on what most PC’s can do and make them optionally better if the PC has access to supporting hardware (e.g. RTX ray-tracing cores).
Besides, video games are drastically complex in comparison to static images 😛
All the cool kids use .HEIF anyway
I use jpeg 2000
You can’t add new and better stuff while staying compatible with the old stuff. Especially not when your goal is compact files (or you’d just embed the old format).
Isn’t that the same as other newer formats though?
There’s always something new, and if the new thing is better, adding/switching to it is the better move.
Or am I missing something about the other formats like webp?
You have to offer something compelling for everyone. Just coming out with yet another new standard™ isn’t enough. As pointed out earlier, we already have:
What’s the point of adding another encoder/decoder to the table when PNG and JPEG are still “good enough”?
PNG and JPEG aren’t good enough, to be honest. If you run a content heavy site, you can see something like a 30-70% decrease in bandwidth usage by using WebP.