Now that the dust has settled, why did JXL fail to get widespread adoption? Why is Webp winning despite JXL being better overall?
>>108847101>JPEG XL's time is on the horizon.I hope so but I'm talking about TODAY. The jump from JPG to JXL is so massive that I expected Webp to be left in the dust. I guess another phrasing for this question is: what the fuck is keeping Webp popular right now despite being objectively worse than JXL?https://www.corewebvitals.io/pagespeed/jpeg-xl-core-web-vitals-support
>>108847017>why did JXL fail to get widespread adoptionbecause of literally 1 guy who works at google who unilaterally removed the feature flag and locked all new PRs and issues made about it. Apple, Microsoft, Mozilla, Adobe, Meta all adopted it but if it doesn't work on most web browsers there's little end-to-end utility. The worst part is the feature flag finally came back after that guy left the company so nothing was accomplished but stalling for 3 years, it's still going to be the new standard format.
>>108847017unless a new thing that has the same level of backwards compat comes along, it will take off now that the chrome team has stopped being so fucking autistic about it
>>108847183I still kinda doubt this. I've heard that there's decoding performance concern on JXL but truthfully I've seen mixed sources on this. Like Anime images decode faster on Webp????
>>108847226>Anime imagesnot relevant for the people that matter
>>108847017So for lossless formats with smaller sizes the conversion is PNG to webp and jpeg to jxl? After some tests this is the conclusion I have. I never tested avif but I remember something about it not really being lossless.
>>108847017Why is everthing still using jpeg and not jxl on my computer? Why does this site not support the format?
>>108847017>what if you jump through a bunch of extra hoops for, uhh.. 20% filesize savings?bluray movie is 20-50gb so doing it for video makes senseinages… why even bother? phone photos already use heic which is efficient, and the rest is irrelevant
>>108847017Obviously, most are not gonna bother updating their CDNs to support something with an unspecified ETA.AVIF has been universally available for a while now, so it's a safe bet over WebP.
>>108850189I know I'll trigger the daiz spam bot but doesn't AVIF work better with hardware decoding? I know HEIC is a piece of shit that causes phones to jam up if you try to software decode a 50MP HEIC on a phone without hardware decoding.Like Webp seems to be the only image format that doesn't need hardware decoding.
>>108849122dumbfuck
>>108850435Hardware decoding is a meme. Even if the device is in the minority category of 4:4:4 capable, it's no guarantee that the underlying software won't just forego it in favor of dav1d.IIRC, Chromium even disabled JPEG acceleration because of bugs.
>>1088505464:4:4 isn't necessary for AVIF to be good though. At least not anymore. >SharpYUV Chroma Subsampling optimizes color information retention while effectively reducing data size, ensuring enhanced image quality and smaller file sizes. This update brings a refined visual experience to AVIF and HEIF images across various applications and platforms. It also means that AVIF and HEIF derivatives have higher fidelity to the original image.https://techdocs.akamai.com/ivm/changelog/may-15-2024-avif-and-heif-image-formats-update-to-sharpyuv-chroma-subsampling
>>108850486But was it because of anime, or because of that retarded adaptation? Someone watched that netflix thing? I didn't, so no idea.
>>108850588Who fucking cares, the point is JXL can't hide behind the whole "only REAL images are important" excuse anymore. If decoding Anime images is problem for JXL then the devs better fix that pronto.
>>108850603I don't care about meme en(de)coder tech, can you let me derail in peace now?
>>108850546It's definitely NOT a meme for HEIC. Phones WILL jam up pike they're infected with a computer virus if they attempt to decode a 50MP HEIC and they don't contain HEIC hardware decoding.AVIF is newer than HEIC so wouldn't it be more CPU intensive to software decode?
>>108850584lol>>108850627Because it's proprietary garbage made to sell products? Newer doesn't have to mean slower. AV1 is basically as easy to decode as VP9 thanks to the effort of the open source community.
>>108850706That's a big extreme. Most real world photographs look just fine with 4:2:0 + sharpyuv when encoded in AVIF look good to me.More important they're like 90% as small as JXL images.
>>108847017Webpiss adoption is a google psyop
>>108851124AFAIK yes, Google will punish you for using 1MB JPGs instead of 700-800 KB WebPs on your website by downranking your SEO (ie you won't be on page #1).However being the devils advocate here: wouldn't the website using WebP images load faster on weak internet connections (ie phoneposters in a busy cell tower area dropping everyone to 1Mbps or slower)?
>>108850970Text is a very common use case. Imagine all the desktop threads that would suffer if they were forced to use chroma subsampling.You can gain similar compression efficiency by adjusting the chroma quantization accordingly.
>>108847226>I've heard that there's decoding performance concern on JXLmy nigger, nobody gives a fuck except phone manufacturersthey'll add the relevant hw acceleration pieces when it's popular enough, and meanwhile every single media platform out there will have saved 20-30% of storage spacedon't delude yourself into thinking:a) people would even notice, orb) people would think of blaming a site for it and not their phone/laptop manufacture, orc) that anybody gives a shit about people on old HW (by definition they are cheap, why would a company give a fuck about entertaining them?) more than they do about -30% to their monthly storage bill with 0 concerns
>>108847017I cant even run jxl on the newest iphonemost software doesnt support it eitherwhy would anyone bother?
>>108852112IF and ONLY IF what you say is true, then 90% of the entire internet would be using AV1 by now especially since it costs $0 in royalty fees. Video wastes exponentially more bandwidth than images do, by a mile.Meanwhile 4chan is using fucking H.264, which should be a grave insult to every single freetard worth his salt here.What the fuck is 4chan going to allow next, fucking HEIC?
>>108847120>what the fuck is keeping Webp popular right now despite being objectively worse than JXLCorruption at google from people in the webp team. End of story.The rest of the industry supports JXL and is strong-arming google into adopting it. It already has been reintroduced in chrome (behind an option) and will surely be enabled by default before too long.
>>108852275>the entire internet would be using AV1 by nownot comparable, because the dominant sources of video on the internet are:1. youtube: google has the money to care about device performance impacts, and the incentive as they also run android and look bad themselves if they don't handle it carefully2. netflix & shit: who have been using AV1 already https://netflixtechblog.com/av1-now-powering-30-of-netflix-streaming-02f592242d80also not comparable because video always had dogshit codec support problems, and the impact of software decoding is significantly more comparatively to images. an image might take a few ms more to decode, while the video might fail to render at target refresh rate altogetherimage and video are related and yet completely different beasts and don't belong in the same conversation at all>4chan is using fucking H.264>What the fuck is 4chan going to allow next, fucking HEIC?4chan is a piece of shit website tech wise and always has been. it's like asking why bestvideosjakarta.wordpress.com serves BMP filesthe answer is because it's old crusty shit nobody is caring for whatsoever, not because of some conscious decision of any kind
>>108852163>I cant even run jxl on the newest iphoneWrong.
>>108847017Is there a ready-to-use (easily pluggable, with permissive license) implementation in memory-safe language available?No?Well, there's you answer.Google has it for image formats they're using in their products (the safe language is not Rust btw).
>>108847017is graphs derived from filenaming or file magick number some jpeg are webm in jpeg name
>>108847017>Why did (competing standard) fail against (owner of monopoly of distributing standards)?We may never know...
>>108852994There is, and it already ships in chrome.>(the safe language is not Rust btw)Pic related. Google stipulated memory safe, and that's what they chose.I hate rust but you're deluding yourself to pretend it isn't a valid answer.
