JPEG XL

https://www.instagram.com/reel/CjTy3YVv08c/?igshid=YmMyMTA2M2Y=

https://brainwaltzera.bandcamp.com/track/itsame-group-hugg

<@794205442175402004> the website sidebar looks weird on chrome android.

Strange. Reported it.

But dead

J2K? On Safari we used it quite a lot, actually. Only recently it is becoming less relevant now webp and avif are also supported on Safari...

It's certainly not jxl, but between jpeg, j2k, j-xr, I'd rather choose webp, just because it compresses more

WebP cannot do 4:4:4 though

no full range color. Bad

Well internet use only

Even I, realizing this, started using jpeg xr q 95 8 years ago, although Now I think it would be better to use webp q100, and it's all the same that yuv420 is there

I have seen images where webp q100 has problematic banding that is visually obvious even without seeing the original

(this is not caused by 4:2:0, but by the minimum amount of color quantization that vp8 forces...)

Oh yea

Just yesterday I saw it looks disgusting

I created a picture in a neural network, its nature is such that it is noisy, I smoothed out the noise in another neural network and compressed it into a webp, and these stripes simply spoil the entire gardient background. I had to compress the original version with noise which visual smoothed out all the band, but the size is all 3 times larger

https://t.me/my_content/8096

webp is similar to avif in the regard that by design they were not meant to preserve any kind of noise at any quality level.

I'll assume you're talking about camera RAWs. the raw not a photo as much as it is a sensor dump that is yet to be transformed into a photo. raw is the whole grains before even roasting.

Avif and webp encoders mostly tend to smooth away subtle textures, like skin, wood, stone, textile, clouds etc.

High-contrast textures they will tend to keep better

lmao I swear last time I tried AV1 it wasn't that bad

though I was doing low bitrate and anime content

they told me it should be fixed now

Yes. Works fine

Maybe you did it at high speeds. My tests show that hevc with specific settings between slow and medium -- outperform av1 significantly in same time. The best thing I expected from her was the advantage of compressing the gameplay factorio at 360p. It's amazing. It's better than avc and av1 avc is better than hevc at all qualities lower than 720p (at the same speed). Except for this case

I've found HEVC to be specifically terrible at anime. It's as efficient as AVC in all resolutions while using more CPU time for both encode and decode. Haven't tested with faster presets, wouldn't be surprised if it was worse, the ringing in HEVC is very very visible in anime content AV1 on the other hand manages to outperform both at anime, though I haven't tested at the same speeds either. Generally I still stick to AVC for anime.

good ol' x264

I encode in H.264 rather than AV1 cause I like my encodes to finish

HEVC is better for anime because all HW decoders support 10-bit, but good quality requires some careful tuning

yes, maybe some first gen hw decoders did not support 10-bit, but still their number is pretty small, as well as the vast majority of modern mobile devices and tvs can decode 10-bit hevc (unlike 10-bit avc, which support is almost non-existent, even though it is also in the standard)

but hardware decoding for H.264 is also irrelevant

software decoders can handle it just fine

trying to discuss 10-bit anime on mobile devices is also a meme cause watching anime on a 6in screen is lol

yes, because it provides lower file size at the same quality

even for an 8-bit source scaled to 10-bit and then encoded

the higher internal precision lets the decoder quantize more

since 1 bit error out of 8 becomes 3 bits out of 10

to decode? it's still easily able to be decoded in software in realtime

to encode? yes much more

for HEVC it can be a bottleneck on older CPUs but not really for H.264

especially considering that anime has a low framerate

I think from 2015-2016 (after UHD BD standardization) 10-bit has basically been a requirement to support 10-bit helps to avoid banding without a huge bitrate boost (dithering/noise is also very bitrate hungry)

typically 24000/1001

perhaps the source was hdr or something like that (wrong tone mapping)?

hevc performs well in anime, surpassing avc by 5 or even 10 times with the same quality Now the question is - how did you compress?

HEVC is the absolute leader, if we are talking about content in which there are a lot of unchanged frames, anime, cartoons, factorio, gameplay of visual novels, this is all in which hevc has a very great advantage

But what probably won't tell you is that hevc is forbidden to use lower than SLOW preset, or specific independent presets, precisely because then hevc will bad even avc, both in quality and speed

I recently developed a preset exclusively for minecraft gameplay compression (from source or bitrate > 8000k), I did my research on all codecs, avc av1 hevc vp9, and only hevc was able to do it well and in an acceptable time.

Especially for viewing it in telegram, I got 5 hours of gameplay in 2 gigabytes, with avg 800-900 kbps But also on the up qualities it looks just great than all

AVM AV2

best frame

ffmpeg -i INPUT.mp4 -movflags +faststart -vf "scale=1280:720:flags=spline" -c:a copy -c:v libx265 -preset medium -bf 5 -refs 5 -g 600 -qp 35 -i_qfactor 1.5 -b_qfactor 2.0 -tune fastdecode -x265-params "lookahead-slices=0:me=star:rskip=0:tskip=1:frame-threads=4:pmode=0:pme=0:wpp=1:rc-lookahead=250:ctu=32:subme=3:merange=57:scenecut=0:cbqpoffs=-5:crqpoffs=-5:psy-rd=0:psy-rdoq=0:max-merge=5:weightp=1:rd=4:limit-refs=0" -f mp4 OUTPUT.mp4 qp 35 < 1000 kbps qp 30 < 2000 kbps qp 25 < 4000 kbps

and Example of a streams qp 35 https://t.me/my_content/8101

vp9 was worse than avc. av1 was worse than hevc at identical speed

any dinky mobile CPU can decode H.264 10-bit in software, and older hardware won't support HEVC decoding at all, so that's irrelevant

what the heck lol, so if you encode with faster presets you get better efficiency?

I've got gigabytes of minecraft gameplay on my storage so I'm definitely very interested to try that out

you use -qp rather than -crf? isn't that less efficient?

will try it out here... also if you happen to have an anime preset for me to try I'd love that

my CPU is a potato unfortunately and it's gonna take hours...

oh wait my work computer has a beefy CPU, time to use the free CPU time on it

can I access files via SSH in FFmpeg?

owo I can using SFTP

damn even my work computer isn't encoding very fast, only peaking at 0.5x realtime

(the source video is 60 FPS though)

looking at what's already encoded, yeah it looks quite impressive, apart from blurred textures which is to be expected most of the edges are well-preserved and there isn't as much ringing as I'd expect

resolution?

how do you develop presets? I used to be an absolute junkie at testing compression but the knowledge of specific x264/x265 parameters is kinda beyond me I know what most parameters are but can't quite grasp at how they affect encoding

1280x720 already

dammit I'm hitting thermal limits on my work laptop and the encoding is throttling T.T

still faster than my personal potato desktop

if you want speed, why not use hw encoder?

because I also want efficiency... also it's not like I want a ton of speed, I just wish it was at least realtime speed because that's a 11th gen i7

I expected more, that's all

hw decoders it is still much less power consumption and means less heat, noise, more battery life most of the encoding community follows hw decoder compatibility (like ref frame limit, buffers, etc.), 10-bit avc existed at first as some experiments, but after hevc became widespread they are almost gone also there is a lot of old hardware and tvs where it is impossible to update the software and 10bit avc is not supported there either

guess I'm really way behind of the encoding community's developments 😦

qp=36 looks impressive for the bitrate, pretty adequate for casual mobile watching, the quality reminds me of the average minecraft stream on Twitch qp=30 looks good overall, better than Youtube quality at the same overall bitrate

i'm more interested in high fidelity encoding and the difference here is only about 10-25%, maybe more for 4k resolution (but avc is also rarely encoded in 4k), for low bitrates the difference may also be more

do you have a test clip? i want to try encoding with those settings compared to hw encoding

I can send you some cut footage from my archives... but I'll need some file sharing website I can use for that, preferably one I don't need an account for

or you could send me a test clip and I'll encode it here and you encode it on your side and we compare

<https://wetransfer.com/> should work

might have to change this setting if it defaults to email transfer

okay I'll prepare a clip here... think I'll pick an ender dragon fight I recorded, is that okay?

(it has a funny moment in it lmao)

yeah, that's fine 👍

what's the file size limit?

for that website

2gb i think

ended up only being 200 MiB, it's a 6-minute clip

uploading

this one is only 960x540 resolution however, hope that's okay... it is high bitrate/quality though

I'm gonna encode with QP=35 okay? to really stress the encoder

uploading 10%

I'm on a slow proxy so bear with me

70%

https://we.tl/t-qe27zsxEH8 here's the raw clip for you to encode

and here's what I just encoded: https://we.tl/t-0MUlil1LkM (sorry I don't have Nitro)

omg the end of that clip is so funny lmao

software encoding with the settings dzgas posted: around 75fps 29.1mb, hardware encoding: >450fps 31.77mb image comparison: <https://slow.pics/c/UlKrNSLv> software file: <https://mega.nz/file/Y7RhyQqA#pmpk-tp2RAud-wD1JkwRS8cfWIQcsGl24MwtcOcVecA> hardware file: <https://mega.nz/file/17YGxZSa#4B25LjsoExdZPmnb92XCgtyjDuSyrYimqA0Wbojg4kA>

would be nice if someone had 1080p or 4k uncompressed footage

welp I could record something at 1080p, just idk what to record

I usually don't record straight 1080p to keep the file sizes down a bit

normally when I record I use hardware H.264 with QP=20

which is pretty good quality especially if we're gonna be reencoding for low bitrates

my graphics card doesn't do hardware H.265 encoding so that's not an option

Capture game play and store in losslessly on ram and encode with h265.

I think your hardware version actually looks better

perhaps blurrier, but with less ringing definitely

I have 64gb ram now

I only have 12

and right now 10 GiB in use

because I leave a lot of stuff open

and because I forgot Minecraft open overnight

and Minecraft takes friggin 5 GiB of RAM

damn there's over 8 GiB of swap usage now

h264 lossless isn't that big

I never liked playing with mods that much, except for server plugins

I play vanilla... ish, because server plugins

I usually play with friends, either on big servers or small self-hosted ones

and there's of course /home and /tpa plugins because big servers have it and everybody asks me to put it on the private servers I host for us to play

haven't played in small servers in a long time tho

I'm gonna try my own approach at encoding the file, in software, but just using x265 defaults and see how that looks in comparison

nah I played yesterday lol, in a large server, but didn't do much

just cracked the world seed and went mining for iron ore veins lmao

got a shulker full of raw iron blocks in about 4-5 hours

just for the meme

because I don't even like that server very much and iron isn't even worth a lot there

last time I cracked the seed on the same server I went into every ancient city and raided all of the chests lmao just so nobody could find swift sneak books anymore...

but then I was dumb and somebody stole the books from me anyway oh well lmao

wow are those mods still being updated?

some people really like older versions tho

and people with really old low-spec computers are stuck in versions 1.7.10 and below, too

there are like PVP servers for those versions just because they're way less laggy and some people still use old computers to play, and they're rather popular

because otherwise a lot of servers in current versions use plugins to emulate the PVP experience from version 1.8

(no cooldown, decreased hit delay, sometimes the god apples give regen 4 rather than a ton of absorption hearts, but axe damage and sword sweep are retained because people like those things)

heck some people are still playing on Minecraft Beta 1.6, which is the last version before hunger was introduced

and some libre Minecraft server reimplementations out there aim for Beta 1.6 too

I'm not into mobile gaming and not into the mobile gaming community so idk for sure but at least from my point of view online PC gaming is as popular as it ever was

like, sure mobile games are more popular, but it seems like it's a new public, it's not like PC gaming has decreased in popularity

newer versions are hell for older machines

they use way more RAM and demand way more CPU power

it's especially the CPU usage in Minecraft that's bad, has always been kinda the bottleneck but nowadays it's especially serious

since 1.13 or so the CPU usage has creeped up significantly with the switch from integer block IDs to paletted namespaced NBT data

I miss the days when it was possible to run the minecraft server alongside a client to play with your friends

since 1.13 I've had to use a separate computer for that

because otherwise it lags like hell

and servers can't take high player counts on these newer versions

nobody seems to know why but the CPU usage goes up in a very steep curve with the number of players online, and beyond 150 or so it becomes unbearably laggy even on the most powerful hardware available

on 1.12 and before, this wasn't as much of an issue, a server running on cutting edge hardware could easily tank 500+ online players

I tried doing a couple of CRF encodes rather than QP, I think the quality is better but I'll have to compare carefully later

there is also svt-av1 with fast presets like 8-10, the more cpu cores, the more efficient in quality/speed than x264 and x265, but on streaming quality https://stolen.shoes/embedVideo?video=https://files.catbox.moe/3rj0sw.mp4&image=https://files.catbox.moe/zesb1t.jpg

will compare that one too

https://files.catbox.moe/3rj0sw.mp4

yeah already downloading here

sucks that in mpv, when cycling through video streams (shortcut: `_`), one of the cycled options is no video stream and it closes the window

can I disable that behavior?

it sucks for doing A/B comparisons

or guess I could instead stack the windows and alt-tab between them

and sync them with Syncplay perhaps

meh I'll do the comparison later

What? Where did I say that?

looks good, what settings did you use?

oops sorry I misinterpreted your message

you meant to say that for good efficiency I should stick to at least `-preset slow`, preferably `slower` or `veryslow` if feasible?

OK, use my preset at qp 25. Well, or you can remove the 720 parameter and do it in 1080p, then it's better to take qp 30, merange=92 and ctu=64

Yea

the footage is all in either 720p or 540p so no point in doing 1080p

I tested your preset but using -crf 30 rather than -qp 35 and the quality seems better with approximately the same file size

still gonna do a more careful comparison later

Well, at 540p the efficiency drops significantly. Due to the fact that the total length of the vector and the shift decreases

just a preset (8-10), nothing more

but if I upscale the video, will I get more objective quality for the same file size?

I don't like how crf works. But in exceptional cases, I use it, in fact, only in one, with exceptional frame statics, for example, for rendering a Chat from a twitch

you can use -qpmin to limit how much crf ends up varying the qp, too

I think it was -qmin?

without specifying which encoder it is

meh

I had that in my personal preset for compressing anime for watching on the bus, though that was H.264

every second one is x265

Dozens of times reading documentation and independent experiments on thousands of examples, I've been doing this for at least 5 years, almost every day, I'm crazy

I think this was my anime encode preset for mobile: `-codec:v libx264 -pix_fmt yuv420p10le -preset slower -tune animation -crf 30 -qmin 24 -aq-mode 3 -s 1280x720`

I've been there before but never got too deep into cherry-picking x264/x265 options

If you want to do this, then you probably need to experiment with a more complex method of calculating vectors

Maybe rd=6 maybe subme=3,4,5,6,7 maybe me=star,full

My most important mistake 3 years ago was thinking - everything new is better than the old

Why 10 bit, for anime?

doesn't 10bit reduce banding and stuff?

Damn, it was so funny about 7 years ago, there were 10 bits of hevc on the torrents, but the hardware decoders were only 8 bits, that's hilarious

because "I heard it gave better compression efficiency".

Mm

Not heard about this

not if your source is 8-bit to begin with lol... but I heard 10-bit was better because it gave better precision for the DCT coefficients and that could yield slightly better quality at high QP

<https://yukisubs.files.wordpress.com/2016/10/why_does_10bit_save_bandwidth_-_ateme.pdf>

x265 took many years to be better than x264, there are proprietary hevc encoders which are better than x265 but they are not available for personal use

av1 for strong compression can no longer be used without 10 bits, av1 have are very good gradient noise smoothing functions that look disgusting at 8 bits

???

See if you take 2 pixels of 8 bits each, and then combine their values 50/50. Then you will get a non-8 bit value

this means that it makes sense to save the original 8-bit picture into a 10-bit picture when resizing, downscale

x265 is better than x264. But only if the frame size is up than 720p.

Even if the progress of x264 stopped at 960p resolution. (In my opinion.) You can always start using a complex tesa me-range=64 or even yuv444

anything

I myself compress only avc in telegram, because it is more profitable than hevc, even sometimes much better quality, but always faster

movie and anime

for medium quality and streaming i will probably be using av1 mostly when support becomes more widely available since there are already fast encoders like svt-av1 (on a cpu with a large cache and lots of cores, it can be faster than x264 with better quality) and also hw encoders but for high quality it is very unstable and slow

If it's not a problem for you to encode av1 in real time on parameters 8 and below, then there's nothing more to talk about

svt-av1 10-bit https://files.catbox.moe/qqisb6.mp4

AV1 has no disadvantages, it is the best in everything, if you do not take into account the speed

https://encode.su/threads/3953-Video-Codecs-user-experience

doesn't open for me :/

vpn?

I'm on a Tor proxy

<:Thonk:805904896879493180>

x265 10-bit (at or near real-time encoding speeds, depending on the cpu) https://files.catbox.moe/ogjkfl.mp4

vs. ^

is this my present vs svt-av1?

probably, didn't check how it was encoded and it's vs x265 10-bit (with my settings) https://discord.com/channels/794206087879852103/806898911091753051/1047190344871186502

well, look, this is important, because according to my tests, both AV1 and SVT-AV1 lose VERY much at identical encoding speeds vs hevc

and firefox doesn't support AV1 inside mp4, why don't you use webm

it's hevc, not av1

ok

hevc vs av1

the last one is hevc vs hevc (with my settings)

hevc vs av1 same encode time

look to the wooden posts that are blurred by the movement of the blocks (av1)

you can write your settings

comparation with my hevc preset. for av1 used SVT-AV1 speed 11 they have an identical encoding time

unfortunately, AV1 is not suitable for GOOD compressing streams on the processor, for some real period of time

yes, but it depends on the cpu, the more threads the better for svt-av1 because it scales almost linearly with the number of threads and therefore it is possible to use slower presets (x264/x265 benefit much less from this)

I'll try encoding with these settings

HAH

but after all, x265 is the most parallelized codec in general, when I first studied it, I was shocked by how many technologies were used there for this

and also, few people talk about it for some reason, but x264 parallelizes by default and without problems into 35 Thread, and this is due to the fact that AVC is a very old codec and can be split into such a number of streams without affecting the image quality.

while AV1 is the technological successor of VP9 in parallelization techniques (bad). But it's worth saying that they tried very hard to make it work somehow

libaom yes, but svt-av1 scales linearly even on multiprocessor 96+ cores cpu x264/x265 have their limits, especially for slow settings

in any case, all this has led to the fact that the best way to compress a One file is to divide it into 20 parts and compress each on one core, then combine everything. unthinkable

this is a good way to do it and there are tools available for this like av1an, but it is not useful for streaming

whoa?

yes, but these are old tests, svt-av1 has significantly reduced memory consumption in the new versions

well, you didn't write about memory, but about cores? you mistake?

<@1034873369314730065>I don't understand you. I just want to say that nothing parallels, this is the *scourge* of av1

given the **monolithic **AV1 structure, as well as VP9, parallelization is possible only by physically dividing the image into parts, this also kills the quality

maybe, although back then there was no 64c/128t on the single cpu like now and maybe I was looking at other comparisons, I mean that svt-av1 has no limits (except resolution) on scalability

but not so much, it's quite possible to do it on 4 streams without problems, but who do you have to be to be able to stream like that? with good preset

I only agree that such parallelism can be achieved on image sizes like 8k by 8k.

I like how AOMENC compresses 360p at 1-2 speeds, it's really a beautiful sight, no one has been able to beat AVC at this resolution since its introduction

parallelization for a single frame is bad for VODs, when it is possible to do it offline and get any number of frames, but for live streaming it does not so much hurt the quality (for av1)

i retested even on my oldest pc with i7 4/8, and these settings are slower than svt-av1 -preset 8, something like 39-41 fps vs 58-62 fps on average https://files.catbox.moe/97fxfc.mp4

encoding with my hevc settings is also faster

x265 for ffmpeg `-vcodec libx265 -preset slow -tune ssim -x265-params me=hex:subme=2:max-merge=2` + `-crf/-qp` i prefer `-crf` + `-threads 0` or needed threads `-pix_fmt yuv420p10le` for 10-bit (a bit slower) they may look "weird", but they are a pretty universal and balanced speed/quality settings (real time on modern cpus), though they could be better tuned for certain content they are also for medium to low quality (not really good for high quality)

svt-av1, tune visual quality (less blurring) https://files.catbox.moe/d4kfcg.mp4

me=hex and max-merge=2 is very bad for minecraft gameplay

I conducted tests with ssim and for me it showed worse results than the default HEVC metrics

these parameters are simply disgusting, you use the SLOW preset that activates many functions that are completely ineffective for minecraft gameplay, But at the same time you choose the worst vector mode

well, speed 8 is already better, obviously better than 11, but how is such a big speed space possible??

of course, if it could work for me at a speed of 8 at the same time as HEVC, I would use it (as soon as its support appears), .... but this... is not

no no you not oldest, i am AMD Athlon(tm) II X4 640 Processor 3.01 GHz

these are not just random settings (they are selected and a bit simplified in terms of speed/quality ratio from a very large number of other parameters), although they are not fine-tuned for something, they are more or less universal and also the results can always be compared to something else and make sure that it will be better (or worse), better visually, metrics is only as some kind of assistant (which is not so difficult to cheat, without real visual improvement)

simpleton

probably different cpu, different versions (I use the latest master build compiled with flags for my cpu) or something else

do you even know how many parameters x265 has and how much each contributes

these are not even all of the parameters

you have the bigger screen?

i mean the oldest cpu that i have, it's about 9 years old or something like that

`x265 --fullhelp`

very funny

to make it easier to compare, here are all the encoded results: https://files.catbox.moe/3rj0sw.mp4 svt-av1 8-bit psnr https://files.catbox.moe/qqisb6.mp4 svt-av1 10-bit psnr https://files.catbox.moe/d4kfcg.mp4 svt-av1 10-bit vq https://files.catbox.moe/ogjkfl.mp4 x265 10-bit (my settings <https://discord.com/channels/794206087879852103/806898911091753051/1047254954500436008>) https://files.catbox.moe/97fxfc.mp4 x265 (DZgas settings <https://discord.com/channels/794206087879852103/806898911091753051/1047088747449167933>)

subjectively, in fact, it was important for me to make exactly 720 and precisely to upload it to telegram, so that every smartphone 6 years ago could play it (because the video is at 720p 60 fps), therefore, for a more honest comparison of presets, you need to remove -tune fastdecode

and of course No 10 bit

yes, when already add av1 support AT LEAST SOMEWHERE, otherwise it's already funny, no twitter, no reddit, no discord, no telegram, SOMEONE HAS ALREADY HELLO

6 year old smartphones, if I remember correctly, had hw hevc, so `-tune fastdecode` is not needed, just need to have limits for ref and maybe a few other settings and the formats support in telegram is also quite weird, on some platforms av1 is supported, but on others is not

this is the support of purely client-creators. that is, not official, I agree with HW. and yet I'm of the opinion that you don't have to have HW for playback of anything

for the same reason, I will adjust the AV1 to the performance of my Samsung j1 2016, which I have in my collection

my job hunt this year is now over, and I'm quite happy with the outcome. alas, it is not related to jxl (or whatever sort of image processing) 😉

Congrats! What's your new job?

Presumably (cause I applied to multiple JDs and it's not yet decided where I would go) an NPU compiler optimization and stuffs.

Congrats!

hopefully (not guaranteed though), and I already have a backlog of changes in the background

J40 does require a higher level of concentration that I can't easily retain, which is the main reason I had little visible progress for a while

noice

it'll be cool to have a third decoder to compare things to

for the more obscure features

latest update

Not true.

AV1 encoders have multiple ways to thread: row threading, tile threading, frame threading, as well as slice threading for other processes.

https://twitter.com/mikjcal/status/1597164383075196928?t=3IF63OzEWHzr9qZavArdDA&s=19

on the possibility of deprecating webp as like flash (from <#822105409312653333> ):

I think flash is actually an edge case, where Apple was strongly involved

Yes, thinking back now, I agree that Apple played a major role in it. (And didn't really play it fair.)

flash was, with caveats like NPAPI dependence etc., much more mature than the contemporary web technologies

adobe could have made flash a de jure standard if they wanted

flash was never a web standard and wasn't really part of browsers (although it became very popular and was often enabled by default as a plugin because there were no other alternatives at the time), it was actually a standalone ecosystem within a standalone app

they were however reluctant, and sidetracked with AIR etc., giving Apple a reason for outright deprecation

you are right, but it was not a far-fetched idea to make it blend to the web ecosystem

at some point you had no choice but using flash as a proxy to more advanced features that web browsers simply didn't have

video playback was a prime example, but there are tons of other examples

it took a whole *decade* for web technologies to catch up with what flash provided

yeah, I mean it's easier to remove than a real web standard or formats in browsers, the same thing happened for example with silverlight or javafx

I'm not very sure about that, for example you can reasonably argue that XHR was also not a web standard at least initially because it was originally an ActiveX plugin

but web browsers eventually standardized XHR but not flash

I know XHR is much smaller than flash though

another problem is that flash was proprietary and closed and there were some alternatives, it could in theory become a standard or fully open, but then it would not be the same flash we are talking about, it would be a real web standard without additional third-party plugins (which is actually what happened over time with the most needed features)

same happens with web game engines, at first it was only third-party plugins, but now a lot of functionality is ported to web standards such as webgl

i think because adopting and approving standards originally took a very long time and did not keep up with the real fast growth and needs for the web, later it was accelerated, perhaps because the role of standards development and promotion has mostly moved to large corporations

about formats, if I remember correctly, none of the formats that have been added to browsers have ever been removed but many of the very popular formats that were only in third-party plugins are now gone, like realvideo/audio, vp1-7, wmv/wma, etc.

https://en.wikipedia.org/wiki/X_BitMap is a counterexample.

yeah, I didn't specify that some might have been removed that were originally very niche and didn't gain much popularity theoretically Jpeg2k could be removed because it was not supported by other browsers, also Jpeg XR was only supported by Microsoft and was not widely used and supported (I don't remember, maybe it was already removed) but for example vp8 is actually a transitional format, originally needed to avoid (possible) fees for avc (for online video), it could also have been removed after vp9, but vp8 was added and used in webrtc as standard (and other browsers added support for vp8 in webrtc pretty quickly), it is also used in webp (which also could have been skipped after more modern formats were added, but other browsers also added support after many years even when next-generation formats were already or almost ready) and after all that, removing them is fairly unlikely anytime soon

I hope at some point BMP can be removed as a web codec

J2K and JXR are easily removed since they never got universal support, so there should always be fallbacks for these since otherwise the page would only work in Safari or Edge, respectively

what if they added jxl into bmp

Lol

I've seen somewhere that jpeg xl could be added to pdf at some point in future, is that still true?

I remember people related to the PDF standard talking about that at some point, no idea what the timeline on that would be though.

I doubt it can be removed without breaking some 1990s era websites that are never going to be updated, but it would be interesting to see if browsers could start shipping certain "deprecated" features as WASM modules for a bit of seperation from the core code.

yeah, I think in the future the wasm implementation for lightweight or very rarely used formats is quite possible

there's minimal reason to remove BMP though because it's such a primitive format

the overhead of maintaining it is very low

<@416586441058025472> Opinion on comic sans?

idk about primitive, bmp can contain PNG and jpeg

what now

I think JXR has already been removed, Edge dropped support for it a few years back (With the transition to chromium)

Indeed. It's only left in older versions, which still are being used a little.

Can it? Aren't you confusing with ICO?

In any case BMP is kind of a mess

i may have said it before, but the wikipedia page is really good <https://en.wikipedia.org/wiki/BMP_file_format#File_structure>

It's basically a memdump of internal Windows representations of images across a few decades of how Windows has represented images, right?

I assume only some kinds of BMP images will actually work in all browsers

Anyway if you want to "clean up" image formats in browsers, removing BMP would be where I would start. I cannot imagine it has much usage in the wild at all, and there is absolutely no reason why it would ever be a good idea to use it on the web.

well, that's how pdf works in firefox

https://youtube.com/shorts/G2fnBl9QUhs?feature=share

I am not confused, though

https://www.youtube.com/watch?v=MVwl3tRzjbg&list=OLAK5uy_nLnXok_OoxNB01xgj8z1SVH2tISL64uv8

Why did Mozilla fire all of the Servo developers?

They could have made a new web browser and obliterated Chrome.

And since money is the only thing Mozilla cares about these days, they could be making more money after taking back market share from Chrome

Even if you're a complete sellout that only cares about money, it's just a stupid move to fire everyone working on Servo. Servo is something that would most immediately and directly help Mozilla's bottom line.

most people don't know that many components developed for Servo eventually made into Firefox anyway

Well if they didn't fire the developers they could have replaced even more of Gecko and made it much faster than Chrome.

And actually attracted attention and boosted market share

they could have replaced certain parts of Gecko with Servo components exactly because *Servo is intentionally made so*

in fact at this point Gecko is actually faster than Chrome in many aspects, contrary to public expectations (for example, JS performance)

I find chrome perform better than Firefox especially on js websites. Firefox is laggy.

Same

Firefox has been noticeably slower and jankier for a long time and Servo has the potential to completely obliterate Chromium

It was rapidly heading in that direction because of its multithreaded architecture

But Mozilla put the kibosh on that

Literally fired everyone working on it

it took quite a while, but nowadays I don't have that kind of problems for most websites (it used to require restart for youtube, for example, but it is no longer necessary at least for me)

I have actually used Servo for multiple times, and it never reached the point where I can actually build something on it. it would have taken a lot of efforts to make it a production-grade browser.

to my knowledge the last remaining piece of puzzle in Gecko's performance is a networking stack

On recent Firefox updates, I find Firefox very unreliable especially on YouTube, sound goes off when I pause and play. I have to pause and play two times to get sound working. Sometimes vp9 decoding stops working too.

everything else has been revamped over the last decade

I have used Firefox in both Windows and macOS for a long time (>10y) and it's substantially better than it used to be

The only thing which keeps me using Firefox is, it has nice and smooth scrolling performance on touchpad on Wayland.

does Chromium work well with Wayland then?

Scrolling is not good on chrome/chromium on Wayland

then I guess it's a Wayland problem (or more accurately speaking, a developer problem with Wayland support) and not a Firefox problem

No. There is a merge request on chromium, it is not merged/implemented yet

platforms with lesser support tend to drift in that way, generally speaking

anyway your mileage may vary, but I can speak with a confidence because I do open up to 1,000 tabs in both browsers combined

they have very different failure modes at this scale, but I found Chrome is a lot worse than Firefox in this regard, and actually it has gotten worse in recent years

Hmm. For me chrome/chromium is very stable and more responsive.

(namely: some tabs spontaneously fail to respond, with no reason, and the only solution is to open a builtin process manager and kill affected processes)

in the casual usage (i.e. not that many tabs open) I see no practical differences between them, except for some cutting-edge features where developers are not hesitant to say they know no browsers besides from Chrome 😦

and when I stretch browsers to their limits Chrome tended to break much earlier than Firefox

Probably you have a high performance cpu and won't notice the difference.

I also use Firefox in my now 8-year-old laptop 😉

I agree that browsers in general demand more computation power than they used to do, but Chrome is no different in this regard, at least for me

Okay. do you use self compiled Firefox or the precompiled one from distro?

I don't compile Firefox from a source and I see no advantage in that (I know this statement will anger some Linux distro users, I know)

that laptop is, for the reference, an Early 2015 MacBook Air

and I think I can't compile Firefox in that machine

FWIW, I just switched back to Firefox from Chrome, and there really isn't much difference. On Windows, the most noticeable thing is font rendering, but other than that my experience has been mostly the same.

Of course, Google apps (Docs, Meet, etc.) work a lot better on Chrome than Firefox, so that's what kept me from switching for a while.

it turns out the font rendering is because chrome doesnt respect cleartype settings

it seems edge has fixed this recently

another main reason i stick to firefox is the smooth scrolling behaviour/settings

microsoft's impulse scrolling in chromium helps a bit but it's still not as good as the old edgeview scrolling

oh yeah and chrome's image scaling is still abysmal

What was your experience?

pretty much same with many experimental browser engines. works for a few selected sites, but does not work well for others.

for instance it was pretty much unusable for Korean websites

Firefox is also much better if you use ublock origin to block adverts

they wrote a blog about it

Will OpenCV support JPEG XL?

that ultimately depends on if someone writes an image reader to decode JXL for open cv

that said, you can always have a separate decoder and feed OpenCV an uncompressed image

thanks

Firefox ignores some as well, forces GDI rendering for some fonts and ignores the system contrast value if it's set to the default setting

I hate modern GDI rendering, so sluggish, it's crazy

I wish more stuff used Direct2D, Lots of stuff seemingly moved to Cairo (And then Skia, like Firefox for a while) which gives better quality than GDI, but still not amazing performance

I was looking at the issue tracker for .NET "winforms", which uses GDI/GDI+ for all custom rendering, and there's an issue about replacing that with something newer

The vast majority of comments were suggesting Skia, using the OpenGL backend over a Google translation layer

And then creating a home grown remoting solution

if only hardware GDI wasn't killed off in windows vista with the new driver model

While Direct2D is a core part of Windows and is built into RDP so it already handles all remoting

GDI was actually quite fast in windows xp

just try resizing the columns in services.msc in an xp VM compared to windows 10 and you'll see lol

iirc they mostly offset that with 7 (re-adding a lot of the old support), but it still doesn't play well with things like the DWM

didn't seem to really do anything from what I saw

also we do have FreeType

I have heard reports that microsoft broke GDI acceleration in some version of windows 10

and that it just silently falls back to software

oh yeah that as well

gotta love modern windows

Wouldn't surprise me in the slightest, also wouldn't surprise me if the GPU side atrophied as well

I remember reading some blog post years ago from a MS dev that went into some detail

Like GDI having specific pixel coverage functions that can't be changed

While something like Direct2D just relies on vertex sampling on the GPU with normal supersampling

rip

<@794205442175402004>fun thing about that Twitter compression post you commented on, you can actually get decent quality images on Twitter

https://twitter.com/A_Fox_Wizard/status/1601505813117698048/photo/1 (only you can view it on Twitter, because I don't want my only posts to just be some test images, so it's hidden for everyone else <:KekDog:892418529117233162>) this link will work for everyone else: <https://pbs.twimg.com/media/FjmwpTkXoAEfLu_?format=jpg&name=4096x4096>

Think the criteria for Twitter not to compress your jpegs are: the resolution can't be higher than 4096 pixels and I think the file size can't be higher than 5MB

It could be something like that, yes.

Isn't freetype library for fonts?

yeah

In Dutch and French the expression is "the ball is in your camp" while in English it is "the ball is in your court".

https://www.riffusion.com/about

https://www.theverge.com/2022/12/16/23511461/apple-google-mozilla-chrome-safari-firefox-browser-speedtest-3

> It’s nice to see everyone working together to make the web better.

hah

> Mozilla says that a benchmark built by several web companies will help provide a “shared understanding of what matters.” The company says that’s important when you’re trying to coordinate across web developers, standards bodies, the groups that build the engines that interpret code based on those standards, and the companies that build browsers based around those engines. > As Google points out in its Twitter thread, though, the companies have set up rules that should help prevent any of them from trying to tip the results in their favor. Nontrivial changes will require approval from “at least two of the participating browser projects” and can’t be implemented if there are strong objections from others, and major changes require a consensus from everyone involved, according to the governance policy.

For some reason, the Chrome team only seems to communicate around jxl on Fridays. Could be there's some reason for that, like maybe they have a team meeting then or something, no idea. But it does seem to be quite consistently that if they communicate on jxl, it's on a Friday.

This is not ideal to have a conversation, because Friday in the US is already weekend in Europe. I always get to see it on my Saturday morning: the addition of the removal note (Oct 28), the "we'll work on publishing data" message (Nov 11), the actual publishing of the data (Dec 2), the removal of jxl (Dec 9) and now the updated decode speed data and the change to `wontfix` (Dec 16). Always on a Friday.

The "conspiracy theory" interpretation of that could be that they're scheduling it so your replies, as the main public proponent of JPEG XL, are delayed somewhat so they can save face for a just a bit longer. Of course, it's probably the former. Some sort of scheduled meeting or agreement.

my take on it is that it's just normal "oh let me do this before I end my work week" stuff

Here in the Netherlands, we usually do a "VrijMiBo" (<https://en.wiktionary.org/wiki/vrijmibo>) on Friday. Perhaps the Chrome team is doing something similar? If so, perhaps too many alcoholic drinks are involved (i.e. the "Bo" part of that word)?

Chrome you're drunk

Or maybe rather "I don't want to deal with this now" for most of the week and then on Friday "right, this still needs to be done, sigh". But yeah it's likely something like that.

sounds like an "agile method" organization with a sprint end every friday 😄

recurring "prove avif's superiority against competing image codecs" ticket

PSA: if anyone here has joined Mastodon, you may want to use FediFinder to find the Mastodon accounts of those you follow on Twitter ASAP, as Twitter is now banning links to other social media platforms. (<https://fedifinder.glitch.me/>, <https://fedifinder-backup.glitch.me/>) Also, if you get the archive from <https://twitter.com/settings/download_your_data>, you can this tool <https://tinysubversions.com/twitter-archive/make-your-own/> to make a searchable archive of your tweets.

chatGPT is real deal

lol

I have no idea what kind of text, it needs to be searched somewhere deep in the archives of 4chan, to tell the truth - to generate something beautiful myself is a very big deal, so I just take other people's pictures, well, or for example I made this apple myself

although I have another photo from that post

I don't remember, there were a lot of posts from SD, it seems like 5 boards, in any case I can only say that now it's lost

tried to make my youtube channel slightly nicer: https://www.youtube.com/@JonSneyers

still very amateurish youtube channel, obviously

whew, the book I’m reading doesn’t always mince its words: > I offer the following simple and consistent explanation for the persistent irrational belief that natural remedies are safe: taking them rots your brain, gradually destroying the capacity for logical thought.

what is "natural remedy" in this context?

prolly something in lines of alcohol or natural drugs

something like daturas can be used as pain-relief medicine too sorta but these are example of something that would definitely rot your memory stuff if used persistently

https://www.youtube.com/watch?v=jvPPXbo87ds nice sequel to the earlier Bézier video

good intro to jxl's Catmull-Rom splines (only doesn't go into detail about centripetal)

we really need to make a "jxl brush" for krita or inkscape or something, to draw with splines exactly representable in jxl so you can save the image as a crazy small jxl file

(also we need to better define what the maximum pixel area of splines is in each profile/level, in a way that doesn't cripple how this could be used in the future, but also doesn't cause DoS issues)

I think it was a general statement about remedies that are “natural” (often herbs, essential oils, this sort of thing)

same as in https://sciencebasedmedicine.org/herbal-medicine-and-aristolochic-acid-nephropathy/

> It has been a stunning triumph of marketing and propaganda that many people believe that treatments that are “natural” are somehow magically safe and effective (an error in logic known as the naturalistic fallacy). There is now widespread belief that herbal remedies are not drugs or chemicals because they are natural.

> Herbal remedies are drugs, plain and simple. They contain chemicals that are ingested on a regular basis for their pharmacological effects. The fact that they derive from plants is irrelevant. The fact that individual chemicals are not purified and given in precise amounts does not mean they are not pharmacologically active chemicals – it just means that when taking an herbal remedy you are getting a mixture of many chemicals in unknown doses.

in a similar vein, Steven Novella (SBM link above) also wrote https://sciencebasedmedicine.org/how-safe-is-cbd/

the book I quoted for is by another author (the same as https://www.researchgate.net/publication/38415685_Three_things_every_medical_writer_should_know_about_statistics)

instead of medicine, I see

so the usual hoax remedies

Update

https://youtu.be/LQdjmGimh04

🛫

https://github.com/fabiorug/Riliu/releases/tag/1.0

Actually that 64gb ram computer is the one which I use it workplace. I use that to compile software. At home I use a laptop which is only 4gb 😶.

Works fine for me. I use discord and 6 tabs on Firefox, few background programs. All work fine on 4gb ram.

I have 3gb zram too.

i thought KDE didn't use that much ram

hmm https://www.reddit.com/r/xfce/comments/kb0d87/i_compared_the_ram_use_of_15_desktop_environments/

wtf

One time I wanted to convert a large image to jxl, but kept running out of memory (16gb physical + 4gb zram), I ended up setting zram to 48gb/zstd and managed to successfully encode the image in about 10 minutes, peak memory usage of cjxl was around 50gb. The experience was weird and magical in some sense, it shouldn't be possible to "download more ram" like this and yet...

14442x21564 It was one of these tiffs: https://archive.org/download/unix-magic

Do you know zram size can be changed?

On fedora default install it was 4gb. After installing opensuse I chose to use 3gb.

I don't understand what kind of zram you invented

Did I say I invent zram? 🤔

<:This:805404376658739230>

Bruh

lol zram is a linux' compressed memory storage thing

https://www.reddit.com/r/ProgrammerHumor/comments/ztzpsj/for_the_programmers_who_use_tar/?utm_source=share&utm_medium=mweb

tar caf // create autocompress file tar xf // extract file This is all I need 99% of the time

<@557099078337560596> From #adoption: > this suddenly reminds me of the old OpenBSD ports controversy with Pale Moon... Admittedly it could've been handled A LOT BETTER from the Pale Moon side

On the other hand I'm not a fan of the OpenBSD dev insisting on getting the word from Moonchild himself despite being clear that the build configuration is not acceptable for Pale Moon

Nor the removal of the browser (not even an unbranded version) from ports

And it's really weird that people insist on using system libs as if it's a silver bullet for security (despite debunking: https://forum.palemoon.org/viewtopic.php?f=5&t=23706)

Nice Xmas doggo

Heh, thanks

I mean, it's not wrong...

https://www.youtube.com/watch?v=a_pMd409bhM

oh yea

https://www.vidlii.com/watch?v=cvcpyD-W1DW

What are you going to do with them

nothing, it's not even my photo, it's just a very funny reminder of what happens without strict standardization. the same could be said about smartphone accumulators