JPEG XL

Pineapple spider

Panda spider

Turtle spider

Turtle pig

Monkey turtle lol

Cat pig

https://youtube.com/shorts/7FOn6EGhPo4?feature=share <@416586441058025472>

what

for a little more context, see https://youtu.be/czTksCF6X8Y which is itself derived from https://youtu.be/XW0W7j04iRQ

and for more “what”, see https://youtu.be/lpvT-Fciu-4

I'm... not sure what that is...

it’s modern

*what*

https://twitter.com/mcmillen/status/1434632359865995271?s=19

Could a proof-of-stake, energy-friendly version of bitcoin be made to work? Assuming magical initialization conditions (say 8 billion NFTs get assigned, 1 to every human, and that's it, no more tokens can be created, there can only be transactions), could there be a way to get a decentralized currency without the energy footprint?

Asking <@799692065771749416>

(but others are free to answer too)

7x energy isn't necessarily 7x compute power... i wonder how bitcoin and google compare in terms of efficiency (flops per watt).

<@794205442175402004> Not with anything resembling the current security properties. PoS has sort of a circular reasoning problem, where you use the currency defined _by_ the chain to determine which chain is valid. The problem is that you need a mechanism for everyone to unambiguously tell which of two conflicting, otherwise valid, chains is the correct one to (at least temporarily) accept. With PoS, you assign voting rights to staked coins. But if those staked coins existed in the chain before the point of forking you're trying to resolve, you get the problem that those coins now exist on both sides of the chain.

<@!164917458182864897> Bitcoin uses exactly 0 FLOPs.

<@!794205442175402004> There are proposed solutions to this so-called "nothing at stake" problem, but they all eventually fall back to relying on making things slow enough that humans can intervene.

Energy is what matters imo, compute power is irrelevant since it's not really useful compute anyway (only indirectly useful)

(That's excluding more philosophical concerns about whether or not you want the largest holders of the currency to be the one that hold all power of censorship over it - at least with PoW that power is held by those able to spend resources in the real world, which is far less influencable than that of just the currency itself)

I think an energy friendly way to create a decentralized currency would be great progress.

I think Chia's proof-of-space-and-time approach is a much more acceptable approach for a lower-power distributed consensus mechanism, though it also has problems. It relies on wasting another real-world asset, defined as "integral of disk space over time", not computation.

(queue people complaining about hard drive shortages)

I guess any approach that relies on wasting resources is problematic

Unless the resource is something that is good to waste, somehow

Unfortunately, that also doesn't work. If there is economic value to the computation, you create perverse incentives.

Like proof-of-removing-CO2-from-the-atmosphere

Ha, yes, something like that.

The action itself has to have no economic value, so if it has to be something good, it has to be a reverse tragedy of the commons thing

Though doing that in cryptographic way that also commits to the chain itself is... wild.

yes, seems very hard to come up with something that at the same time works cryptographically and inherently rewards something that improves a tragedy of the commons

Seems like I'll have to catch up a bit with messages...

https://deathbysheep.com/track/phantoma

My 3 girls play rock-paper-scissors-lizard-Spock

They were playing rock-paper-scissors yesterday, so I taught them that variation

https://i.redd.it/paxahb343xg61.png

I mean, the current method of Databases running on mainframes seems to work OK.

Yes, but that's a centralized solution that ultimately depends on and gives unlimited power to the central authority (they basically can and will control inflation rates and thus basically own all the money - or at least they have the power to assign any percentage of it to themselves at any point by just 'printing' the appropriate amount of money).

(which by the way makes it silly for a government to claim that "we don't have the money" to implement Good Thing X, Y or Z)

I think the main current problem with it is that the companies that currently run the mainframes are for profit. (e.g Visa or MasterCard) what we probably want is a not for profit payment provider, who's transaction prices reflect their actual costs.

Ie, a fraction of a penny/cent/yen/whatever.

they also censor whoever they want

True. That seems to be what was happening with OnlyFans.

not quite from what I have read

Transactions are not the biggest problem though imo. I think it's a bigger problem that "mining" is free for some (central banks and to a lesser extent regular banks) and impossible for others.

apparently, the problem with OF was https://www.bbc.com/news/uk-58255865

Must be a shitty job, being a content moderator. Literally watching the goriest, most disgusting stuff the automated systems are tagging, to decide on what side of acceptability it is on.

Probably sounds fun at first sight, but it must be so mind numbing and nightmare inducing after a while...

Yeah. There is meant to be a lot of mental health issues amongst Facebook's moderators, etc.

Just hire some 4chan and BestGore visitors ¯\_(ツ)_/¯

They'll actually love this job <:kekw:808717074305122316>

nah. t. just trust me here

https://twitter.com/scienceshitpost/status/1326034806950391808

I was working on https://lifthrasiir.github.io/roadroller/ for a while, that's one reason I was unable to follow this server recently lol

(technically still a compression stuff, but I guess this channel is more appropriate)

Perhaps <#805176455658733570> will be better also Brotli? https://caniuse.com/brotli

I thought it is not the image compression so here is more appropriate, but anyway

it already beats Brotli in terms of compression ratio, especially given that the output includes the (tailored) decompressor

but it is very slow for the obvious reason and only makes sense for demos 🙂

the current scoreboard FYI

I think Brotli also has room for improvement with higher potential

is there a zopfli equivalent to brotli?

As far as I know, no

`brotli -11` is basically what zopfli does

I wonder what could be done if we would 'port' some of the ideas of jxl's entropy coding to general-purpose compression

As far as I remember -11 is the same as -best in the comparison, but still, considering how Zstd has changed during all that time and added slower compression modes, I think Brotli also has room for further optimization and denser compression

yeah, very likely

also for better speed-to-compression tradeoffs

btw, you may want to compare with large window brotli (-w 30 or so)

There is nothing inherently image-specific about the concept of HybridUint encoding with an MA trees + cmap context model, an lz77 option and an ANS tokenizer

Thank you for the info, I never thought to tune brotli parameters before

of course you would have to replace the MA properties by different properties that don't refer to 2D neighbors, and also different predictors to put in the MA leaf nodes

and you'll also likely want to add a useful window initialization

about roadroller... is it basically arithmetic coding + lz77 but with a small decoder and self-contained?

rANS, context mixing, some preprocessing

ah, no lz77?

x86 demoscene folks long had a similar thing for a while, so all I've done is the same thing but in JS

rANS is the same as a LIFO range coder, right?

Yup

With simpler renormalization I guess

Also, is such a large window supported in browser decoders, as there is usually a warning in the encoder that it may be an non-compatible format?

No LZ77, but remedied with a large number of context models

that's the advantage of doing it LIFO, right?

Yes to my understanding

with chance adaptation, if there are enough contexts then basically the ctx model can approximate lz77, that's also how flif does a reasonable job on things that are lz77'able

in jxl we dropped adaptation because it's bad for speed, but we still need to understand better how we can combine lz77 and ctx modeling more effectively

there's a funky redundancy in how repetitive stuff can be either coded via lz77 matches or via baking it into the ctx model, if you have a MA ctx model

I think your decision is right, it took a while to figure out the suitable parameters for adaptive models and still Roadroller has to run a search for them (only reasonable cause it embeds its own decoder)

I think if we can figure out how to crack that nut and make good encoder decisions there (i.e. combining MA tree learning with lz77 matching), there's potential for significant improvements in lossless jxl

Anyway, I've read quite a bit about compression algorithms but it's the first time I've ever made it into a concrete code XD

I believe so

that's always fun 😛

I imagine LZ77 can theoretically do better than bounded context models - but not sure in practice when it is you start seeing wins

(also IIRC it is faster to decode, but that's secondary for that usecase)

It's sad that Brotli didn't get as much support and development from the community as Zstd and even attempts to add Zstd to browsers and replace Brotli, although this is very irrational

I've tested this, but the default of -w 16 (I guess) seems to be enough for the current use primarily because the input size never exceeds 64 KB

fair enough

paid font, similar to stena and suduwe but i think is monospaced

looks ugly but relaxing

353 kb this screenshot

389 kb segoe ui

353 kb arial

397 kb suduwe

this load faster

as fast as arial

I will get headache with those fonts.

you won't, because you won't use them

a very empirical statement

I will get headache with those screenshots of those fonts

only if you look at them long enough (you won't)

jokes aside, do people actually get headaches from fonts?

maybe, if they are illegible

Sorry if last discord preview Is bright

The only default font i use is lucida console in notepad

Custom font Forever

this is the best ever

https://bottosson.github.io/posts/colorpicker/

this is <#794206087879852106> 🙂

ooh that's interesting <:FeelsReadingMan:808827102278451241>

https://youtu.be/3vZY4ITcU8c

https://godbolt.org/z/z3vbGa7bK for people that want to see some *weird* C++

(I had some fun today trying to put together the most esoteric C++ features I can name)

TIL that emoji blending is a thing

🐱 and 🙈 combine to this

Same thing

10/10 emojis

https://www.youtube.com/watch?v=Li-f1BX0XEo

https://twitter.com/nucliweb/status/1436301702488018950?s=20

https://jsfiddle.net/y9d7krxv/ another abomination I've made recently (relates to Brotli)

what is that stuff it decompresses into

I don't want to spoil the fun, but I can confirm that it should decompress into 122,784 bytes of data

probably uses whole lot of brotli's dictionary because I see a lot of html-ish stuff what doesn't make too much sense

correct! it is _the_ Brotli preset dictionary.

the resulting Brotli stream is large (for the obvious reason) but the code generating the stream is not

> coolfeet.php what was brotli dictionary trained on <:Thonk:805904896879493180>

it is word-based, so it'd be probably `cool` `feet` `.php` individually

though so but wanted to b funny

I've done specifically because if I can somehow extract the preset dictionary it can be used in small JS demos

imagine that 1KB JS is somehow fetching more than 20KB of words out of nothing...

I mean... https://www.reddit.com/r/programming/comments/k6ysmz/google_brotli_dictionary_contains_sonyguysfuckpipe/

so apparently naming things cool and feet is v efficient

lmao

unfortunately I can't use embedded bitmaps in WOFF2 because they would be sanitized out by OTS

that's legit annoying tbh

yeah Japaneses seem to complain about that a lot

theres https://github.com/khaledhosny/ots/issues/200

OTS?

OpenType Sanitizer, used to filter the font file before giving it to the OS rendering stack

embedded bitmaps in particular had an actual exploit using them

that's probably why they are outright filtered out

huh

so the state of security in font rendering is so horrible that we are passing them through some external program rather than trusting the libraries to gracefully fail? xD

unfortunately yes...

in Windows a part of the rendering stack is in the kernel (!) so it is much more problematic

good design, that

There were also experiments with custom Brotli dictionary https://blog.cloudflare.com/brotli-compression-using-a-reduced-dictionary/

https://spy47.bandcamp.com/album/ticket-to-sweetland-2

oh no that vile bot returned me early adopter role

https://darkclover-records.bandcamp.com/track/riniot-feat-incomer

FWIW, you could exclude this dictionary by compiling with `-DBROTLI_EXTERNAL_DICTIONARY_DATA`. See also: <https://github.com/emscripten-core/emscripten/issues/9960>.

```c const int main[] = {}; ```

best valid C program :)

oh god

https://twitter.com/nick_kapur/status/1436908703567933440?s=19

Interesting way to sign a petition

Not very easy to add more signatories though

Just alphabetical order or something would also work to "anonymize" the "leader"

That is also not friendly to adding more signatures though, if it's on paper

are japanese runes easy to alphabetically sort?

there are so many of them

No idea if it makes sense, I guess you could sort by unicode codepoint?

no such thing back then. also dont they have like 3 systems for that? i guess one could convert to kana for sorting but sounds complicated

i think it's usually just consonant then vowel

5 and 5

in computer systems the chinese characters are sorted by unicode codepoints

but by hand i tihnk it's usually pronunciation in kana

Japanese "runes"? They have three different writing systems. Hiragana, katakana, kanji. Hiragana and katakana have a specific (modern) order, equivalent to alphabetical order for English. They can also be ordered according to a poem. Kanji can be grouped into radicals, then stroke number. https://en.wikipedia.org/wiki/Japanese_writing_system https://en.wikipedia.org/wiki/Iroha https://en.wikipedia.org/wiki/Radical_(Chinese_characters)

its kinda joke and kinda how golang and plan9 C refer to unicode characters (https://9fans.github.io/plan9port/man/man3/rune.html)

is there even a word for what combination of these 3 systems is. just saying japanese graphemes or writing system feels too unfun

they are all referred to as kana I think, Though that might only refer to hiragana and katakana

because kanji is just hanzi

Testing a Solitaire gameplay VP9 4:4:4 encoding (spoiler: happy ending 😁)

(Thanks <@!321486891079696385> for the encoding guide on Reddit, my first VP9 encode was quite crappy)

I can send anyone clean `yuvj444p`+`pcm_f32le` version for test encoding purposes 🙂

just made myself an omelette with cheese (an _omelette au fromage_, if you will) and felt like a meme incarnate

Why is that memeish?

because of https://youtu.be/2kArCRjT29w

Amateur. https://www.youtube.com/watch?v=njFw1NOAu3s

reminds me of https://youtu.be/GAoFDzVhHdw for some reason

he also has one on pizza: https://youtu.be/Jy3uREg7v_o

https://kotaku.com/wow-this-quake-clone-is-smaller-than-a-word-doc-1847668782

https://news.ycombinator.com/item?id=28520221

Also Q1K3 uses Roadroller compressor https://discord.com/channels/794206087879852103/806898911091753051/885117677885091851

nice game

but I keep dying

yeah, I too didn't get past the first level partly because of the lack of aim

Hi, ive been trying to compare images using the butteraugli command line tool and i was wondering if one of yall could help me with something. I think the first number it puts out is the max distance error of the image, however im still sorta confused about what the second number means. Do you know how to get a "average" score of the image instead of just the max error? I feel like that would be more accurate for my testing

that's kind of what the second number does

https://en.wikipedia.org/wiki/Norm_(mathematics)#p-norm

first number is inf-norm (max)

second number is 3-norm (or iirc it's actually a weighted sum of 3-norm, 6-norm and 12-norm or something like that)

1-norm is mean absolute error (though in case of Butteraugli that would be not RGB MAE but rather something like MAE over some heatmap computed using a Laplacian pyramid in XYB, or something)

2-norm is root mean square error (RMSE), which is basically the same as PSNR

the higher the norm, the more it's penalizing the worst error

the lower the norm, the more large areas of small error can compensate for small areas of large error

(e.g. getting an easy background right will result in MAE or PSNR looking good, while the artifacted foreground can be quite bad)

Thank you so much, ive been trying to wrap my head around this for a while now <:KoroneDorime:876123503479361587>

https://www.youtube.com/watch?v=81rBzfbFLiE > Codex, create a lossy great quality image encoder!

Are you low key mean jxl isn't good enough?

No, I don't mean that, jxl is best still photo image encoder for now(av1 also a strong encoder), but I guess in future we have chance get a AI-base image encoder.

HEIC good, jxl bad

And I think jxl still have to implement some quality improve for graphics(non-photo).

I think an AI encoder for a classic codec like jxl would be cool

Fully AI-based codecs are also coming, but that's kind of scary to me in terms of fidelity

too ontopic tbh

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

Beautiful places in game.

🙁 https://www.theguardian.com/technology/2021/sep/16/home-computing-pioneer-sir-clive-sinclair-dies-aged-81

I think it's the same as any other business email

jon@cloudinary.com uses google suite gmail, spam filters against regular spam are good

Only remaining spam are personalized manually written emails by marketeers who want something from me to get more traffic to their site

you could also do things like using `veluca93+github@gmail.com` for github and then set up filters on gmail or something... 😛

That could defeat automated spammers but humans could go to the main inbox...

but they may be honest enough to say “I found your profile on GitHub”

In my experience automated spammers are much much worse 🤣

by now, even they may have picked up the trick

I wouldn’t be surprised if some of them have an `s/\+[^@]*(?=\@gmail\.com)//` now

(to be clearer: ```shell echo "veluca93+github@gmail.com" | perl -pe 's/\\+[^@]*(?=\\@gmail\\.com)//' veluca93@gmail.com ```)

wut game is this

Sky children of the light ( for android / iOS /switch)

regex 🥶

Ah sorry for spamming

Better to do that in <#805414133788180491>

Yeah

p sure yes

Why does it say that you're deleted?

His username is deleted. (He renamed it to deleted)

https://soundcloud.com/zytherprime/shinjinrui

Duel in the Mist - Inazuma(稲妻) OST Battle Theme MV > https://www.youtube.com/watch?v=FYP2dM8RTFM

yes

Wow, the composition is insane.

I like shamisen sound 🙂

https://youtu.be/c5XIDt90VgY

https://discord.com/channels/794206087879852103/794206087879852106/887782619222245478

Although I also think it's a typo, but still, it doesn't change the meaning (sRBG/sRGB) But, if it was RGB and BGR, then they are slightly different things

https://c.tenor.com/cYzOFdtjlzsAAAAM/serious-looking.gif

https://upload.wikimedia.org/wikipedia/commons/4/4a/AMD%407nm%2812nmIOD%29%40Zen3%40Vermeer%40Ryzen_5_5600X%40100-000000064_BG_2042SUS_9JF6228V00014_DSCx4%40IR.jpg

https://cloudinary.com/blog/for_developers_the_html_picture_element_explained

>>>#ontopic

https://www.reddit.com/r/headphones/comments/pt8gv3/wow_so_this_must_be_why_flac_is_better_than_mp3

old copypasta

<https://encyclopediadramatica.online/FLAC>

i need those fonts i have two i need complete series

i will ask in discord av1 if anyone knows thailandese

http://www.fontpsl.com/webpage/home/index.php

ifixit based

https://youtu.be/zKYG2TiBk9k

holy shit if this rumour is true tho

<:monkaMega:809252622900789269>

That's why I use https://www.ifixit.com/Device/Fairphone_3

well it's a double edged issue imo with the iPhones you get guaranteed software updates for around 5-6 years but if there's any hardware issues you're kinda sol on the newer devices with most androids you only get 0-3 years of updates but replacing parts is 0 issue. You can make an argument for unlocked bootloaders and custom roms on android but google *thru safetynet* and the phone manufacturers appear to be trying to make it harder

it's unfortunate how the fp3 in contrast with the fp2 ships with a low end / midrange soc tho

simple, use landline phone, and real computing device for other stuff. bonus: no one can track your location

`real computing device` Can't use that thing on travel / bus and with day long battery.

laptop is an option, some last pretty long

no artificial support deadlines, restricted repair or installation of software, can just use modem for portable communication

people consciously trade their freedoms and independence for that small bit of portability & utility ground left to cover, and then act as if this outcome wasn't desired

it's a set of trade-offs that led to current state of matters, and it works well enough for the most, as this results in revenue for makers and evoluves portable computing

I just want a decently powered linux smart phone. I don't even need android apps.

https://youtu.be/5pYRatrnnPo

If it has LTE or 5G

🤔 but everyone will know where u live instead

no issue when switching batteries besides a non genuine warning

looks like the rumour was wrong then

that's nice

afaik it doesn't?

It doesn't have cellular network support

i'd imagine the m.2 slot is m key only or m&b key?

is it even feasible to do a dual e and m key slot?

just plug usb modem

I don't like hacky solutions

yeah probably too annoying and wire/socket would wear out if used frequently

Yes

Pine phone is good. Needs sometime to mature though

yeah was abt to mention that

a bit weak on specs but premise is p good

though considering price, specs kinda make sense

its p cheap for phone

https://mahoukichi.bandcamp.com/album/postcreator

https://twitter.com/ZoneNature03/status/1441956188585820171?t=jj3LNddbEcycij6uWwc7Zw&s=19

Haha so cute

Until you see the rest of the photos

Cute:

Not so cute:

<:KekDog:884736660376535040>

nice

<@!219525188818042881> My friends like your tank

Lol. Rip tank though <:KekDog:884736660376535040>

. . [🦊 . ███████)▂▂▂▂▂ ▃▄█████████████▅▃ ◥🍪▲🍪▲🍪▲🍪▲🍪▲🍪◤

<:FoxTank2:761241439400624199>

Most likely <https://store.steampowered.com/app/599140/Graveyard_Keeper/> (but, those aren't my screenshots)

https://arstechnica.com/gadgets/2021/09/discord-libreoffice-and-other-app-stores-come-to-the-microsoft-store-in-windows-11/

libreoffice on microsoft store

https://gitlab.gnome.org/GNOME/gimp/-/pipelines

jxl dev

macaw font full

best serif ever for readability

otf versions

elegant

😩

reading jxl encoded love live doujins feels so based. living in the future

yup

https://9to5mac.com/2021/09/30/usb-c-branding-cable-confusion/#disqus_thread

https://twitter.com/cramforce/status/1443962459723755533

hoping to have fun with another type of Modular

v nice

it probably cant run linux tho right

🤔🤔🤔🤔 https://www.phoronix.com/forums/forum/phoronix/latest-phoronix-articles/1282476-facebook-is-aiming-to-make-compilers-faster-using-machine-learning-with-compilergym

sadly not, indeed

Doesn't work with wine?

I suspect it might “work” but I wouldn't be sure about latency, good VST support and so on

but fun can be had with other synths too, such as ZynAddSubFX, which does run on Linux

Works with strawberri melk?

Very likely yes

How much of encoder improvements is better defaults, vs actually changing anything about how the encoding works? I imagine there are a lot of cases with heuristic improvements where a certain density was possible before with some arcane options tuned just right, but after improving the heuristics that becomes the default and people then get better images.

only few things are exposed as arcane options you can tune from cjxl, most are just choices the encoder makes

https://android-developers.googleblog.com/2021/10/android-12-is-live-in-aosp.html

> AVIF image support - Android 12 adds platform support for AV1 Image File Format (AVIF). AVIF takes advantage of the intra-frame encoded content from video compression to dramatically improve image quality for the same file size when compared to older image formats, such as JPEG But, still not Jpeg XL <:SadOrange:806131742636507177>

Wait for android 13 (´;︵;`)

tbh given current amount of bugs being found in libjxl id say its reasonable that its not in yet

What avif decoder is in android 12? Is it dav1d or libgav1?

Smooth gif. It looks like a video and has more than 256 colors. 🤔 https://developer.android.com/about/versions/12/images/splash-screen-gmail-example.gif

Most likely libgav1, if nothing has changed by the release

Question is who will use android avif support and when

If you want to use avif in an android app, you'll want to package your own recent dav1d, because the android avif decoder either is not there yet, or it's very slow and outdated.

(for the same reason, even libjpeg-turbo does get packaged in android apps, since a recent libjpeg-turbo has way better ARM optimizations than an old one from older Android versions)

Note that it's in particular the devices that run older android versions that need the most recent, best-optimized decoders the most

https://gif.ski/ ?

May be

Jxl is bigger than gif. 2.71mb vs 2.94mb

libgav1.

It does pose a small problem though: even today, libgav1 is quite a bit slower than dav1d, and threading is... not optimal. That's not counting the fact that Google does not update media components with the Google Play store.

The yearly release cycle combined with lots of devices that don't get updated makes Android system support not a very good target for modern codecs that are still improving their implementation.

That's what bothers me: Google could "easily" update media codecs(software decoders) through updates via the Play Store for a few Android versions now I think.

Android could certainly use a more flexible distribution mechanism, it's silly how there are basically no shared libraries that are useful/up to date, so apk's at the same time are very conscious about package and install size, yet are kind of forced to statically link and bring their own copy of anything for which the version from 5-10 years ago is not good enough.

encoded at -e 9, very slow though...

Does lossless h264 or hevc compress it better?

Or vp9

They are all bigger than gif lol. 6.53mb for x265 lossless. 3.37 for vp9 lossless

strange 🤔

i managed to get x265 and vp9 both to 3.23 mb lossless, but still bigger

oh i see, there is some dithering

possibly why?

yuv?

Am I doing it wrong `ffmpeg -i splash-screen-gmail-example.gif -c:v libx265 -x265-params lossless=1 spash.mkv` `ffmpeg -i splash-screen-gmail-example.gif -c:v libvpx-vp9 -lossless 1 splash.webm`

i was using slowest encoding options

for ffmpeg vp9 it would be -deadline best -cpu-used 0 -lossless 1 not sure about x265 in ffmpeg

wouldnt you want to just use x264 lossless

x265 is worse

2.44 mb for x264

better than the gif worse than the jxl

<https://engineering.fb.com/2021/09/13/core-data/superpack/> yo have u seen this? <@!219525188818042881>

Yes

But kinda want to see third party comparisons and comparisons with other compression algorithms. Especially since BBC's shitty crippled AV1 vs VVC testing which caused me to trust those things even less <:KekDog:884736660376535040>

this one is quite interesting 🤔

<https://engineering.fb.com/2021/09/13/core-data/superpack/>

it has some strongish dithering at some points not nearly as bad as what gifski spits out tho

48k opus is so good

<a:adogeping:682300957597368353>

<@!272986016242204672>guess it was you again <a:FrogStab:821038589345136682>

yo <@!321486891079696385> what u think of this audio test?

i'll do a 5k bitrate one just for u ❤️

i don't have opusenc on here <:PepeHands:808829977608323112>

Then download it <a:ThinkingSphere:821038590091329557>

pog

cryptominer time woag

which one is opusenc and which one is ffmpeg 🤔

i didn't use any fancy settings for either

<@!794205442175402004> I hope you can answer this: How long does it typically take until I "Immediately get free extra quotas" after tweeting my "Cloudinary 'Spread the Word' link"? I posted a tweet a few days ago https://twitter.com/KrishayRizwan/status/1444283155041636353 (and followed Cloudinary) but didn't receive the credits.

No idea, I'll ask around

what's your email or account name?

dronkk it

Store voltage, drink voltage.

Some discussions about AVIF and JXL https://news.ycombinator.com/item?id=28759181

<@!794205442175402004> hey, i just wanted t give you some feedback on your ISO 'petition'. mostly positive. but i'm not sure here is the place for it (it would be a kinda political i guess). so if you want to talk about it you can dm me and then i can probably dm you back.

👍

Looks like I can only dm you if we become discord friends

in my experience it seems acceptable to send someone a friend request if you only want to dm them. at least, people always do that to me (people from a server where i have turned off DMs for a reason) <:kekw:808717074305122316>

I found it... the ultimate virginity protector <:KekDog:884736660376535040>

Oh no, it has text on the bottom that says: "SOME ARE PC, WE ARE GAMING"

gamers

Cute dragon

do they also own the life sized lucky?

https://twitter.com/TechAltar/status/1443749297250607108

AMP pages do sort of suck tbh. I much prefer a decently optimised initial website.

The main thing I like about AMP is also it's greatest weakness, any site can rehost the content and you don't know if it's unmodified or not

> don't know if it's unmodified https://amp.dev/documentation/guides-and-tutorials/optimize-and-measure/signed-exchange/

I knew a bit about that, but I've got no idea how widespread it is since I don't use Chrome (Or search for news articles much actually, mostly browse sites I trust)

i thought only chrome & chromium forks could use amp

lmao they recommend stackoverflow for community help

In general we support the latest two versions of major browsers like Chrome, Firefox, Edge, Safari, Opera and UC Browser. We support desktop, phone, tablet and the web view version of these respective browsers. Beyond that, the core AMP library and built-in elements should aim for very wide browser support and we accept fixes for all browsers with market share greater than 1 percent. In particular, we try to maintain "it might not be perfect but isn't broken"-support for iOS 8, the Android 4.0 system browser and Chrome 41.

🤔

https://blog.chromium.org/2021/10/renderingng.html

but is it as good as gecko

IDK, I use both Firefox, Chrome / chromium. Chromium based browsers feels fast and responsive than Firefox.

<:Thonk:805904896879493180> https://twitter.com/jxlapp

Jira Excel, haha

looks like it was created in 2021

most 3-letter acronyms have multiple meanings, I guess

there's also a Java Excel API which is called JXL

https://en.wikipedia.org/wiki/JXL

Junkie XL, quality name

https://www.swissinfo.ch/eng/smuggler-busted-with--a-bit-of-swiss-cheese-/47000686

me_irl

Totally off topic, but I have no opportunity to ask such a distinguished and knowledgeable audience otherwise: is there some library or a bunch of algorithms or techniques or something that may help to predict how well an image will compress based on its characteristics? In he old world that would be a choice between a JPEG and a pngquanted PNG and in the new world it would be one between AVIF and JPEG XL or different modes of either of ’em? Basically, something like automated part of point 1 here: https://github.com/guardian/grid/issues/1835#issuecomment-210920733. Doesn’t have to be perfect, but would help avoid (or, at least, partly inform) making a bunch of different encodes and comparing their similarity with original and their filesize (TBH, an automated way to do the latter would also be good to know about :D)

sounds like a job for either handwritten heuristics or neural network training. and no this isnt "totally off topic"

i suspect something what would detect patches of solid colors or primitive gradients could be helpful but theres probably much more than that

oh also <https://github.com/libjxl/libjxl/pull/466>

https://www.youtube.com/watch?v=6TFjWe67Pgg

is it encoded from the gif or the original video?

i'd imagine you'd go video > png > jxl

libjxl, and imagemagick with jxl support landed on archlinux repos so maybe one day we can hope ffmpeg working w it too. things are moving

https://youtu.be/dZStUzYE8Uw

https://maverssoft.itch.io/archangelnemesis-s1e1 this looks interesting

Guess what I have been doing today

(together with ~70k others)

aww no kkomrade emote

next, you should get them to protest ISO paywalls

you can add emotes yourself ithink

i can't in this server since i'm not a mod, and i don't have nitro so i couldn't add it to a different server and use it here

just try it. can just steal .png to bypass nitro thingy

oh yeah i mean i could upload the image but i can't use it as an inline emoji or a reaction

unless it's priv limited to early adopters or something like that but i can add whatever stuff

no i mean like

yeah i can't access that

i see

then just send png i can add that

https://cdn.discordapp.com/emojis/659512168861138965.png

done.

<:kkomrade:896822813414011002>

or just "accidentally" leak them to us <:kekw:808717074305122316>

<:kkomrade:896822813414011002>

Masks are still required indoors, e.g. on public transport, but not outside.

This was the first big demonstration since the BLM one (and that one was not allowed but I was there anyway, together with many others - all with mask though, then)

https://pbs.twimg.com/media/FBRUcuuVkAAB-uc?format=jpg

> All he needed was a bath. Rescuers were finally able to clean the curry off of the herring gull after he put up a bit of a fight and covered the veterinary team in curry water.

<:KekDog:884736660376535040>

Whoa, the kids just saw a deer in our garden

Our garden is not particularly large, and I live in a suburban area near Brussels

So that was a bit unexpected

Can someone send me or link avx2 cjxl windows x64?

Scalar is very slow 🐢

don't the ones from https://artifacts.lucaversari.it/libjxl/libjxl/latest/ work?

How high are the garden fences?

Thanks

we have no real fences, just bushes

https://www.instagram.com/reel/CVDTqLbj-5V/?utm_medium=copy_link

old

https://jxl.moe

what is it with .moe domains

i just thought i'd snag it when i got embed.moe

I expect some anime jxl drawings. 😔