Thread 106283710

I'm bored so I decided to see what all the noise about modern image formats was and use picrel to test them. This isn't just any PNG btw, I compressed it with oxiPNG and used max compression for the newer formats to make the challenge more relevant to anyone using a computer that isn't a potato.

PNG = 32,961 bytes

AVIF = 21,361 bytes
https://files.catbox.moe/6l1d0k.avif

WebP = 10,174 bytes
https://files.catbox.moe/obfjjs.webp

JXL = 9,719 bytes
https://files.catbox.moe/f2qokk.jxl

JXL absolutely dominated, achieving ~71% better compression efficiency and WebP achieved ~69% better compression efficiency which is still NOICE. AVIF achieved ~35% better compression efficiency which isn't too bad either but it's puzzling how a 2010 image format (WebP) is outperforming a 2020 image format (AVIF).

>inb4 "you should have cropped it to 1488x1488"
Yeah, I fucked up. I don't feel like redoing everything DOUGH.

>>106283710 (OP)
There was some autist that turned pics into 1 frame webms and reposted. Something to do with file size/image quality. You seem to know about these things. Do you have any idea what he was on about?

>>106283735
1 frame webm is the same as webp

>>106283735
I mean that is what formats like WebP and AVIF do, they store a video keyframe and pretend it's an image. However they slightly differ from regular video files because they can also store and compress lossless image data that comes from a PNG.

>>106283752
>>106283768
>he was just shilling webp support on 4chins
Thanks. I understand now.

>>106283710 (OP)
Literally who cares

>>106283813
>t. seething
lolle

>>106283824
It was a question.

>>106283845
There is niche in /g/ for image quality compression autismo. Not enough for a general thread, though.

>>106283857
I don't see it frequently. What's why I asked. It seems beyond niche especially since storage is so cheap

>>106283875
nta but thats not why you asked
you want to do negative reinforcement on all sorts of classical autsitry
i see a pattern of tranies just like you being negative towards legitimate autismo

>>106283906
I'm sorry my post upset you so much. You don't have to get overly deeply analytical about it. Its just a question.

>>106283922
>You don't have to get overly deeply analytical about it
wouldnt that make things easier?

>>106283875
anon's lolicon collection got a tad too big and he needs a more efficient way to store cunny

>>106283937
Not by the definition of the words overly and deeply and analytical.

>>106283981
'depends on the natural and acquired inclinations
in my case alertness is the default state even when i sleep

>>106283710 (OP)
I don't want to play Where's Waldo. What did you hide in the image?

>an actual image codec comparison and not just a retard crying about 4chan not adopting modern image formats because they cater to Pentium 4 PCs

Fuck, I hope I'm not dreaming.

>>106283813
>>106283875
It's fair to point out that with better lossless compression you get to a point where lossy compression at least for images no longer becomes necessary on the web for like 70% of images on the web. PNG 1080p screenshots for example especially unoptimized by oxiPNG use multiple MBs of space which is fucking retardedly bloat. Modern lossless image compression hovers somewhere around 1 MB for 1080p screenshots or less if it's CGI And that's just for images in general.

So instead of deep fried memes we could be preserving the quality of images on the web.

>>106283710 (OP)
go back to your containment board

>>106283710 (OP)
I click JXL and it doesn't work, looks like it lost

>>106284061
OP's source didn't come from a screenshot of twitter's anus made by a literal "who?", though.

>>106283710 (OP)
>max compression
Max lossy or lossless?

>>106284086
I can't tell them apart but I'm on a phone so maybe someone can magick them to verify if they're lossless or not.

>>106283710 (OP)
Member when made that flag and psyopped librards into adopting it? Back then when it was just faggot day jews were using that day to push anti straight white agenda so we made the anti straight white male flag to identify our enemies. We added niggers and spics to it too and made it look like the Palestinian flag to mess with the jews knowing their libtard golems would fly the flag in their jewish masters presence. That was 10 years ago and its faggot month.

>>106284186
>memberberries
dw, the world keeps turning with or without you
it'd be better if you didnt sit on your ass for the past 10 years, but ids not yet hobelets

>>106284072
Google is being a faggot kike about JXL right now. Safari on iPhones and Macs support JXL no problem.

>>106283710 (OP)
You're delusional if you think JPEG XL is going anywhere. It's a dead-end format backed by a handful of nerds while the world sticks with proven winners like PNG and WebP, and your shilling just exposes how out of touch you are with reality.

>>106283710 (OP)
Hi Daiz. AVIF is already widely adopted and used and absolutely crushes all other formats in real life applications.

>>106284841
PNG is widely adopted.

>>106283710 (OP)
png is the best because it was the first and it was made for human.

>>106283710 (OP)
>oxiPNG
Either it isn't that great or you suck at using it.
That png doesn't even use a pallete and the bit-depth is 8bits even though I count about 11 colors.
A palette with a 4-bit bit depth (index in this case) means you can store 2 pixels per byte.

>>106286112
According to onlinepngtools there's over 1000 colors in that image so this wouldn't work.

But I don't see over 1000 colors...

>>106284186
Mushrooms mentioned.

>>106283752
it's actually webp 2 (vp9), webp is a still frame of vp8

Someone post the Satania picture already to put an end to this shill thread

>>106284056
Deep fried memes haven't existed for like five years

>>106283729
You're really weird. Nobody thought that
Now try it with images that have more than 11 colors. You might vaguely begin to understand what you're trying to discuss.

Sorry I'm late

png or death

So basically avif gives you the best result if you want to really bit starve you image since that's what video codecs were designed for and jxl for when you care about fidelity and want your image to be as small as it can be.

>>106283710 (OP)
Good thread OP
But using computer generated imagery is cheating. Some algorithms take into account the real world textures and stuff. Benchmarking them on a single type of image might paint a misleading picture.
Can you do the comparisons with some pictures of nature, buildings and animals?

>>106283735
I think you meant gifs instead of webms because i remember a post about that a few days ago

>>106288394
>jpg

>>106284376
enjoy your CVEs dipshit
https://desuarchive.org/g/search/text/gf7dyq.jxl/
https://github.com/advisories/GHSA-23ff-wfv3-xrvg

>>106283875
Storage may be cheap but bandwidth is not. Every good morning picture you flush down on the internet blocks the pipes a little more.

>>106288586
Wtf is up with that JPG? I have to drop jpegli quantizer to high 30's to recompress it to the same size.

>>106288677
Cloudflare automatically compresses images, and basically every website is behind Cloudflare.

>>106288037
>Nobody thought that
I did.

Hmmm
>>106288888

>>106288696
it's just high-ish resolution and very noisy

>>106288448
kys

>>106288908
How was it compressed in the first place then?

>>106288918
>he struggles to understand the concept that lossy to lossy is less efficient than lossless source to lossy
all that noise in the original jpeg has been re-encoded into further noie which helps reduce the filesize
...but re-encoding the additional noise means that's now detail that has to be preserved instead, of course you aren't going to get the same efficiency as the first generation encode.
this is basic common knowleldge of how lossy codecs work, retard.

>>106288869
So what? Only the people who work for Cloudflare should discuss compression?

>>106289097
What do you think the point of my post was

>>106289109
What was the point of your post?
Do you go on other threads and tell people not to discuss phone prices unless they work in sales or not to discuss windows performance unless they work for microsoft?

>>106288869
cloudflare automatically compresses images... by losslessly optimizing them
that has nothing to do with this discussion
retard

>>106289138
So are you going to answer the question or

>>106289143
Yes, you're right - compression has nothing to do with compression.

>>106288448
just compressing everything down to a tiny fixed filesize is a useless test. you don't save and use images like that

a practical comparison should determine what is the smallest filesize/maximum compression ratio applicable before the image degrades to a certain threshold. then for each format compare filesizes at that quality point

>>106289151
I think the point of your post was that since Cloudflare automatically compresses images for almost every website it is pointless to consider the bandwidth issues that might arise with poor compression, hence my worries about bandwidth issues were invalid.
But Cloudflare does not do that and even if it did, there are still messaging applications that do not serve the images on Cloudflare CDN's. I also have a website where I do serve images without using Cloudflare optimizations.

>>106289162
>doing what is done to 99% of all images on the web is a "useless" test that is not a "practical comparison"
Keep the cope up my friend but you can't stop it