JPEG XL

it's noticeable

So the Russian one? I don't know Russian; what timestamp?

always

always speech

There are a lot of background noises and sound effects; without the original to compare to and not knowing what they are saying I don't know which noises are supposed to be there.

<@710263461803982850>

damn Vorbis has artifacts of the same kind, but they are almost inaudible

the original AC3 looks... well, there are no artifacts on the voice

I don't quite understand what's going on, but maybe I'll use Vorbis

The AC3 is 192 kb/s, the Opus is 64 kb/s. The noises/sound effects that I thought sounded like they might be artifacts are in the original so I guess they are intended.

oh no... I was afraid that you would say the same

just all 10 seasons are damaged in this...

i did comparison tests and yes, Vorbis does allocate more data to stereo, but it doesn't help because the problem is in the source

<:Opus:805856410235437068> ok

I was sure of this because the fact of the stereo artifacts was characteristic of OPUS

At low bitrates the stereo width is one of the first things it cuts, but 64 kb/s shouldn't be bad.

yes

As for me, Opus sounds better in this case

Yes.. it has a higher frequency, but as for me, there are also a lot of artifacts

what if you lowpass more before encoding

it seems that's what the vorbis is doing

The original is still limited to 15 kHz

I can only say that OPUS has very accurate weighting coefficients for each block of frequencies, and stereo Separately, but I'm not a programmer to look for and fix this

what i mean is that the vorbis is cutting it hard around 3400hz

and the opus is up to 5khz

so what if they both had an input that's already lowpassed at 3400hz

i can

imo it's vorbis' loss by default for not being able to encode higher frequencies than the opus

for opus is 3440

one moment

nono, 3440 for 1720

for 3150 use 3150*2

6300 hz

<@288069412857315328>

interesting

of course linear prediction works well by eye

But as I hear it's a lot more artifacts

on no

vorbis is stereo

lol he can

mono

<@288069412857315328>

I would like to say that VORBIS is now a professional codec

for games, for working with music, for archiving sound of non-standard frequencies

it 4000 hz 9 hr is decoded in 5 second

opus in 10 min

Games have been using Vorbis for a while, Halo for PC was the first one I saw

OPUS in its form, it seems to me, is the ideal LOSSY format, which is the best by default, which doesn't care on old foundations, frequency, noise level, hearing sensitivity, accepted in audiophile circles, and does everything to say - this is the best way (this is especially important due to the use full range against in all bitrate cases)<:Opus:805856410235437068>

i am so confused

I know it. for a very long time, almost since its release, and it still has no competitors for NOT all the obvious reason - this is the fastest codec for decoding

and of course free

well. just....just keep listen to opus 128 kbps which in 2022 is really indistinguishable from the original

but for myself I use 141 bitrate at which OPUS uses the maximum stereo width value of 16 khz

for all my 60 hr music

perhaps it would be very interesting to create a VORBIS-OPUS hybrid so that VIRBIS would be at low frequencies and OPUS at high frequencies.......... although in general, purely technically, I can do it even now, just later by combining 2 frequencies

lol That sounds good...

only the OPUS codec be incorrectly, but everything works - if you cut off the bottoms, then the minimum that can be set in opusenc encoder is 21 kbps, which in fact are 33 kbps - which suits me

hm

the problem is not entirely obvious, but it is found in the fact that at such low bitrates OPUS play stereo is bad, even if you encode from stereo by force

Wouldn't surprise me if a real time voice chat codec was designed for mono sources primarily

so experiment failed

and yet OPUS is not a "codec for voice chat" OPUS is a standard that contains 2 codecs. Completely different, CELP(music) and SLIK(voice) SLIK force dont use after 64 kbps

of course, there are big questions about why **DISCORD **allows NOT use SLIK only, but they know better (haha)

Here's the difference in quality (per eye) between uploading a 1080p video in 2D (for which youtube uses AVC codec) and uploading it in full-res Side-By-Side (3840x1080 in total) with 3D + aspect ratio metadata (for which youtube doubles resolution, uses 7 times the bitrate, uses the better codec, VP9, *and* reserves a 16:9 viewport which prevents cropping the thumbnail and empty space on the sides when the video isn't in fullscreen) Side by side slider comparison with more methods in the drop-down list <https://imgsli.com/MTE5MzM5/0/3> Playlist: <https://www.youtube.com/playlist?list=PLVDbhbTh1d8cmE47xRTU7zXy3hKBXxrhW>

Looks Like there is a comparison on the bitrate and frame size on which CODEC are completely UNcompetitive

<:PeepoDiamondSword:805394101340078092>

Very Looks Like it is AVC/x264

a codec that cannot be compared to any newer codecs size up 720p due to Known Technical Limitations its True? <@740013717114192023>

and of course. it makes no sense to compare videos by resolution on youtube, because he intentionally encodes videos with **ceremoniously **low bitrates, ~~due to considerations of capitalism~~

ok

I didn't like it so much let me show you

1.

ah lol

ok

wait what

AVC is not that bad. After all, many cameras still record to AVC.

no, I want to say that I don’t understand at all what the conclusions of the message from <@740013717114192023>

vp9

avc

3D videos are just 2 simultaneous views of the same thing from slightly different angle. in the case of youtube, they're side by side, the left view on the left. the conclusion of my experiment is that if we upload a 3D video to youtube and use that bitstream filter to double the height, youtube won't just double the display height but also the resolution and increase bitrate and thus, increase quality noticeably.

I usually upload 3D videos but just noticed how bad this 1920x1080 2D video looks, especially when moving fast through grassy terrain https://youtu.be/xxY_xaFrYm4?t=2420

perhaps, okay, the main problem here is that AVC does not support more than 1080, both technically and in terms of algorithms

so if I had uploaded a 3D version (2x 1920x1080 = 3840x1080 in total), a single eye/view from the 3D version would've looked a lot better than the single view (2D) version, because of the higher bitrate and codec <https://imgsli.com/MTE5MzM5/0/2>

I must say right away, for those who do not know, 4k in AVC is implemented by playing 4 independent streams at 1080p

but yeah, I get the higher quality has more to do with the bitrate than the codec

in the case of the 3d version, yes, YouTube simply cannot give more, but if you turn on 1080p vp9, then it will also be Bad

this judgment is not true, but in this case it is true

I wish youtube used MVC (the Blu-ray 3D standard), which encodes 3D more efficiently by making the second view just the difference from the first one

As far as I know, AV1 natively supports 3D video, so let's wait.

if only there was a way to force it on youtube 👀

🔥 my pc is wait

I can say that Youtube's AVC encoder settings are not the best. I've once uploaded a 240p video, it was an old footage from an old digital camera, a higher resolution video doesn't exist. Youtube just compressed it so badly, that it looked like a MPEG1 video from 1995.

Youtube's AVC encoder settings are fuking shit

to store some videos. I download 1080 resources, why do I compress to AVC 360p on my settings (the best for AVC) and get better quality for the same 360p Youtube bitrate

I would like to remind you that the maximum AVC encoding speed settings are like HEVC fast but ~30% better

because youtube uses hardware encoders, unlike you they have to process millions of hours of content

I don’t know which ones, but obviously not the CPU

that is why i said hardware and not software

maybe they designed the chip just to encode youtube? I don't know (at all)

iirc yes they do have their own custom designed hardware for it

I am glad that after 20 years, for video sizes less than 360p, there is a worthy replacement for AVC - AV1

neither HEVC nor VP8 nor VP9 are competitive with AVC at these resolutions

Those digital cameras and smartphones also use HW AVC encoders and the video looks alright. So the problem is not the encoder itself.

stop shoot video in 4k - better up the bitrate

frame size less - can compress better algometric<:AngryCry:805396146322145301>

Also 1080p often can do 60fps and 4K doesn't.

increased **efficiency **bits per pixel

they also use a relatively high bitrate, it will always look good if you allow it to use enough data

it's a matter of time and I don't like it

I'd rather have 60fps than 30fps.

in general, it infuriates me, people still use bitrate and CRF instead of CQ

for an example mine uses ~35mbps for video, not exactly the type of bitrate youtube uses for their videos

Why can't, give as much Quality as you want, why BE the CBR stream, if we're on the Internet, it's all so infuriating

oh yes, a static picture in the frame, 6000 bitrate please.

most encoders should be smart enough to not use a ton of bitrate on static frames

unless your doing CBR, in that case its expected then

this is the problem, that EVERYONE can, but it is written in the parameters, either ABR or CBR or CFR

all encoders in cases of a static frame can submit several BYTES

CRF should be basically the same as CQ on a static frame, the main difference is in motion

I don't need to go far, so I did, 8 hours of video 600 kb

it's a delusion

what

Last I checked Twitch uses 6000 CBR max (for non partnered streamers) and advises against VBR because of network-related issues or something

<https://help.twitch.tv/s/article/broadcast-guidelines?language=en_US>

My favorite kind of videos are those exploding into bunch of artifacts once every second. I find it kinda funny.

CBR is technically bordered VBR, YouTube also has CBR, and the video cannot be less than a certain bitrate, for 720p it is about 300 kbps Forced bitrate

this is why most PUBG/Tarkov/Fortnite streams are unwatchable IMO

no u can stream up bitrate

I think only a handful of partnered streamers get to stream at 8000kbps (maybe a bit more)

huh?

for example here 8000 https://www.twitch.tv/uselessmouth

on a side note, I once streamed at 120fps (still at 6000kbps) and it actually worked. even the VOD was 120fps but it required a 120hz+ disdplay ofc

i still dont understand what you mean by this, can you clarify?

I believe that no one needs more than 60, except for companies that want to sell it (for video)

yeah that's what I said. AFAIK only partnered streamers can use 8000kbps

can't watch you use YEP on this Discord server this streamer popularized ":yep: cock" for all twitch russians Using a smiley in any vulgar topics or hints

ok

obs

wait the Twitch.

promised to start implementation AV1 at the end of this year

AV1 will be very useful especially fro those with poor internet connection. Many internet provides still provide poor upload speeds.

2nd most viewed PUBG clip on twitch viewers be like: this is fine

🪤

This looks worse than a 3hr movie squashed using DivX in 700MB.

Ahaha no. Not the codec. The size of video. Streamers either continue on the same bitrate, but with better quality, or not, it all depends on the goal - to reduce the size or increase the quality. It's just that av1 is like opus, it's a sooo big leap that it's a little difficult to benefit from it, or earn money, if av1 is so hard then why be a twitch partner for up bitrate...eh, they'll come up with some bullshit

This is the problem of the codec being used beyond its capabilities Large frame size Low encoding preset Key frame limit of 2 seconds Locked bitrate cbr Badly done Everything, and got bulschit

Yeah but if your upload is limited to say 2000 kbps, you can't do much about it. A better codec can push more though the same channel.

The question is obvious - why did you decide that the upload is limited

Because so many people have garbage internet, because better connection isn't avaiable to them.

At what point from sending the packet to receiving it on the server, it is indicated to me that the limit is 2000 per second

There is such a thing as radio, for example, television, where every byte counts

And what is the Internet?

yes i am

too

can confirm. a bunch of streamers (especially beginners) settle with 3000-35000kbps, mostly because of poor upload speed and sometimes because of the *viewers* download speed, since transcoding (availability of lower resolutions) isn't guaranteed for non-partners (even affiliates sometimes don't get it)

But these are all misconceptions that no one understands

The video does not have to be of this bitrate The video does not have to be with a delay of 2-4 seconds from the streamer It is not necessary to have the best frame resolution for good quality But everyone says we need CBR It is necessary to have a response of a second or better 2 And of course, a very low bitrate for pixels, so that there would be porridge

Why doesn't anyone stream 3000 bitrate for 480p30? To be more precise, all the stages of bitrates and frame sizes should be shifted one down

If what you're asking is why is the bitrate so limited on the internet even though analog television and radio can transmit a lot more raw data, I think that's because of digital infrastructure being a lot more complicated to set up, but at least it's usually a lot cleaner and less prone to interference from what I understand

No, I'm saying that unlike TV, you can do better on the Internet for the same amount of data if you manage them correctly and work correctly with algorithms.

Well, I also have to say that the twitch hardware encoder for bitrates other than the original is fucking shit, much worse than YouTube, and in the obsalute it's about Avc VeryFast on less

lol I don't think TV (as in, analog over-the-air or cable television) is comparable most people's internet connections. AFAIK video and audio are transmitted losslessly so if you calculated the bitrate required, it would be orders of magnitude compared to the average internet's bandwidth

Twitch, Amazon, hundreds of servers around the world don't make me laugh

For 720p, everyone will have a bitrate,

Even the most backward phone ADSL gives 7000

But if you have 2G then yes, this is a problem

but yeah, the point of codecs is to minimize the bandwidth required, just like you can fit about a hundred MP3 songs when only about a dozen CD-quality songs used to fit

3G upload 5000

I still don't get the point you're trying to make lol

That's why analog TV is no more, they can use the same channel to broadcast 15 SDTV streams instead of one.

that streaming services should use a better codec/preset?

Only their quality depended on the crookedness of the person who uses the encoder, as well as on the bitrate, 128 of course was so-so, it was a pity there was no opus at that time, because opus is only 4 times more expensive for decoding

btw here's another comparison of AVC and VP9 in the same resolution tho a small difference in bitrate <https://imgsli.com/MTE5MzU3>

And besides, mp3 was not the first, no-

Why do you keep comparing anything to vp9?

is vp9 fast? No, avc is faster vp9 is the best at its speed? No, av1 is better at the same speeds

👉 <:AV1:805851461774475316> <:H264_AVC:805854162079842314>

no One codec can be faster than AVC with fast preset and give better quality

I'm not just comparing the codec alone, but the 2 encoding presets as a whole used by youtube for the same resolution*

No codec is better than av1 at the same encoding speeds, yep

av1 is out of the question because it isn't available to me in this case, otherwise I would've included it in the comparison

Youtube will soon abandon vp9 I think they are replacing all the previous equipment for encoding it, in order to encode av1 on the new one

only for new videos or old videos too? and if so, does youtube even keep a copy of the source uploaded file or will it just use their highest quality conversion?

Yes, Google decided not to waste time on your video lol. But I do not know that there is support for such 3d for av1, I know for sure that there is video for 360 degrees. You can just search for Top videos on YouTube, all videos with 100k+ views must have av1

Yes, YouTube accurately stores the originals of all videos, for additional encoding, from the very beginning of existence, first for the vp8 codec, then vp9, after av1, And it's also funny, opus, all the old videos suddenly sound better for this reason, although hardly anyone noticed

So Youtube is still keeping those WMVs from Windows Movie Maker?

3D is just 2 views on the same frame, so any format/codec supports it (although some might be limited by resolution, like someone mentioned regarding AVC). the *real* question is whether the format can encode it *efficiently* like MVC

Ah, thank god. I suspected as much, which is why most of my long 3D uploads are very close to the 256GB limit lmao

Ahaha

it's mind-boggling how youtube has enough storage for every video with its variants and original upload lol

Files are files, but you won't get access to the source

1. Buy the google 2. Take a video

I've been using 100-130Mb/s for my 3D videos. and the local files look almost indistinguishable from realtime gameplay

Oh, don't talk about such bitrates, this is 10 times beyond human perception for me

I don't think it is

at around 80Mbps I start noticing compression artifacts, so I used 100 which is more reasonable than the 130 max in shadowplay

Remember one thing though: 3D high action videos are quite a bit harder to encode than simpler content 😛

Is Youtube still so confused if the uploaded video doesn't have any audio track? As I recall it always took very long to process.

3D really isn't that different from regular videos. for practical purposes, it's just 2 videos in one, usually side by side or stacked one on top of the other. the real complexity comes from using a smart encoding algorithm that's stereo-aware to encode it more efficiently (like sharing pixels that are the same on both views or treat them as consecutive frames) I'm not sure if youtube does anything like that. Like if I upload a side by side video and then upload it again, this time telling youtube it's a 3D video, it will use a different encoding, like lower bitrate and will look a bit worse <https://imgsli.com/MTE5MzM5/1/2>

1 mbps Enough for everyone<:ReeCat:806087208678588437>

also, here's a slightly counter-intuitive comparison if you don't know the extra details: my local file 3840x1080 (in total) looks better than the uploaded one that's encoded in 3840x2160 (notice the far grass, grass behind the minimap, details on the trees and overall sharpness) <https://imgsli.com/MTE5MzM5/3/4>

the major detail being that the one on youtube is ~50Mbps and my local one is 100Mbps

Anyway, I've also tried to fit a video on a floppy, but at the same time, I wanted to play the video on 486 PC. So modern codecs were out of the question. The bitrate limit was so the floppy drive can keep up reading the file realtime.

I've used MPEG1, worked pretty well. It provided much better quality than those AVI codecs avaiable in Win95. The CPU had difficulties with the MPEG audio itself, so the performance was not very good. Without audio, it could do 15fps in a small 160x120 window.

I had to dig up an old program TMPGEnc because fffmpeg couldn't encode MPEG1 in 15fps for some reason.

https://youtu.be/b-2wP9QNmp4

As the graphics card lacks any video support, playing the video in other than original size was very choppy. You could see how the framebuffer if filling each frame.

And yes, it was fun.

fair enough <:HaDog:805390049033191445>

meybe Can av1 level 2

AV1 on 486DX2? I don't think that would work.

😈

Also opus doesn't seem to be viable option as the CPU struggles with old MP3, it can play it at 11025Hz mono.

dont use mp3 loool use VORBIS

10x or more faster

I can try vorbis.

👍🏻

It is true that opus is the most difficult to decode, for the same reason it is not implemented, for example, in bluetooth headphones

AVI files of that era used either 8 bit PCM or ADPCM.

the other did not exist.

There's a nice picture viewer for DOS. It supports JPEG and PNG. The VGA card has only 1 MB of memory, so it can do either 24bpp in 640x480 or 16bpp in 800x600. The program can load pictures of any size as long the decoded picture fits in RAM, which is 16MB. It can scroll over the pictures that won't fit on the screen. So obviously I've tried to give it some nice big pictures. It can load full size JPEG photos from an older camera. Decoding big PNGs is pretty slow on 486 though.

should try webp

it is the same blocky as jpeg, but I do not know if there is a viewer for it for such hardware

It would be challenging to find software with webp support that can run on DOS/Win95

maybe first you need to compile LIBwebp for DOS...

I have only basic knowledge of coding. I'm not sure how difficult would it be to port libwebp to DOS as memory management in DOS is pretty complicated.

I'm not a programmer too. but, because WEBP is written in C99, there shouldn't be any Legacy problems... in theoretical

The author of the PictView program ended up coding his own PNG implementation. Perhaps it was easier than attempting to port libpng to DOS.

no way... PNG is already old enough to be on DOS at one time... although most likely it was much more profitable to use JPEG due to what storage capabilities were then

In 1995, PNG was bleeding edge technology. MSIE that comes with Win95 doesn't support it yet.

Before that there were images formats using RLE compression like PCX or LZW compression like GIF or TIFF. Images in DOS usually used maximum of 256 colors.

JPEG was its own thing. It works completely differently compared to other image formats of the time.

I agree, it was magic, and complex calculations

i love this moment (open gif) https://youtu.be/gHUCAshJus4?t=1579

Libjpeg still has an option to produce output using only 256 colors

Decoding JPEG by those ancient computer is not particularly fast either.

I know but 256 colors means grayscale, right?

and not a fact that from view of information storage, maybe it wasn't profitable, because of the small sizes

it seemed to me that you are talking about the palette?

or Just 4:0:0

256 color patelled images were a norm in DOS. I don't think JPEG could support palette.

Grayscale yes, it's juts the 8bit Y channel alone.

> perhaps I need to change the translator, otherwise even it seems to me clumsily written

Although early VGA hardware couldn't do 256 shades of gray, it had only 6 bit DAC.

it seems to me that given the small size of the images of that time, the importance of pixels was much more important than creating noise on JPEG images, especially considering that people preferred to take 1 bit of the image or even 4 bits at most, the standard 16 colors. and that's why PNG just didn't make sense on machines of that time, and that's why I AM surprised that this format was made in the distant future, and is still relevant, even though it's already old, but legacy

The screen resolution was low, so images were small in size by nature, pixels were large.

it was a good time when people appreciated every pixel....

256 color palette was the lossy compression used for complex images at the time. 16 colors were used for simple graphics, GUI, icons, etc.

<:Google:806629068803932281>

Anti-aliasing was not used either. CRT did it automatically.

Not Anti-aliasing... is hardware blur

CRT color / shape magics are really interesting tbh

games used to pretty much exploit physical phenomena to make it work as good as they did

yes... and that's why there is such a thing as 4:2:2

because it was the horizontal stripes that were smoothed because of the display technology

and interlacing

(brr)

oh no interlaceee

but

interlacing is a problem of America and their NTSC technology. we had a half-frame display for the size of the entire monitor ...maybe it was better, not for me. the flicker I was noticeably more

I mean, it was like this, half of the screen is one frame, the other half of the screen is another frame

not sure what you mean, PAL and SECAM used interlacing too

I want to say that it was of a completely different type And did not tear the picture across the entire area, but only in the center at the point of contact of the half-frames, which means that the movement of objects looked Normal without breaks

half-frame? wasn’t the interlacing on a line basis in both cases?

https://upload.wikimedia.org/wikipedia/commons/1/1f/CRT_image_creation_animation.gif like this

I'm saying that the technology is different

right, but as far as I understand, not in terms of how interlacing works, or is it?

half frames

I have never heard of such half-frames

to begin with, the entire upper part of the screen, then the entire lower part of the screen

well the SECAM

that's why I've never seen any frame breaks. and these lines, because We was no such problem

<@604964375924834314>but this was a problem because the original NES console did not work for us. although we also have different voltage levels. and assigning the network frequency

and have never had similar problems... But, we had a **tearing **before it became known on PC <:HaDog:805390049033191445>

tearing yeas

I would like to add that the more SIZE of monitors became over the years, the worse such a half-frame display was, because if You is close to the TV, then the half-frame flickering will be noticeable, but if from far away, everything was great, but for Close interaction with the monitor it is better to use NTSC

Libjpeg has an option to do some color quantization to output the decoded image as a palette indexed buffer - this feature probably dates back to the days when most graphics cards were used in 256-color mode so this would be how to display an image.

Yes this makes sense.

cannot be... after all, JPEG cannot store palettes...? you definitely mean the palette and not the Truncation color format FFFFFF > FF

The conversion to 256 colors is done during decoding the JPEG.

JPEG has nothing to do with palettes, but libjpeg still has an option to convert a decoded jpeg to a palette image just because in the early 90s you couldn't do much with a full color buffer

meh crutch

well, I understood, nothing interesting, just transcoding after decoding, and as I understood it was done for each JPEG block

It's one of those things you would today no longer have in a libjpeg API

well, but there are old versions that no one seems to have erased, the legacy was killed, well, their Do

<:banding:804346788982030337>

8000 hz 16 kbps AAC vs OPUS vs MP3 vs VORBIS

and

PNG supports all sorts of color types and bit depths. I'm using most of them. The most common are 24bpp RGB and 32bpp RGBA. On the other hand, the least common are grayscale PNGs, in particular 4bpp and 2bpp grayscale are very rare.

PNG only supports what conveniently fits in bytes: 1,2,4,8, and 16 bit per sample.

Everything it does (filtering and entropy coding) is operating on bytes, and is not aware of samples and channels

That means only the 8-bit case actually has real prediction, for lower or higher bit depths the prediction doesn't really make much sense

In my opinion, the bit depth relevant for delivery is going to shift from 8-bit to 10-bit in the near future. For capture, 14-bit is currently common. PNG is not very useful for 10-14 bit — it can do it, by padding to 16-bit, but prediction and entropy coding will be pretty bad.

PNG was designed for content that was common at the time, to replace formats like GIF, BMP or PCX. Many Windows programs were using BMP as an intermediate format, to store pixels in memory. As BMP doesn't support grayscale directly, grayscale PNGs were usually encoded as indexed color with grayscale palette.

Additionally, 4bpp and 2bpp grayscale formats require images to use those exact 16 or 4 shades of gray, which is unlikely ta happen unless the image was created taking that into account.

png is no interesting, lossless webp is best for the compression today

WebP is great but it can't preserve the original pixel format. Lots of art was drawn with limited colors, especially artwork for games. PNG has legacy support, it can store graphics in the original form, like it was on the original CGA, EGA or VGA system, including the original color palette. Thus PNG can store losslessly and reversibly any of the old image formats.

I didn't understand what the problem was

For example, I have a 16 color BMP image, I can convert it to PNG and later, if needed, I can convert it back to the exact BMP file. If I convert the BMP file to WebP and then convert it back, I will get a BMP with 24 bpp colors. It has the same contents, but the information how the pixels were originally encoded, is lost.

It's not really a problem, and the feature is more like a side product than a primary intention. But it's nice to have.

After all, I can just store the original palettes separately and encode all images to JXL or WebP, later I can reconstruct the original format by assigning the pixel colors to palette colors.

yes, that's right, webp is not so universal

for some reason it reminded me of vorbis & opus

the good news is that Vorbis and PNG were made in the years when there were already thousands of scientists and technically knowledgeable people to take into account a lot of things and do something for the future

it just so happened that PNG and VORBIS have become legacy, but at the same time they have very flexible settings, which are generally they don 't in modern codecs

that is, a lot of technologies that could make new codecs even more versatile simply DO not

against this background, JPEG XL looks like a technologically-stuffed monster

which can generally DO everything that was ever needed from codecs

not really, if you only have very flat images then sure

as far as I know, in terms of compression ratio, WEBP beats PNG In all possible cases in general, any arts bit (8 bit RGBA webp), any type of image

it beats png but that doesn't make it the best

better not better

png isn't the only lossless image format

not "better", but "more versatile"

PNG is more universal

litlle more

yes, at least write your own. from this his support will not appear

what

no support = no format

what

???

is the only

it's not though lmao

and webp

and more avif

and....its all

i like it

Lossless JXL can do better than WebP, although it sometimes requires some tweaking. But that could improve when the encoder gets better.

16 bit PNG can be used to encode images in arbitrary bit depth . PNG provides sBIT chunk that indicates the original bit depth before the image was converted to 16 bits.

Yes. It just doesn't compress very well for > 8 bit.

sBIT can indicate which bits are significant, but it's a widely ignored chunk and in any case you need to give the padding bits some value, which will mess up entropy coding.

PNG has no context modeling, not even a very basic one to distinguish the high bits from the low bits in case of > 8 bit.

The main purpose of sBIT is that pixels in the original format can be restored. When upsampling pixels to PNG's native format, the empty lower bits should be filled by repeating the highest bits. Predictors in PNG are very simple, they work strictly on bytes, but they're applied to corresponding bytes in pixels. n-th byte of a pixel is predicted by n-th byte of the previous pixel.

Of course, lossless JXL is superior to PNG. There's no doubt. PNG is showing signs of its age.

Is libheif dead?

Why would it be dead?

libheif has pauses in development/maintenance. It happens, time to to time.

When I was experimenting with MPEG1 video, I've tried to encode 1080p video using ffmpeg to MPEG1 and see what will happen. To my surprise it produced a playable video file. I thought MPEG1 was limited to maximum size of 352x288 as MPEG2 was used for higher resolution.

profiles and levels are typically only really relevant for hardware implementations. software decoders tend to not care much about that, and just play anything that is a valid bitstream

remind me again, what's the recommended method to encode AV1, particularly for 3D? 👀

metod? I can only say that you need to use AOMENC, in ffmpeg the speed parameters are 4 or 3, you can’t do less(won’t be enough life)

I was having issue with AOMENC (it doesn't generate a new file or show any progress, just hangs for a while) but I tried lossless AV1 encoding in FFMPEG (libsvtav1) and my 320MB test video is now 8MB which seems *way* too good to be true.

does anyone know how to check if 2 video files are identical?

does svt even support lossless? I think it does yuv420 only...

I suppose you could let ffmpeg compute psnr, should be inf if they are identical

you can't just look...video

and aomenc be in ffmpeg

libaom-av1

ohh that's right. I'll try that one then

I guess it wasn't lossless, after all. unsurprisingly lol I used ```bat ffmpeg -i iA.mp4 -c:a copy -c:v libsvtav1 -preset 8 -crf 0 -metadata:s:v:0 stereo_mode=left_right oA_AOM.mkv ``` then ```bat ffmpeg -i oA_AOM.mkv -i iA.mp4 -filter_complex "psnr" -f null NUL ```

That seems quite lossy, psnr 29 is what you can roughly expect from mozjpeg at q60.

on a side note, about what we were discussing the other day > We are working with Twitch on the next generation of game streaming. AV1 will enable Twitch viewers to watch at up to **1440p 120 FPS at 8mbps**; a feasible bitrate that can reach most home-broadband and 5G users. > source: > https://www.nvidia.com/en-us/geforce/news/rtx-30-series-av1-decoding/ > The Twitch code leak did confirm it's being worked on too.

120 fps? Why that many frames per second?

I've seen streams with bad tearing because the game was rendering faster than the capture could support

Also if everything supports it, and the bandwidth usage is within limit, why not capture at a higher framerate? Better to discard when watching than when recording

More important than higher framerate then is to do proper vsync when capturing so frames are always single frames and not the top of one frame and the bottom of another.

I also don't get why a game would render at more than 60 fps — even 30 fps is probably sufficient w.r.t. human vision, though game mechanics ticks should probably be faster than that to assure good enough reactivity

It's mostly response time, I've got a 144hz monitor and you can definitely notice the "fluidity" compared to a 60hz one

They're at 300hz+ this year

it's more of the time gap between the last rendered frame and the monitor update cycle being shortened, when being able to render more frames, it closes that time gap more, and makes it feel more responsive and smoother

at least I think it's something like that

Yeah, the game main logic loop normally blocks until it's time to render a frame, so it can only process input then

At 30fps it can only poll the mouse 30 times a second, 60 times at 60fps, etc.

Also things on screen (enemies etc.) have their position updated more often, so they don't "snap" between positions as much

There is nothing preventing a game from polling the input at 120 fps and updating game physics at that speed, while only rendering a frame once every 4 ticks, so at 30 fps.

Displays with a higher refresh rate should be better in terms of ghosting and stuff like that, but I doubt that having content rendered at 120 fps and displayed on a 120 Hz monitor can be visually distinguished from content rendered at 60 fps on that same 120 Hz monitor.

Even 30 vs 60 fps I think is hard to actually see. It does make a difference at the capture side: a 60 fps movie will have frames with less motion blur (because exposure time has to be shorter) so it will look different/sharper, but I wonder if playing just the even numbered frames at 30 fps would really look that different...

I can easily distinguish 60 90 120 fps. What are you talking about

When all are displayed on the same display at the same display refresh rate?

I limit the fps in game engine settings. Is that a problem?

I am just talking about how much fps of image info our eyes/brains can process meaningfully. I think 60 fps is a safe upper bound for that (and probably even 30 fps is kind of enough; normal cinema movies are only 24 fps after all). Of course for many other things, more fps is better, including: - game engine/input speed: of course lower latency is always better - display refresh rate: faster refresh means less ghosting etc, so in general more accurate display - capture framerate/exposure time: less motion blur means sharper frames

refresh rate has nothing to do with ghosting

and movies are only 24 fps because this was decided over a century ago and many (old) film makers and film fans like this choppy look.

Well it depends on the display technology of course to what extent ghosting is a thing

Difference between 30 and 60 fps is actually pretty noticeable. Also higher framerate gives impression of higher video quality because there's more visual information.

I mean without applying lossy compression. Of course with the same lossy settings (say same bpp), a 60 fps version will have fewer artifacts than a 30 fps one.

It's not only the bitrate. Motion is an important part of video content. 1080p at 60 fps has higher perceptual quality than 4K at 30 fps.

Of course, in this comparison, 4K would be better in static scenes with little motion.

However, most videos would benefit from higher frame rate.

this conversation reminds me of this video lol https://youtu.be/hjWSRTYV8e0

It's confusing that games treat display fps and game engine fps as synonyms — yes, for gaming, the less input lag the better, obviously, especially for continuous inputs like mouse or VR movements. But that's about input handling (and how it affects game physics and game camera movement), not about graphics rendering.

Also I suppose most games don't actually produce an absolutely consistent framerate, but will aim at some fps and if there's too much stuff to render, the framerate will drop. So in that sense, of course 30 fps is not enough if there are occasional drops.

And another major difference between game engines and video cameras is that video cameras will include some motion blur when capturing (the exposure time is typically close to the duration of a frame, usually half of it or so), while game engines produce frames without motion blur so you basically need to render more of them and let the display ghosting and human persistence of vision produce the motion blur that makes it look smooth and natural.

I assume it's just as easy to render two frames than it is to render one frame that has motion blur, which may be why games do just produce more frames instead of trying to produce a consistent 30 or 60 fps with the appropriate amount of motion blur included

something something brain is 60fps but eyes are infinite fps

i tend to turn on motion blur in non twitch games

if i dont get at least 120 fps

Yes, there's no input lag issue in video. However, people have sense of motion. Despite it's impossible to see the individual frames, people can feel the smoothness of the motion.

one person asked to use svt and i explained him

Yes, and the difference between 24 fps and 48 fps can be seen (though it's somewhat subtle already imo). But I don't think more than 60 fps is useful from the perceptual point of view — assuming those frames are captured and displayed in a good way, including the natural motion blur etc, and without compression artifacts etc.

Yeah, I'm not sure how useful would be more than 60 fps for watching video, 50-60 fps feels pretty smooth. I don't watch movies very much but in sport, low frame rate is atrocious. 30 fps is very choppy, it feels like watching a cartoon.

120FPS is very useful <:YEP:808828808127971399>

I can't tell myself, because I don't have the equipment for playing 120 fps video.

I wonder if there is research showing in a blind test that people can see the difference between 60 fps and 120 fps. Maybe some people with exceptional vision can see it, but I suspect it's like sampling audio at 192 kHz: perceptually irrelevant (but of course it can still make sense to do it, in particular if you're going to slow it down in editing)

No, people have no trouble seeing >240FPS in research.

The differences do become lower, but they are still there.

Are you confusing persistence of vision, where flashing a light even for just a microsecond can still be perceived, with actually seeing higher framerates?

Perceived differences of actual higher framerates(IE, higher smoothness, less panel motion blur, etc).

Can you point me to research showing that? I have a hard time believing it

I am not talking about display technology framerate, that's another thing

i dont think i have excptional vision but to me personally its very easy to see the difference ¯\_(ツ)_/¯

Yeah. Once you go HFR, you can't go back.

When I went from a 60Hz monitor to 144Hz, it was an eye opening experience.

Here is an interesting actual study on the matter: https://osuva.uwasa.fi/bitstream/handle/10024/10196/UniVaasa_2019_Huhti_Juho.pdf

Even simple animations like application window minimize and maximize look smooth and realistic oh high refresh rate monitors. 60hz display feels choppy now.

Again, I am not talking about monitor refresh rate. I am sure higher monitor refresh rate is correlated with better perceived quality.

What I mean is a 120 fps video displayed on a 120 hz screen versus showing the even frames only, at 60 fps, displayed on the same 120 hz screen.

Any 60 Hz LCD display will effectively display frames slower than 60 fps because all LCD displays have some amount of ghosting: some traces of the previous frame (and even the one before that) will remain visible...

You gain nothing* from displaying 60 fps at 120 Hz versus 60 at 60. Ghosting happens with 60 Hz but also 120 Hz and the amount depends on the panel type and the specific model. But you could have a good 60 Hz TN panel with less ghosting than a 120 Hz VA one. Maybe ask Cloudinary to sponsor you one for research purposes. * the only exception is a less noticeable jelly effect: <https://testufo.com/scanskew#speed=2.0&direction=horiz&bounce=1&density=4&center=1>

other images av1 wikipedia italian edits

look is absurd

waste of time mostly

https://www.youtube.com/watch?v=zvTyWK3KYEc

watch at 480p

I would assume that even if the per-frame ghosting of the 120 Hz display is larger than that of the 60 Hz one, the ghosting over two frames (when each frame is duplicated) should typically be lower, no?

ghosting is typically measured over time and not as "per-frame ghosting" since pixels in transition don't care about when their next transition is happening*. The smoothness is merely affected once enough ghosting remains after one full refresh cycle. So ghosting becomes more important the higher your refresh rate is. *but they relax over time and turn towards white (that's why adaptive sync monitors can flicker when the refresh interval gets too large)

from pc point of views it's garbage always who's gonna bother reading all this

Tomorrow we'll talk about png

my brother in christ try this on a 120hz+ monitor. the difference ain't even subtle lol <https://www.testufo.com/framerates> <https://www.testufo.com/framerates-versus> ofc responsiveness is a major reason to go with higher FPS/Hz https://youtu.be/OX31kZbAXsA?t=1823 but it's irrelevant from a streaming viewer's POV. I think the main reason why we'd want higher FPS streams is just cuz it looks nicer, especially videogames. live action/film is different cuz most people are just too used to 24fps but more FPS=more details per second at the same resolution + more comfortable/easier eye tracking + less motion blur (especially if used with strobing) games aren't like cameras (which blend all the light in movement/panning captured during sensor exposure). instead, a shitty monitor with low hz will just blend 2 separate moments without the in-between blur, which is why it looks a lot worse

The choice of 24/30fps for video is a regional thing anyway, down to the utility frequency in use

Areas that use 50hz power used 24fps video, 60hz regions got 30fps video

don't you mean 25fps for 50hz? but yeah, europeans got the even shorter end of the stick when it comes to the refresh rate of the PAL standard even the audio in some retro consoles was slowed down to compensate, at at a lower pitch https://youtu.be/iefCo8RV84o

Ahh, that's what's confusing

TV is 25fps, Film is 24fps

"film destined for television is photographed at 25 frames per second." "Theatrical features originally photographed at 24 frame/s are shown at 25 frame/s."

aka analogue video standards are trash

In today's digital era, using "regional" frame rate is not needed anymore. TV broadcast is just a video stream, it could preserve the frame rate of the original material. It's time to purge those old standards. Modern TV's can take anything between 24-60 fps. If anybody is using an old CRT to watch TV, they need a conversion box anyway and the box could do the frame rate conversion for them.

they could, but keeping it same framerate is more stable re bandwidth use & frequency allocations

They could just stick to the same bitrate. If they're doing the framerate conversion, the effective amount of pixels remains the same anyway.

If they just duplicate each n-th frame.

And in case of frame rate reduction. Slightly lowered image fidelity would be less loss than removing each n-th frame.

Isn't that basically what interlacing does? Half the vertical resolution, then pack 2 frames into a single encoded frame

Interlacing is not exactly the same thing. In the interlaced video, only a half of the lines is updated each frame.

Interlaced video can be interpreted either as full frame rate in half resolution or half frame rate in full resolution but none of there is the right interpretation. It's basically an analog process, how CRT draws the image.

it still makes sense to support both 50fps and 60fps given that AC frequency still influences the flickering rate of artificial lighting

(in fact, even more so now that it's LED instead of incandescent)

the rationale has essentially moved from playback time to capture time

I suppose that professional LED lighting is powered by DC power, so there shouldn't be any flicker.

e.g. ad displays in football matches are notorious for this

LED displays are multiplexed, so they actually display just one row at the time. It's more practical to control only rows and columns of LED matrix than control each LED individually. That's different from LED light fixtures, where all LEDs are lit constantly.

Why can't power grids just standardize on a single AC frequency? Also: why can't they deliver power to end consumers in DC? Wouldn't that simplify a lot of things, considering nearly all appliances need to convert AC to DC first?

I guess, it would be too much work.

And different frequency is not a big issue anymore as most electronic devices use switching power supplies.

Many appliances can actually run on DC power. However I don't think the increased complexity of DC distribution would be worth it. Modern electronic devices don't care about mains frequency. Household appliances don't need to be able to work worldwide.

if one can buy their own lighting then sure, but there is definitely plenty of LED lighting “in the wild” that flickers

Also, DC converters of this magnitude would be massive.

There is a reason DC converters for power transmission are currently reserved for continental power stuff.

https://blog.kasson.com/gfx-50s/mitigating-lighting-banding-in-gfx-es-images/

https://www.reddit.com/r/lightingdesign/comments/9qoe95/how_to_top_led_banding_in_photos/

Low pwm frequency?

from what I understand, there isn’t even a need for PWM for it to occur

Actually I'm not against shooting video in framerate matching mains frequency. If it would yield a better result, it makes perfect sense.

I was talking about conversion from one framerate to another, which is unnecessary and may introduce unwanted stuttering.

Just wait until you see Japan, it has 2 separate power grids, one 50hz and the other 60hz

I just wonder how many appliances still need AC power and use the specific frequency for something (like CRTs did)

anything that has mostly spinning parts wants AC (and even better if they had triphase but oh well)

in particular washing machines

before switching power supplies were a thing, DC-DC conversion was a pain and AC-AC conversion was much much simpler

with SMPS, AC->DC is I believe not much different from DC->DC, you just add an extra full bridge rectifier and call it a day

(disclaimer: I don't *really* know any of this stuff)

Modestly modern CRT TVs and monitors don't care about mains frequency either. Even old basic VGA graphics used both 60Hz and 70Hz refresh rate.

in PAL regions, various games on the Sega Dreamcast would let you choose whether to run the game at 50 Hz (default) or try 60 Hz in case your TV supported it

mine did

60 Hz mode also used the full height of the display whereas 50 Hz had (thin) black bands

https://segaretro.org/List_of_50Hz-only_PAL_Dreamcast_games > The Sega Dreamcast was the first home video game console to support a "PAL60" screen mode - that is, to run video games at a 60Hz refresh rate without extra borders, as is the case with countries adopting the NTSC video standard. […] > > Unusually for its time, it is thought that 70-75% of PAL Dreamcast games have built-in support for 60Hz-compatible displays, and in many situations, 60Hz support can be forced through other methods. In contrast, very few PlayStation 2 releases offer a 60Hz refresh rate as an option, and support remained patchy for the Nintendo GameCube, Xbox and even Wii.

On most game consoles, support for PAL 50fps mode was an afterthought as console games were usually made for Japan and US in the first place. Game speed was usually tied to framerate, as a result games usually run slower on 50fps PAL systems.

Though sometimes you get better resolution on PAL versions.

I guess British computers could take advantage of the higher resolution, those were designed for PAL system after all.

Yeah, historically.

Though I think the Wii was the last console that had that difference.

As it isn't reay a thing with newer stuff that uses HDMI.

there are AOMENC developers or specialists here, I had extremely unpleasant artifacts associated with an incomprehensible distribution of key frames

11.886 sec -- the keyframe is SET on one frame before the real key frame

oh no

this is a problem throughout the entire anime that I have encoded, literally every keyframe is not a key

enable-keyframe-filtering no result, neither parameter 2 nor parameter 0

it looks like some kind of bullshit, maybe I made a mistake somewhere myself, wait a minute, I'll check SVT-AV1

well, what a stupid mistake with letters in ffmpeg doc