"New" as in still not widespread used of course.
I didn't read much in the last years related to codecs and came to this codec page by accident today...
AV1 instead of H264-H265 and AVIF hopefully instead of JPG sounds good.
Unfortunately these kind of changes don't come easily if at all mostly.
This is the Wiki page:
https://en.wikipedia.org/wiki/AV1
Here are some image comparision examples for AVIF too:
https://netflixtechblog.com/avif-for-next-generation-image-coding-b1d75675fe4
Oh man I don't like this. It's got the clarity/smooth algorithms built in that Topaz Labs made popular. They make things flat and lack detail while at a glance looking clean and crisp. To me it changes the whole artistic medium, from detail oriented to flat toned like a cartoon. The AV1 example image in that article fixes compression or noise in the original image, which could literally destroy shots in film where particles/dust is a intended effect.
Smoothing functions are by-default employed in both DTV boxes I've had and just never used them. They're earlier models of smooth features and it looks like mosaic filter applied at a small scale. Just terrible.
Any examples Jordan?
Found this:
https://squoosh.app/
From here:
https://reachlightspeed.com/blog/using-the-new-high-performance-avif-image-format-on-the-web-today/
I enabled AVIF support in Firefox. But got errors.
Hmm. I enabled it and tried AVIF but got bad results even with max effort and quality. Really bad banding and compressing blocking.
Hard to see in a compressed video but can barely see the added blocking from AVIF compared to original. It doesn't even appear to be adding any of the smoothing/clarity filters that the specification examples showed.
The comparisons on that site look promising.
Quote from: WAS on September 02, 2020, 10:09:38 PMHmm. I enabled it and tried AVIF but got bad results even with max effort and quality. Really bad banding and compressing blocking.
Hard to see in a compressed video but can barely see the added blocking from AVIF compared to original. It doesn't even appear to be adding any of the smoothing/clarity filters that the specification examples showed.
I can't really see it in the video Jordan. But i really doubt that this codec would be so bad when it looks like they all want to use it.
There are the other factors like royalty etc. But still...from what i read around, with similar file sizes this should be the same quality and actually higher.
If it is not, as you say, then it sucks of course.
I still can not see the AVIF comparison in Firefox. I will try with Chrome.
I looked ones more at your video. It might be that your AVIF comparison doesn't work too.
All other tests look finished and show the file size comparisons. But with AVIF it just keeps working in the lower right side.
I tried with Chrome. The look depends on the kind of image too as always.
Just a quick test. Chrome got errors and had to refresh the page etc. on the large file especially.
Still early to say something concrete but to me it looks promising
Quote from: Dune on September 03, 2020, 02:09:39 AMThe comparisons on that site look promising.
Did you tried it with Chrome Ulco? If you got no errors how did the large image look and did you get errors while testing?
I don't use chrome, just firefox and that looked good.
Ok. Thanks.
It just looks terrible. Smoothing people work into flat tones is just going to ruin representation of mediums. Literally a dusty attic scene would be flat atmosphere, no dust particles as they would be smoothed out like it does with the barn sky.
If it works like the Firefox version and is just jpeg like compression I'd be much happier. There is no filter effect with the website compressor, but their examples for video streaming at just bad. Topaz Labs is a curse on art, and not happy about it's type filters in a field meant to preserve a person's vision.
With topaz you can do a horrible composition in PS and run clarity and clean on it and it'll look like a digital painting like the barn preview.
I have a feeling that you are comparing a bad, filtered (Topaz labs?) approach to a general use case.
You can use very high compression in JPG's and say the same for example.
Here is another page with some examples. AV1 looks at least on par with HEVC (in those examples at least).
https://www.androidauthority.com/av1-codec-1113318/
JPEG does nothing of the sort filter wise.
Did you look at the specification images on the netflix blog from it in action on netflix?
While this is good for low quality, as it's better than blocks, the same is also employed in their high-bitrate example which is unacceptable of an art medium like film, and should be as close to the source as possible, which the JPEG accomplishes, albeit with some noticeable blocking. However, the detail of the scene is preserved. It would destroy partical grain effects. What if the barn is meant to look rustic and that smooth effect makes it look new. What if a lot with lot. Sand, dirty details, all will but smoothed and flat tone. In fact you can see blocking is still a thing, it's just being smoothed out with the filtering. The AI/Algorithm has issues with the tree branches on the right and tries to preserve them, showing the underlying blocked compression like JPEG.
Here is a side by side, you can clearly see the filter in action ridding the scene of detail and creating easily processed and streamed flat tones. It's in a ZIP because for some reason large PNGs have been failing in uploads lately for some reason.
Hmm...i think what we are talking about now is the subjective part about these kind of tests.
For me those blocks and banding in JPG's are unacceptable in these days.
We have seen here many threads about this problem as you know most probably.
Filtering or whatever, these kind of problems is good, if it is in an acceptable quality margin.
For me the blurry low bitrate image looks better then the low bitrate blocky JPG.
And with higher bitrate the AVIF just looks as it should too.
I understand you but my preference is just different.
I don't think that an image that does have artificial blocks in it does have more detail.
It looks just bad with blocks. But you may say it is just more blurry then blocky with AVIF...
Anyway... It is too soon.
I don't know if i will be really happy with this format until i see more and can test it as i want more deeply.
And by the way your file example is a comparison of AVIF to the original image.
You should compare it with the JPG to the original. The JPG is just blocky.
Source is of course the best comparison of what's actually happening and to pay attention to the details lost. I already mentioned you can look at the high bitrate example and see the same filter employed, and same lack of detail. So no matter what on Netflix with that current method of AVIF you get a cartoon smoothing effect. If that could be turned off like in the examples online using just AVIF compression that would be wonderful, as I don't want me image augmented when at high quality. This is shared by many who always turn off smart picture functions on their TVs which do the same.
And of course I know it's a matter of taste but the initial rebuttal seemed to be "I don't see it". Which imo is disconcerting to even having a discussion about it if those fine details aren't immediately noticeable being lost. I am always a knit-pick with details. Like no offense to Ulco but it drives me nuts all his rock formations are smooth for the most part. :o I see to much of the base noise formations.
For a comparable size of avif and jpg, the result of avif is very much better. If you compare original (left), with avif (right) and jpg (under), the jpg shows awful blocking for the same size, while avif does a pretty good job. Of course with high compression you loose detail, but that needs finetuning for specific saves.
I am not so negative about this.
Yes looks better to me too Ulco.
I found a version of Gimp with Avif support. I don't remember from where.
The file name is "setup_gimp_avif.exe". There are some plugins too around for Gimp.
These are still work in progress i think.
From just 3-4 tests i made it looks like if you don't go too far the files look good.
After a certain threshold the blurriness what Jordan don't like begin to change the look.
The JPG version kinda look more detailed (when high compressed like the Avif...but not very high).
But i think that is more because of the blockiness that feels like detail (but isn't) but actually gets worse in edges.
Testing is clumsy. I will wait more to see what happens in other software.
But a control to handle the filtering-blurines might be nice, especially for people like Jordan (no sarcasm) and maybe for different kind of images.
It's in progress, I figured, so I'll await any development. If it's good, it'll become a new standard, I guess.
Quote from: Dune on September 04, 2020, 02:20:56 AMFor a comparable size of avif and jpg, the result of avif is very much better. If you compare original (left), with avif (right) and jpg (under), the jpg shows awful blocking for the same size, while avif does a pretty good job. Of course with high compression you loose detail, but that needs finetuning for specific saves.
I am not so negative about this.
Even on my phone I see immediately loss in detail, and on a grungy door that's unacceptable. It's really an example why not to use it imo. ::)
If it is so great, why not run all your renders through it and destroy them? topaz has the same AI clean algorithms. Get rid of all that PT noise.... And details xD
Not to mention without the filter, it just saved JPEG artifacts anyway, only when the filter algorithms are used in conjunction. These features are part of Netflix's use of AVIF. Throw any gradient at it
"Even on my phone I see immediately loss in detail, and on a grungy door that's unacceptable. It's really an example why not to use it imo"
Jordan i think you are so accustomed to the way JPG images do look that you are a little biased against this.
Look at these resized and cropped images of the image Ulco posted.
Actually the Avif file does have more detail then the JPG one. And without those blocks.
I tried it with the other images there too.
All similar high compressed Avif files does have more detail then the JPG ones and without those blocks.
Try it yourself Jordan. You will see that the opposite what you are saying is true.
Until i tried this myself i was a little worried too about that filter.
At least in those examples Avif just looks better from a detail preservation point.
I think WAS compares the avif with the original, but that's not the point here. Of course there's loss. It's the comparison in same file size (!) between avif and jpg. And that's so much better.
No I compared to the detail in in the compressed images, or otherwise, being lost entirely. It's OK that you like the soap opera effect or smoothing filters, but a lot of people don't like it, and hate what it does to their movies, for example from consumer reports: https://www.consumerreports.org/tvs/turn-off-these-3-features-in-every-tv/
It seems maybe you're too trained to look for blocking rather than the actual substance of the images, it seems. A grungy door, rendered smooth and all surface detail removed, for example. This goes a long way with hair, facial features, etc, the whole reason we wanted blu-ray, and now more normal 4k-8k content. This is creating the illusion of "better compression" by simply hiding it with filters. If you go lossy, for speed, you'll get block compression. If you go high, you'll get more background/surface smoothing and edge reconstruction based on that lossy jpeg-like image. You can go lossless as well, though it takes awhile for some reason.
The smoothing filter has a spacial control, which would be cool if it was exposed in config on Netflix. But I'm definitely not the only one who does not like the smoothing filters of AV1/F and x265. Especially when it's so bad it's just a worse version of the soap opera effect that was ruining flatscreen TV sales because of default smart picture settings.
There was a phone released, I can't even remember the name of it, but it came with Avatar loaded on it as a special deal to show off the 1080p screen. What you got was a barely 1gb version of the movie heavily compressed with HEVC or something removing all the glorious detail so the handset could even run it at 1080p resolution. It didn't look the best and I remember other people upset on android central too.
Quote from: WAS 05/09/2020, 09:24:05
...
It seems maybe you're too trained to look for blocking rather than the actual substance of the images, it seems.
...
Now this is the part we are having a hard time to communicate. This was the exact thing i said that you have a bias.
The other parts in your last post does have reasonable things that i tend to accept too mostly.
What i not understand is, our discussion isn't about Topaz, Tv smart features or whatever (that i accept like you, as i said).
Our discussion is, if those images (Ulco posted from that site) does have more detail (Avif versus JPG).
The Avif files does have very clearly more detail.
And if you think that JPG does have in those images more detail i am really surprised.
In the jpg version many lines of the wood are just gone. And as a bonus you get blocky artifacts even.
If you still think that in these images Avif does have less detail i wont say anything more.
Even the PNG file sizes are hinting for what i am trying to say.
Higher detailed images do have in general higher file sizes too with PNG.
I wanted to see how Avif would look with an ordinary telephone camera shooted image.
This is our cat. I used Avif version of Gimp. As it is still a work in progress caution is needed of course.
But Avif and JPG aren't even comparable in this test.
JPG even with its 2 times the file size (JPG 28 KB versus Avif 12 KB) looks just horrible.
(I had to use the PNG version of the Avif file here of course)
I did different tests too. Some were so good for Avif i got suspicious if there is any problem related to encoding (there still might be).
Anyway. If this format will succeed i don't know. But it looks very encouraging.
And a better animated image format instead of GIF would be very nice with this too of course :)
You still keep defaulting to the most extreme compression methods which is negligent of my whole point. I'm talking about high-bitrate streaming. Even in Netflix full bandwidth example, the image is altered and still smoothed over with the filter. Netflix's current high-bitrate is great, there is no blocking, and I can see original film grain. This codec will rid that, inherently, by how it smooths all surfaces.
This isn't good. Many people do not like it. And that's why it's related to THE SAME type filters in smart TVs, the Soap Opera Effect, and HEVC, or topaz, ANY of them.
To sum up, I do not care about the extreme-case scenarios of heavy compression, I have a problem with it's filtering in general.
I also find using Web-Based JPEG compression to be negligent as well, it's been obsolete for almost 2 decades, contrary to what Planetside or Oshyan will have you do here. Lol Web exporting for JPEG/GIF has been advised against since I took web design in the early 2000s. It destroys images. Higher quality JPEG can perform well with little artifacts. If someone can't load a 500kb image, they just simply need to upgrade. This hasn't been a issue since even Dialup and the whole reason image-compressor algorithms were included in most dial-up services.
Microsofts already taken note of these concerns and Microsoft AVIF has the ability to disable filtering altogether like Mozilla can, where Netflix will have smoothing hard-coded on. And, of course, ironically, with that filter disabled, you get blocking just like in jpeg.
I also don't like that your argument seems to be based in convincing me the lack of surface details is better. Especially by using most extreme-case scenarios. As an artist I'm surprised you're not more concerned. If my films were released on Netflix I'd certainly be concerned. And with the way things are changing with film releases, people may have to specifically think about the cinematography just because Netflix will kill any subtle things like the tinkles of dust in a black background. I imagine The Mummy would look pretty bad with liquid-like sand rather than grained sand. Can't imagine what the sand storms would look like. Smooth clouds, after all that work in particle simulation. Or like if a website wanted to feature an artists work, but then heavily smoothed and compressed the entire thing which is not representative of the artists work. One of the main reasons a lot of people use PNG only only, to prevent web systems from compressing the images.
Quote from: WAS on September 05, 2020, 02:14:53 PM...
I also don't like that your argument seems to be based in convincing me the lack of surface details is better.
...
:o :o :o :o :o
I give up...
What i am trying to say is quite the contrary. You seem to read some posts quite different then intended. Or i am very badly wording them.
Avif does have more detail (in these examples) so i like what i see, because they have more detail. That is what am i saying.
Anyway. No need to talk more about something that will be or not around 2-5 years in the future.
Your definition of detail, especially preserved, is quite different from mine, and again, you're basing that on extreme compression, not expected viewing bitrates. Unless you're on early 2000s DSL, the examples you keep using wouldn't matter. It's the fact even the highest bitrate is augmented from source, let alone what's currently available thorugh netflix.
Here is my brown dwarf with as less compression as possible with CAVIF, it's completely destroyed. Compared that with the source, and max quality JPEG. I had to view the AVIF with firefox cause windows viewer doesn't open it, though I thought I read it supported it, including rotation.
AVIF is not for everything, especially large surfaces with fine details. You'll also see with gradients, and banding is just as common as JPEG (It's smoothed by the filter so not blocky in most areas), but apparently JPEG at max quality has far better results compared to it also applying the filter over compression, which is negligent at lowest compression.
https://drive.google.com/file/d/1eWRrqxV_fy_sQh6heYryMIK02XBnV7K0/view?usp=sharing
Additionally, compression with AVIF is slow, like I mentioned before, there's lots of tests about this too, making it a "higher-end" codec to begin with which puts older systems, and smart TVs at odd, so hopefully they keep legacy codecs. Here's a compression comparison chart, and AVIF is almost always slower: https://docs.google.com/spreadsheets/d/1TE5iLE08oV90EqOmFHnzBLwiPtQSs1XAvI3QfoMgKQM/edit#gid=0
This alone is why many codec methods are still not that widespread, decades on, because despite size, it comes down to decompiling or the act of compression by fabricators.
Also to note there is obviously some sort of AI/Sampler at play because even with lowest compression, it heavily compresses my dark brown dwarf image, down to 88kb. Sure great file size, but it ruins the image. I'd have to use lossless, at which point I might as well use JPEG so it's widely supported, and looks right.
I give up too.
Maybe actually try AVIF on your artwork Ulco and see how people like it lol It destroys my art, as I assumed it inherently would by what it does.
I honestly think maybe your guys's eyes are getting bad, you guys have both not noticed quality concerns in art on numerous occasions.
Anything that augments a image beyond compression is taboo in digital art. For laymen sure. Whatever floats their boat, but fundamentally not understanding what something does to a image and trying to argue about it with negligent testing instead of use-case examples, is just negligent, and dumb. Why not try it out?
The fact you don't understand how it achieves small file sizes by simplifying the image is astounding. And than try to argue it has more detail. Lmao
I have seen you more then ones being very stubborn and tunnel visioned Jordan.
You think that we don't know most of the things and concerns you are saying.
I disabled the smart settings in my plasma TV 7 years ago right when i bought it for example.
That is a different discussion i didn't want to go because what we talk here is only so so related. Yeah yeah filtering bla bla bla...
You can be sure that we know many of those things.
I used so many encoders that you might only maybe know them from Wikipedia.
One thing i hated when i was a teacher was the "that old teacher don't know anything at all attitude" as we probably all were when we were young.
But please you are not a teenager. You even didn't trust me when i said i use this since 30 years in an old argument. And didn't even tried-test it when i said it.
And no i can be wrong of course. And so can you too. That is ok.
There are old threads still around me being wrong. I have no problem with this. This isn't a race here.
"Why not try out?" is another of your insulting suggestions.
This thread is all about trying... how can't you see this?
You didn't trusted me and didn't do any test (on our old argument). But now you accuse us with this?
I don't say that this codec that i am just recently being fully aware is the next holly grail.
It does have its problems as all Lossy formats does have them.
But it just looks better with the same file sizes as JPG. And with smaller file sizes it looks much better then JPG.
If you need those smaller file size or not is up to the user.
But you can go much lower then JPG and still got a better image then JPG (Yeah subjective). This is just an option like in JPG so what?
When a codec can go so much lower then the other, there is a possibly that it is even better with higher file sizes.
So yes...testing with very small file sizes does have its merits too.
If you don't want compressed lossy formats then use PNG or whatever.
But i don't know in which land of the internet you live but mostly it is still JPG everywhere...unfortunately.
And i have done many "not negligent testing instead of with use-case examples" so i tend to think that i am not "just negligent, and dumb".
Why are you sometimes so aggressive? Really why?
Nor Ulco nor i (hopefully) haven't made any personal accusations.
It is just a bloody fu**ing codec. We even don't know if we will use it in the future.
Come on Jordan if you wrote all those things in a Facebook page where every kind of people is around
you maybe (actually not but :D ) would be right.
But here everybody does have in this or that degree knowledge about these things.
I got bored of this thread because of your post and didn't want to post more about it.
But after your last post i wanted to write my findings anyway. At least others might find them interesting maybe.
You can read what i think about it below more clearly.
I hope you can get over your "interesting"! attitude and see what i mean.
I used standard HD images in most of the tests and some 4K ones.
Some were ordinary photos, some were TV signal test image like images and some were digital renderings.
1* With same high file size images and even with a little smaller file sizes (low compression) AVIF looks the same as JPG .
2* When you begin to use higher compression AVIF files looks much better all the way.
3* But after a certain compression threshold blurring (Ohhh yes there is blurring. We are not stupid Jordan)
begins to creeping up in AVIF files.
Small noisy parts of the image begin to look more smooth (and yes the blurring is maybe in the high size images too but just not clearly seen i know).
4* AVIF files does have after that threshold more blurring. But it preserves lines and middle detail better then JPG. Small detail is blurred away.
But intermediate detail is blurred but still there. JPG files begin to show ugly blocking artifacts that are much much worse with smaller file sizes.
Small and intermediate detail is lost more in JPG when compressed higher (very small file sizes).
5* AVIF files can compress especially lines and text much much better then JPG. There isn't even any race here really.
Looks like AVIF could be very good for clipart and similar kinda illustrations and in cartoon movies where small noise especially is absent.
6* AVIF looses data in a better more pleasant way then JPG. That is the most subjective part of course.
If you don't like compression (lossy codec does what a lossy codec is designed to. It looses data...tadaaaa) don't use it.
One of your examples is suspiciously too low quality (small size).
I could use very different compressions, file sizes easily with Gimp.
Don't know what you used for that but it could be a problem for testing as you should be able to use different options for compression.
For example i got in some tests higher file sizes with AVIF files then the original when i used very small compression.
Funny thing is i didn't wanted to respond because i know it would be too long ::)
This is the Avif Gimp export window.
I'd leave it at that, Kadri. Clearly stated (as ever) ;)
:)
Quote from: Kadri on September 06, 2020, 03:38:04 PMI could use very different compressions, file sizes easily with Gimp.
Don't know what you used for that but it could be a problem for testing as you should be able to use different options for compression.
I used raw libraries with command line wrapper, CAVIF, which are actually what's being used to test the format, not GIMP. Like the trimmean results above, showing AVIF to be ridiculously bloated.
If I wrote this on a Facebook page where actual artists with good version exist, they would likely support me, as the art community never supported HEVC becaue of it augmenting images, and likely won't support AVIF either. This is why after well over a decade HEVC never got popular, despite file size and "quality".
What you perceive as better is not better. It's inherently simplifying the image. The fact you don't understand the codec, again, is the root of all your problems here.
If we were on Facebook, you'd be surprised how contradicted you would be by artists such as quote form Reddit: " ... as it completely kills small details and gives encode 'wax', smooth look to everything". That's just one post of many people hating the augmented look changing the medium.
And against you still keep fooling yourself. A max quality JPEG will inherently have more detail than AVIF because, again, what AVIF does. You cannot argue that away with opinion, it's a technical truth.
And honestly, at this point it may be insulting because I have no respect for someone who can't see what's in front of their eyes. A max quality JPEG hardly even alters the image like a AVIF, even again all of Netflix max bitrate exmaples. Every single one of them is has surface smoothing and blurring. That's inherently lack of detail, no matter how you want to argue it. Just cause there is an illusion of detail like a pixar film, doesn't mean it's there.
And no matter how you look at it, you cannot justify changing the image at all when it comes to peopels artwork and film. It's negligent. Another reason HEVC failed. Changing the look of film. If YOU are ok with that, that's fine, but don't act like it's normal in any artistic perspective. You would be deplorably negligent.
Another thing many users have proven is Netflix is fooling audiences by using destroyed JPEGs that do not match the bitrate compression of any software, photoshop, gimp, or otherwise. Playing up their format change with falsifiable hyperbole. Deplorable there alone. An 80kb JPEG doesn't look remotely as bad as their examples. That's why I was pissed when you would ever use a web based JPEG export to compare too, which no one uses. Even Netflix's lowest bitrate looks nothing like their examples, and much clearer.
The fact you argue for a format which changes the medium and look to terribly smooth and waxy looking sad of an artist. I'd new want to view and augmented format of a film outside their artistic cinematography.
Use some difference comparisons of encoding from source compared to JPEG. Least JPEG honors the source. And yes, pixel density and grain, is part of detail. Lol Silly Kadri.
Quote from: WAS on September 07, 2020, 02:25:11 PM...
The fact you argue for a format which changes the medium and look to terribly smooth and waxy looking sad of an artist.
...
I don't argue for Avif. I just compare Avif to JPG.
Every lossy format is bad for art. Avif just looks the better one compared to JPG with high compression.
With small compression (high file sizes) the discussion becomes mood.
I would love to use at least something like PNG whenever possible.
You seem to think we want to use Avif so badly just that we did a comparison here.
You assume things we didn't say Jordan.
:-X