AV1 is a royalty-free, open spec codec with completely open-source reference implementation backed by Google, Amazon, Netflix, Mozilla, Intel, Microsoft, Cisco, and pretty much everyone except Apple.
AV1 is now considered nearly ready for bitstream freeze meaning the codec spec will be frozen, and any video encoded with the reference implementation there on will be playable on any AV1 decoder in future, basically signalling it will be ready to use.
AV1 currently achieves better compression than H.265 or VP9 preserving the same level of perceptible detail and will likely get better as the implementation is further optimized.
It makes no sense since they don't have significant patents in the H.265 patent pool either (it's mostly Samsung and BBC tech).
David Wilson
Looks like they are waiting it out until a codec wins.
Carter Turner
only 7 years before Sup Forums supports it!
Henry Adams
they are going to create a proprietary codec that only apple can use just like Metal API
Hudson Thompson
AV1 hype also fuck apple
Jackson Thomas
wow a codec that is free of patents?
maybe we can see some foss firmware again if people stop fucking licensing MPEG-LA garbage.
Michael Mitchell
Well what's the fucking holdup
Evan Adams
>HEVC on suicide watch! >Netflix 4K and Babbylake is finished! >no wait Intel and Netflix are behind AV1 too?!
Samuel Ramirez
> safari doesn't support webm > safari doesn't support some other codecs thus some html5 content like flashtro.com doesn't work > now safari won't support AV1
Jack Perry
and on this note, BBC R&D is backing AV1 as well.
Brandon Reed
Probably this. Apple added webp support in the beta version of Mac OS so they are at least considering new formats.
Ryder Bell
Are they going to use the "webm" standard container for this or is there going to be another matroska derived container?
Gabriel Roberts
It's always WEBM
Logan Hernandez
Will only know if it's good if fansub groups use it.
Ryder Wood
You could just use a good browser.
Nicholas Smith
Wake me when there's hw-decoding for it.
Charles Garcia
Well with intel, AMD, and Nvidia all members, it shouldnt be long.
Cisco as well though they wont be doing consumer stuff.
Daniel Russell
It's not supposed to actually work, it's supposed to look pretty and make you feel smug inside.
What did you think you were buying apple for?
Landon Robinson
It'll be up again to Daiz and animefags to get the codec anywhere before normalfags notice it.
Nathaniel Rivera
Ho much better compression are we talking here
Carson Mitchell
Twice as good as H.265 is the target. It's a next-next-gen codec.
Isaac Barnes
Only 8 years until the scene adapts it, only 6 months until the anime scene uses it
Jonathan Baker
That's always the target and it's never feasible in real-world scenarios. Every time someone says some magical new codec will cut filesizes down the only thing that ends up getting cut down is quality.
Lincoln Adams
Don't be religious, if you look at the facts Mozila's Daala was already edging ahead of HEVC. Now with Google, Cisco, Netflix and all the other big players behind it if you think the quality is going to regress you're not being honest with yourself.
Nicholas Price
Where can I get a copy of the reference encoder? I want to see actual comparisons to hevc
Liam Jenkins
Actually, what happened to Daala?
Why have they not made a push together instead of a new one? If Daala was not good enough they could have made something like what was done between Mantle and Vulcan.
Grayson Williams
No idea about AV1, but Daala has one referenced at their home page:
Daala is being partially reused in AV1. Daala itself is actually more of a research project than a single codec nowadays.
Luis Young
>Actually, what happened to Daala? Daala, VP10, and Thor have all been absorbed by AV1
Evan Harris
Daala is using some very innovative technology to avoid patent trolling by the MOTION PICTURE EXPERT GROUP and all the HEVC ADVANCE people. But as a result it's hard to combine most of Daala's technology with traditional codecs, and using it as a base for the next big codec is taking a big gamble.
Some of the Daala tech is being lifted into AV1 but it takes time and the results are just on par in terms of quality. Mostly it's good for avoiding patent litigation.
Oliver Peterson
It's all open source, but with all the experimentations going on in AV1 master at the moment it's not a great time to test the encoder.
Go ask in #aomedia @ freenode if you really want
Angel Robinson
What's happening to daala and thor now? :'(
Nathaniel Ross
serious question, why does this exist when x265 is getting more and more traction these days ?
Charles Howard
I don't really care about how great of a time it is to test. It's near bitstream freeze so it can't be all that different than the final spec.
If it takes forever to encode or crashes or something I'll obviously be forgiving, but if it's worse quality than hevc at the same size for all but a few cherry-picked examples, I'll be critical of it.
Because it's royalty free and supposedly a better format than h265
Liam Phillips
I guess this has nothing to do with AV1 itself, I just get a little annoyed whenever I see "twice as good compression", because it gives distributors like Netflix an excuse to bitstarve their streams even more than they already do, thus quality will suffer, but they'll pretend it's some incredible new development anyway and write a blog post patting themselves on the back.
Jacob Scott
...
Jeremiah Morgan
>I don't really care about how great of a time it is to test. It's near bitstream freeze so it can't be all that different than the final spec. Last I checked the quality is there, but the encoder with all experimental features enabled would be tremendously slow (as in orders of magnitude too slow.) because of the large parameter space to explore.
Alrighty then. Not sure what's the current situation with what features are and aren't merged and behind what runtime flag, though.
Please post back when you have results. I'd like to compare what you find with AWCY results (which are all public).
James Torres
>AV1 currently achieves better compression than H.265 or VP9 Any decent source on that? I haven't seen any test of recent builds yet, only some 6 moths old stuff when it was still almost pure VP10 code.
It's not really a hard thing to believe. It compresses better, but the tradeoff is that it currently takes *way* longer to encode. That's supposed to improve with time, but so was every other previous codec
Carson Rodriguez
Currently encoding a snippet of Big Buck Bunny 1t 1080p from yuv420 to av1 Encoding at 2.4fps on my shitty laptop, so it's gonna be a bit before I have results
Jaxson Young
Given who the stakeholders are, half would be more than happy with encoding times being slow if the decoding time or energy consumption of the hardware implementation is better. The other half if the compression to quality ratio is better.
Leo Watson
Second pass is getting 2.5 frames per minute and has an ETA of 162 hours
Welp
Hudson Hill
Did you run it with --enable-av1 --enable-experimental --disable-daala-ec ? Otherwise it's gonna be shit
Chase Reed
LMFAO
NO PERFORMANCE [ecks dees internally]
Daniel Torres
>--disable-daala-ec
Really?
Nathan Clark
>on my shitty laptop see this is why I got a 5820k.
I dont always need it, but encodes go muuuuch faster.
Nathaniel Gray
It's currently not quite ready, so yeah
Kevin Evans
Huh. Maybe I'll hold off on the HTPC NUC then, wait for AV1 hw support.
David Diaz
None of these are runtime options and av1 is the only available codec and is used by default
John Flores
Why would they add bloat that doesnt run on embedded cpu processes?
Juan Evans
If you want an HTPC to last 5-10 years, wait at least a year, maybe a bit longer.
HDR 4k with AV1 or H265 will be the primary content for the next decade, so waiting for good hardware decode support for both is a good idea.
Brandon Jenkins
that's why people dont usually try to do intensive video encoding on shitty laptops
Mason Williams
I'm encoding a 30 second clip. I could do it with x265 in about a minute.
Jaxon Reyes
Those are the awcy config that map to runtime options
Aaron Jenkins
not with the CPU you couldnt, maybe with a hardware encoder.
Jack Gonzalez
Kabylake/Apollolake has decent hardware decode for H265, correct?
Brody Nelson
have you done a side-by-side comparison of encoding to similar bitrate and resolution in each codec?
Landon Cox
For H265 yes, but that looks like it will only matter for big movie/TV releases.
No one wants to pay for H265 royalties so the vast majority of internet content will likely end up AV1, youtube, netflix, amazon video, etc.
Evan Perez
The AV1 encoder is far from being finalized, so that would be a pretty silly comparison.
Today there are JPEG encoders that get double digit % gains and the format has been frozen forever ago.
Landon Powell
Sure, but what's the vast majority of "free" internet content going to be? Pirates don't care about patents.
Leo Jenkins
it would be silly, yes, but the other user was implying they knew how that comparison would turn out and i was curious
Elijah Ward
If the groups care about file size and quality above all, then AV1 is better both on objective metrics and perceptual tests.
If they care about fast encode, H.264 will always be good enough since connection speeds keep getting better.
Landon Cox
AV1 or H.264 likely.
H265 just doesn't appeal to most people in that scene.
Austin Bennett
No I wasn't. I was implying that my laptop was not the reason for it taking a week to encode a 30 second clip. No matter what quality, that's ridiculous when I can do a normal encode with a normal codec in a few minutes
Xavier Nguyen
>AV1 was pure VP10 So it's basically >shit >That's always the target and it's never feasible in real-world scenarios. Every time someone says some magical new codec will cut filesizes down the only thing that ends up getting cut down is quality.
>Because it's royalty free and supposedly a better format than h265 Just like vp9 was great
>Daala is being partially reused in AV1. >Daala itself is actually more of a research project than a single codec nowadays. >Daala, VP10, and Thor have all been absorbed by AV1 >Some of the Daala tech is being lifted into AV1 but it takes time and the results are just on par in terms of quality. Mostly it's good for avoiding patent litigation. >Now with Google, Cisco, Netflix and all the other big players behind it if you think the quality is going to regress you're not being honest with yourself. and that's why >Given who the stakeholders are, half would be more than happy with encoding times being slow if the decoding time or energy consumption of the hardware implementation is better. The other half if the compression to quality ratio is better. it's shit
Nathaniel Sullivan
Is this a spambot?
Connor Morales
>>the encoder with all experimental features enabled would be tremendously slow >That's fine. I expect that and am not testing it for speed
Blake Hughes
>No matter what quality, that's ridiculous when I can do a normal encode with a normal codec in a few minutes Holy shit are you trolling? We explicitly warned you that the current state is experimental and extremely slow while they develop a new RD model, what the fuck did you expect?
This is why we can't have nice things.
Ayden Barnes
>my glorious effort comment >spam go kill yourself with your bleeding edge streaming video
Levi Sanchez
You're not making any sense. It's not even wrong, you're just rambling.
Angel Perez
So do you work for the movie industry then?
Nathan Harris
I already did... with ur edge :^)
Hunter Long
I'm not blaming it for being slow, I'm just saying I can't actually complete the test because I'm not going to wait a week for it.
Stop being so defensive, christ.
Noah Gonzalez
>2027-10 >effortposting kys
Bentley Perry
I'm just saying, what did you expect.
Benjamin Powell
I expected it to take maybe 5-50 times longer than x265, not 1000
Ayden Wright
its exacerbated by the fact you're using a shit laptop desu senpai
Thomas Gomez
>the encoder with all experimental features enabled would be tremendously slow (as in orders of magnitude too slow.) >I expected it to take maybe 5-50 times longer than x265, not 1000 lrn2read
Bentley Cook
No, it's not. Learn how to ratio
Jace Robinson
even a quick wiki search will tell you they adjusted their royalty fees
It really puts it in line with x264 royalties, and that is widely adopted, so what is the problem ?
Leo Brown
Oh no, you wish it were that simple.
Xavier Morgan
That's just HEVC Advance rates.
They have to pay MPEG LA and HEVC Advance, and I've seen threats of a 3rd patent pool. At that point, not knowing what rates might be like in the future, businesses willl wash their hands and walk away.
Businesses HATE risk, and x265 is risky.
David Bailey
>puts it in line with x264 royalties
are you high user?
The maximum cap on x264 devices per year was $5M
Now it's $20M?
Adrian Russell
>serious question, why does this exist when x265 is getting more and more traction these days ? x265 still has no traction kek also, av1 btfos it in general
Cooper Robinson
Okay so, I'm encoding a 10 second second clip at 480p on my laptop that will be done in another 15 minutes, but I've also gone ahead and built the encoder on a server with a xeon and gave it a 10 second clip at 1080p and that has a 2 hour ETA.
So results forthcoming.
Brody Price
You know, you could also check the AV1 project's own daily results reports.
They'll probably have more scientific measurements than whatever you're going to come up with.
Bentley Murphy
They'll also probably have biased measurements where they encode shit that plays on the strengths of their codec
Zachary Bell
Not really, no, that would be useless. They're trying to make a good codec, not to pat themselves on the back.
AFAIK they test on some standardized videos and on a bunch of Netflix samples.
Josiah Wilson
>AV1 project's own daily results reports. link?
Brayden Davis
It's the site made by the Daala guys and adopted by the Alliance for Open Media: arewecompressedyet.com
Ian Collins
>AV1 currently achieves better compression than H.265 or VP9 preserving the same level of perceptible detail Sauce?
