How come there is no information about this anywhere?
Apparently AOM/Google are already breaking compatibility with AV1, they are reopening the specification and making a changed format. Supposedly to incorporate changes requiested by hardware partners for decoding support.
If AV1.1.0 decoders in GPUs/phones don't support AV1.0 it is going to suck but at least there is not much content out yet. But it would be nice if they publicly announced this instead of being silent.
Other urls found in this thread:
download.opencontent.netflix.com
diff.pics
twitter.com
a revision a year keeps the cattle in fear
>How come there is no information about this anywhere?
Because it's being discussed in private mailing lists exclusive to AOMedia members.
Nothing is set in stone. Most changes approved until now seem to be backwards compatible, but the possibility is still there if it benefits hardware implementations.
If you hang around in relevant FreeNode IRC channels you'll get some tidbits.
Yeah but in the meanwhile, a format that might likely become dead is encouraged for broad use.
Bu I guess it would be a bad PR so it is understandable to keep it under wraps.
who cares? only goolag use it anyway
Nothing (HW) got release, so I don't see much problem since SW can deal with this kind of change.
I just find it weird since those companies are members of aom, so seems like they're only working on shit now. At least they need to comply and release something in less than a year.
>Using AV1
There's your mistake.
It seems that if you make an AV1 encode now, it possibly won't be compatible with later AV1.1.0 decoders, which might include all or most of the AV1 hardware.
Youtube and Netflix are fine with this because they can reencode everything. They only stream stuff instead of selling it and don't want anybody to keep the files.
For piracy/home use it is something you want to be aware of though, if this compatibility breakage happens.
>AV1 encode now
>For piracy/home use it is something you want to be aware of though,
Because you basically can't with the godawful state of AV1 encoders.
It's not really relevant since it's not deployed anywhere.
Breaking the specification freeze at this point is fairly safe. At most it would be an inconvenience for the people working on software encoders.
those just need small changes, it's a PITA for hardware guys. So either the AOMs are dumb or it is actually the hardware guys that want this - more likely I think.
But it probably means some delay in their availability, because the effective bitstream freeze got pushed back by at least half a year.
could be a result of finding some things were not as ideal as they thought once they actually got someone to spend real time working on silicon
Yeah I wouldn't but it seems people actually do encodes.
based
Totally Microsoft move, Google.
More like Google, they have a tradition of making one video format after another, pushing the spec out early without proper time for testing and then immediately jumping on a next thing.
What?
DEAD
ON
ARRIVAL
Codecs can work out multiple AV1 versions for back compat, only old encodes won't get future hw accel. it's no big deal as not all h264 profiles got hw accel etc.
AV1 will be harder on CPUs to decode though, so software fallback in TV sets and plastic boxes won't be easy.
It shouldn't be a big deal to people who has enough horsepower to encode AV1 today.
Was mentioned by user in previous thread a weak earlier.
It's probably the hardware folks who requested that in the first place. Not a big deal considering encoders are painfully slow and nobody except google and netflix is using it right now.
Oh, I missed that.
Would Jow Forums be interested in some content getting encoded at slowest speed for demonstration purposes?
Seeing this now only.
encode some anime pls
youtube already has a lot of live action videos
Any anime in particular? I guess it has to contain action, be grainy (so a film transfer). Multiple bitrates have to be tested for. Obviously encoding ripped blu ray sources would be best. Is there any anime that has worthwhile 4k scans?
AV1 is still too new for anybody to be using it for anything but testing.
What changes are they making? Will it reduce quality?
rav1e or aomenc?
Aomenc. I assume aomenc offers most quality.
It seems there are some additional restrictions (which is why it will be incompatible) that make hw decoding easier.
That potentially can lower quality but impact likely won't be big.
Yes, even though rav1e is making progress.
Then good luck user. You'll be encoding at around 0.01-0.05 fps for 1080p footage.
everything is making progress, it isn't really relevant that it does, when the journey is 3 years long.
two questions
>will it affect youtube-dl
>will it affect IDM's videograbber
Of course it wont. How the hell would a bitstream change affect a tool that downloads files from the web?
I don't know, user. It's still nice to know that a second actively developed encoder is out there. Nobody wants to repeat the whole libvpx fiasco after all.
I have several machines that are better than standard PCs so I'll split the encoding between several.
10+ years back when Xiph and freetards of the internet lead propaganda war pretending that *theora* is supposedly fit to substitute H.264, you kept hearing about "making tremendous strides" all the time. Man the memory still kind of pisses me. As if AVC encoders didn't have any development then. Ahum. Those were the heydays of x264 ffs. The crazy kids there actually kept raising the bar all the time then, so the reality likely was that theora was getting farer and farer from the goal in fact, those times, because it had much less development.
Anyway, to explain it better - of course making proogress is good. But the problem is that the encoder is at the start, it's very unfinished, onyl now starting to add basic stuff like adaptive keyframe insertion, ffs.
It's not like making progress is something unexpected and pleasantly suprising. They have to be making a lot of progress daily just to be able to make a reasonably finished and polished encoder in 2-3 years. That time can't really be shortened, it's fuckton of manhours to make an advanced encoder.
Basically, just have to stay realistic and not adopt pipe dream expectations.
Sorry for old fart rant.
I wish I could say the same. Mine are all worse.
To be clear I also have high hopes for Rav1le myself, it's not like I want to shit on. Since Google never really impressed with libvpx (= libaom now), we gotta hope in Rav1le. Hopefully the devs will be able to absorb the knowledge in this field achieved by x264/x265 without getting burned by the same lessons and having to reinvent all that again.
And you think x264 was started and finished in a month? Or that it generated optimal streams using every available tech defined in the spec after a single year of development?
That's the point my dude. It started in 2004(!). It got really got in 2008-2009. Then decent improvements 2010-2011 or 2012. Look at how many years it spans. And that was reasonably fast development too.
It got really good in 2008-2009 I mean.
I don't disagree with you, this takes time. That was exactly my point.
I need to update this again since I've added a 3570k and a 3570.
Nice collection, damn. What do you have that for, do you run animu encoding internet cafe?
>Basically, just have to stay realistic and not adopt pipe dream expectations.
I don't see how my remark about rav1e making progress implied any high expectations. To be more clear, my intent was to say that they are making progress quality-wise (not just focussing on encoding speed) and that there isn't such a glaring quality difference as there used to be just a short while ago.
Yeah of course. I meant that they are probably still far from general usability.
I tried in august or september and it was quite bad.
>They only stream stuff instead of selling it and don't want anybody to keep the files.
I don't think anyone is selling AV1 right now or at any time soon because of the compatibility and not just because of the possible issue with 1.0 and 1.1, but at the market as there's no smartphone/smarttv with AV1 so they can consume the content. Also, that would only be a problem if we talk about physical media, because digital you could as download in another format (and they can update it as well).
For piracy you can always (re)encode from source or on-the-fly if anything. I don't think that's a problem, since we would be talking about hw decoder on smarttv/tvbox again, as software decoder should handle well if you're playing on an attached PC.
Which is also funny because they delayed like three times / more than a year.
This delay can be good, maybe they can finish/promote image formats so we can finally ditch jpg/png and fucking gif already.
It's not even fucking out yet. of course shit will break.
looool you're fucking retarded, it'd take years even on threadripper CPUs to encode only a small library of movies.
Nobody is replacing their libraries with AV1 when it hasn't even been released yet.. and even if they WERE they can do exactly what you just said, reencode their library, just like netflix and youtube can.
No, the format was supposed to be finalized in march (when they announced bitstream freeze) or at the latest in june (when it actually happened for real - ye olde marketing).
>minor version increment
>breaks backwards compatibility
Google's doing great with those diversity hires.
I don't think it's that bad, if you use shittier settings.
Personally I do prefer using good ones with HEVC, though.
I'd say AV1 is probably more in the hands of necbeards. Although maybe those are managed by various mbas and such who decide what to do, like, let's push the finishing of the spec one year earlier, that will make us look so good.
Another reason to stick with ITU-T formats like H.265.
Can someone link to some AVI HD videos? I'm curious how hard my computer will choke decoding in software.
*AV1
I wish we had format that would be developed with MPEG/ITU profesionalism but at the same time benefiting from the giant freetard mindshare VPx/AV1 receives.
>HEVC decoding in ffmpeg grossly under-optimized 5 years after the format launched and while it is being used in the wild
>fine with everybody
>100 people rush to write code for a second AV1 decoder like lemmings although the encoders will be ready in 2 years, it's only relevant on youtube-type sites and nobody really needs support particularly pronto
>internet cheers wildly
download.opencontent.netflix.com
Get the mp4 ones IIRC, I think the IVF ones were borked
Not really. Nobody except companies wed to the AOM are even going to be making that kind of call for a couple of years.
The licensing is by FAR the largest reason people are cheering for AV1 over HEVC. It's not surprising that people are excited by the fact that a free codec might finally come out in front of a proprietary one.
HEVC is free too, when you pirate and/or use free software players. Big corps like netflix and google have huge enough profits to absorb any costs they would incur.
There appears to be a horrible memory leak when playing those in mpv.
i tried a bunch of weird codecs back then
theora, snow, vp7... (and yea, even vp7 claimed to be better than h264)
>I wish we had format that would be developed with MPEG/ITU profesionalism but at the same time benefiting from the giant freetard mindshare VPx/AV1 receives.
basically "i wish x264 existed"? you know x264 implements H.264, and MPEG standard, and is a foss codec, with lots of 'freetard mindshare', right? it's existence coupled with the shit licensing of H.265 is why nobody really cares that much about H.265
Me too. VP7 was also quite slow.
the one thing I gave up on was Dirac. The options were way too incomprehensible, I had no idea what to try to tweak and it was not very promising so I skipped it.
even in x264 times, lots of attention was diverted to dead stuff like vp8.
AVC as format could be improved a lot - as HEVC shows.
>encoding 1080p video with libaom
>11 seconds done so far
>started 2 hours ago
>stock FX 8350
pray for me
nobody in the piracy world touches a format that isn't tired, tested and well supported. it took them years just to move from a .mp4 to .mkv container, if that's any indication of how slow things move in that world.
seconds done so far
>>started 2 hours ago
you wot? on a 8 core cpu? fucking google.
youtube has av1 encodings available for some videos already. they're probably just testing though. pic related.
well the bitrate is kind of low, 14 seconds now
psudocode for a super simple AV1 encoder that is not much slower than the reference encoder.
while(calculate_average_rms_error(encoded_video, source_video) < desired_threshold) {
encoded_video = generate_random_bytes(desired_file_size)
}
>using av1 trash over based hevc
lmao
MPEG LA please leave.
mad you'll be missing out on all those patent royalties, aren't ya?
>How come there is no information about this anywhere?
Because it has no impact given that nothing has been implemented in hardware yet.
Goddamnit you fucking googles, a breaking change is a MAJOR version change
>AV2 before AV1 is even a thing
take the glue out of your mouth
>implying that matters
Jewgle releases like 10 versions of chrome a year
>HEVC is free too, when you pirate and/or use free software players. Big corps like netflix and google have huge enough profits to absorb any costs they would incur.
Netflix and Google does not 'absorb' those costs with their profits, they forward that cost onto the customer.
More importantly, smaller sites like for example Jow Forums (ever heard of it?) can't carry those costs, which makes video standardisation on a royalty free codec absolutely great.
>video
>Jow Forums
I wish, but the lack of VP9 support on webm's here made it pretty clear Jow Forums is not for videos, we use silent VP8's as a Gif workaround not as a true video supported platform
Why would they upgrade to vp9 when they knew AV1 was coming ?
VP9 existed when Jow Forums first got webm's we have been asking since day one, long before AV1 was announced. it isn't like moot has 4cha re-encoding the videos, he just has to host them, the browsers play them natively the same way they play vp8's.
What does it matter if the encoder is FOSS? The standard is the problem. AVC is also licensed only more straight forward, not like the clusterfuck with HEVC where you have 3 patent pools.
AV1 wasn't widely known about for most of the time VP9 has been available.
Also, it's still going to be a while before AV1 becomes widely used, and in the mean time VP9 would give a significant improvement over VP8.
libaom only offers tile-based multithreading yet. Not sure what defaults ffmpeg is using, but he likely doesn't fully utilize his CPU.
How much better is hardware video decoding at saving battery charge? That's pretty much the only benefit I can see for HW; I already play all videos on my phone in SW mode, since sometimes there are issues with HW and it's just easier to do everything in SW instead.
>VP9 existed when Jow Forums first got webm's we have been asking since day one
VP9 had very little hardware support back then, also VP8 is less demanding which helps if you need to fallback to software decoding.
Unless Jow Forums is dead by the time AV1 hardware rolls out late 2019/early 2020, it will support AV1, everything on the web will.
I hope so, I still don't see why we can't have VP9 in the meantime, literally just get rid of the stream parser and you're done
My old iPad doesn’t even play AV1. Doubt this thing will last long.
anime piracy on the other hand
>they forward that cost onto the customer.
Not if the customers don't let them. Youtube in particular, but Netlfix too if we were able to pressure them with cancelling subsciptions and the like.
AV1 complexity is huge so in its case, probably substantial.
Or even legit anime releases. A small company called Animagia recently licensed the Chuunibyou movie in Poland and released it with VP9 encoding as a .MKV, and now they want to try AV1 for their new movie release in January.
Ugh, VP9 for anime.
10-bit VP9, with VP8 used for the web streaming version. I mean, he didn't have much of a choice - trying to sell anime in a country like Poland pretty much means running on razor thin margins, and paying the codec fees certainly wouldn't help.
It's true that for a small guy, it is probably like that. I just hope it doesn't artifact like in the old times. Does 10bit fix the banding and blocking it caused? IIRC there isn't much if any psyrdo in libvpx to keep texture.
See for yourself, on their first release, Amagi Brilliant Park: diff.pics