Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > VP9 and AV1

Reply
 
Thread Tools Search this Thread Display Modes
Old 20th May 2018, 00:06   #681  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,336
Luckily encoders can choose to not use some feature if its deemed too slow, especially since 8-pel is optional anyway and has to be signaled in the frame header.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 20th May 2018, 07:49   #682  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,556
Quote:
Originally Posted by nevcairiel View Post
Luckily encoders can choose to not use some feature if its deemed too slow, especially since 8-pel is optional anyway and has to be signaled in the frame header.
Ah, I'd been under the impression that it was a required feature. With a permissive spec, it doesn't matter.
foxyshadis is offline   Reply With Quote
Old 21st May 2018, 14:54   #683  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,746
AOM v0.1.0-9559-g59d2aa958
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 22nd May 2018, 23:31   #684  |  Link
TD-Linux
Registered User
 
Join Date: Aug 2015
Posts: 34
Quote:
Originally Posted by Shevach View Post
I remember a paper "Motion-Compensating Prediction with Fractional-Pel Accuracy", by B. Girod, 1993 . If we omit a mathematical part of the article and go to the conclusion part, it's written (in my wording): for blocks 16x16 of TV video resolution 1/4-pel motion accuracy appears to be sufficient.
Others are correct in that 1/8-pel is inherited from VP9. That said, there are many reasons why the results from that 1993 paper may no longer be valid - for example, VP9 operates on up to 64x64 blocks (and 128x128 for AV1), meaning that spending an extra bit for a more precise MV can potentially have a much bigger payoff than in a codec limited to 16x16 prediction blocks.

It's also a relatively cheap feature to add - the subpel filters are already pretty large, so it's just adding another set of taps.
TD-Linux is offline   Reply With Quote
Old 31st May 2018, 15:44   #685  |  Link
Shevach
Video compressionist
 
Join Date: Jun 2009
Location: Israel
Posts: 126
Quote:
Originally Posted by TD-Linux View Post
Others are correct in that 1/8-pel is inherited from VP9. That said, there are many reasons why the results from that 1993 paper may no longer be valid - for example, VP9 operates on up to 64x64 blocks (and 128x128 for AV1), meaning that spending an extra bit for a more precise MV can potentially have a much bigger payoff than in a codec limited to 16x16 prediction blocks.

It's also a relatively cheap feature to add - the subpel filters are already pretty large, so it's just adding another set of taps.
The problem is not spending additional 'bin' per vector component but how many comparisons are added for motion estimation in 1/8-pel MV precision?
Shevach is offline   Reply With Quote
Old 1st June 2018, 15:40   #686  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,746
AOM 0.1.0-9658-g265d15d46

New CLI option:

Code:
            --enable-fwd-kf=<arg>       Enable forward reference keyframes
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 1st June 2018, 16:36   #687  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by LigH View Post
AOM 0.1.0-9658-g265d15d46

New CLI option:

Code:
            --enable-fwd-kf=<arg>       Enable forward reference keyframes
What does it do?
mzso is offline   Reply With Quote
Old 5th June 2018, 03:07   #688  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,736
Quote:
Originally Posted by mzso View Post
What does it do?
Something like Open GOP or RADL, I'd guess.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 9th June 2018, 11:06   #689  |  Link
Mierastor
Registered User
 
Join Date: Nov 2010
Posts: 15
"The new V76 is essentially the successor to Arm’s previous high-end video block, the Mali-V61, which was announced back in 2016. Understandably the world of video encoding and decoding doesn’t evolve at quite as brisk a pace as GPUs, so Arm generally only revises their video blocks at about half the frequency. ... this processor will not include any support for the upcoming AV1 codec. While the bitstream specification for the eagerly anticipated codec was released a couple of months back, the timing was unfortunately after Arm had already completed the V76 RTL (never mind the fact that the specification isn’t closed yet). So it’s going to have to be the next video block after the V76 before Arm can include AV1 decode support. ... The very high encoding requirements of AV1 also mean that even after a decoder ships in a phone, we’re unlikely to see a full-featured encoder in a phone any time soon."

https://www.anandtech.com/show/12835...k-video-future
Mierastor is offline   Reply With Quote
Old 9th June 2018, 11:11   #690  |  Link
nevcairiel
Registered Developer
 
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,336
Its fascinating how the PR stunt from AOM managed to mislead everyone into thinking the bitstream is actually done. And months later, its still not actually done!
__________________
LAV Filters - open source ffmpeg based media splitter and decoders
nevcairiel is offline   Reply With Quote
Old 10th June 2018, 23:02   #691  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,986
I guess it's still "wait and see".

I'm still not sure why I'd use AV1 as an OTT operator delivering 4k content. I have to make HEVC for everything that exists today. Even if AV1 ends up being a bit more efficient (and this comes down to encoder implementation) it will still cost me a huge amount of money to encode my library in both formats, so why would I?

I guess it all depends on what clients end up supporting it.
Blue_MiSfit is offline   Reply With Quote
Old 11th June 2018, 09:00   #692  |  Link
Mr_Khyron
Member
 
Mr_Khyron's Avatar
 
Join Date: Nov 2002
Posts: 203
Socionext Implements AV1 Encoder on FPGA over Cloud Service
http://socionextus.com/pressreleases...cloud-service/
Mr_Khyron is offline   Reply With Quote
Old 11th June 2018, 10:27   #693  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,746
AOM v0.1.0-9742-g4e7b6f08f

Somewhat related to AV1: Google would like to patent (r)ANS, a speed optimized kind of Arithmetic Coding, specifically for its use in a video codec – despite Jarek Duda (its main inventor) having released this algorithm already to Public Domain in 2014...
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 11th June 2018, 19:35   #694  |  Link
iwod
Registered User
 
Join Date: Apr 2002
Posts: 756
Quote:
Its fascinating how the PR stunt from AOM managed to mislead everyone into thinking the bitstream is actually done. And months later, its still not actually done!
May be I am Old? Does any one remember On2? I mean if anyone who has been on Doom9 long enough should know. On2's "marketing", has been the same for years, even after it has been acquired by Google, and even many of the On2 employees left ( May be only the engineering left and not their marketing? ), their marketing hasn't changed a bit. And now it is Open Media Alliance, which is still pretty much ( Google + Mozilla ) + many others, with Av1, google is taking the majority of responsibility.

Quote:
I guess it's still "wait and see".

I'm still not sure why I'd use AV1 as an OTT operator delivering 4k content. I have to make HEVC for everything that exists today. Even if AV1 ends up being a bit more efficient (and this comes down to encoder implementation) it will still cost me a huge amount of money to encode my library in both formats, so why would I?

I guess it all depends on what clients end up supporting it.
One of the reason were HEVC Advance were changing OTT operator % per stream. Which was ridiculous. It wasn't until this march did they decide to stop this terms. And Youtube ( Google ) and Netflix has a huge incentive to stop using HEVC because of this. Not to mention HEVC listening is a bag of hurt. Google and Netflix want to provide uses AVC as base and use VP9 / AV1 for everything else. So as a consumer if you want Youtube or Netflix 4K content you will have to buy a STB that support Vp9 or Av1. They won't be doing ANY HEVC content. One of the interesting thing is both Youtube and Netflix cant be watched in China, and while many "new" or "info" likes to claim they are the biggest in the "world". That "world" does not include China. Similar to how Amazon or eBay likes to claim they are the biggest X in the world, when compared to China they are at least 4 - 5 times smaller in sales volume. Since China is in Region 2, they paid much less and most of the devices are already shipped with HEVC, in fact they are already streaming HEVC whenever they can.


Quote:
Somewhat related to AV1: Google would like to patent (r)ANS, a speed optimized kind of Arithmetic Coding, specifically for its use in a video codec – despite Jarek Duda (its main inventor) having released this algorithm already to Public Domain in 2014..
This has been mentioned multiple times, not sure if they are really against patent codec or against codec that don't use their patents portfolio.

Anyway I really hate this licensing terms and price discovery period. It has happened with AVC, and it is happening again with HEVC. The group or their members want to extract maximum outrageous prices, and waited for years of failure in market before they relent. I sometimes wonder if there are any more thing we could do with AVC to further improve it as an baseline.
iwod is offline   Reply With Quote
Old 12th June 2018, 22:30   #695  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,736
Quote:
Originally Posted by Blue_MiSfit View Post
I guess it's still "wait and see".

I'm still not sure why I'd use AV1 as an OTT operator delivering 4k content. I have to make HEVC for everything that exists today. Even if AV1 ends up being a bit more efficient (and this comes down to encoder implementation) it will still cost me a huge amount of money to encode my library in both formats, so why would I?

I guess it all depends on what clients end up supporting it.
Assuming the prior codec has full penetration,
codec_value=decoder_penetration * quality_@_perf advantage

So AV1's success is driven by decoder_penetration and quality_@_perf. The latter is important; 15% better at 10x encoding time doesn't really count unless encoding time was already more than fast enough. When live encoding is needed, or if encoding compute is a limit, MIPS/pixel is the limiting factor, and so AV1 implementations will compete with H.264, VP9, and HEVC at the same MIPS/pixel.

If devices wind up providing just H.264 and AV1, than the decision driver is whether the added compute, storage cost, and cache dilution is worth it. Even if compute was free, at a large scale supporting another codec is a huge operational expense and system complexity increase.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 12th June 2018, 22:58   #696  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,736
Quote:
Originally Posted by iwod View Post
So as a consumer if you want Youtube or Netflix 4K content you will have to buy a STB that support Vp9 or Av1. They won't be doing ANY HEVC content.
Netflix is certainly delivering UHD on devices that don't have VP9 hardware decoder support. And VP9 is a particularly challenging bitstream format for software decoding without high single-core CPU performance (this is much improved in AV1).

Tons of SmartTV and STB-like devices have way less CPU than a typical phone, so software decoding at 2160p is a non-starter. And replacement cycles are way slower for TVs and STBs than for phones and tablets.

Even if AV1 is an incredible success, companies would still have to deliver HEVC for legacy living room devices in 2025+. Heck, they will still have to deliver H.264 in 2025 for a number of device categories.

Personal computers and phones/tablets are by far the easiest markets for fast integration of new codec support. And lots of enterprises were still doing corporate content in WMV/VC-1 until they'd deprecated XP and Vista.

Massively improved technologies get the market moving, but even the best improvements can take a long time to become universally available in many markets.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 13th June 2018, 00:58   #697  |  Link
amichaelt
Guest
 
Posts: n/a
Quote:
Originally Posted by iwod View Post
So as a consumer if you want Youtube or Netflix 4K content you will have to buy a STB that support Vp9 or Av1. They won't be doing ANY HEVC content.
So your claim is that Netflix will cut off the 10s of millions (though probably actually more) of devices that are already streaming 4K HEVC from their service? Yeah, right... People are not going to buy a new TV or STB because of some silly codec war.

Last edited by amichaelt; 13th June 2018 at 01:00.
  Reply With Quote
Old 13th June 2018, 02:49   #698  |  Link
IgorC
Registered User
 
Join Date: Apr 2004
Posts: 1,315
It's not like VP9 isn't supported by any smart TVs.

I have bought 1,5 years ago LG smartTV. This year after upgrading the Youtube aplication it starts to support VP9 (stats for nerds reports it) and plays Youtube 4K without single drop.
IgorC is offline   Reply With Quote
Old 13th June 2018, 07:03   #699  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,986
Quote:
Even if AV1 is an incredible success, companies would still have to deliver HEVC for legacy living room devices in 2025+. Heck, they will still have to deliver H.264 in 2025 for a number of device categories.
Exactly, and this is why (unless CDN delivery costs are the vast majority of your opex) I don't see why a premium OTT VOD service would use AV1 anytime soon.
Blue_MiSfit is offline   Reply With Quote
Old 13th June 2018, 07:08   #700  |  Link
iwod
Registered User
 
Join Date: Apr 2002
Posts: 756
Quote:
Originally Posted by amichaelt View Post
So your claim is that Netflix will cut off the 10s of millions (though probably actually more) of devices that are already streaming 4K HEVC from their service? Yeah, right... People are not going to buy a new TV or STB because of some silly codec war.
Well I should have removed Netflix from that sentence, I am not sure if all of their Catalog are in HEVC already. But Youtube certainly aren't doing any HEVC at all. ( At least for now )

Yes, The point is, HEVC is already included in most TV or STB.

Quote:
Originally Posted by Blue_MiSfit View Post
Exactly, and this is why (unless CDN delivery costs are the vast majority of your opex) I don't see why a premium OTT VOD service would use AV1 anytime soon.
I wish we don't have the same problem with VVC.

Last edited by iwod; 13th June 2018 at 07:13.
iwod is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 03:43.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.