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 24th December 2019, 22:24   #1981  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,556
Good to see that VMAF-wise, SVT-AV1 has reached parity with vpxenc and x265, at least for 8-bit, and is only a little slower than x265. That's a pretty big milestone, and the codebase is still in a lot of flux, so there's probably some headroom.

BTW, SVT has supported y4m for a long time now, they just haven't updated their readme.The encoder guide in general is pretty outdated, but I want to see where the holiday cleanup goes before I send pull requests.
foxyshadis is offline   Reply With Quote
Old 25th December 2019, 05:56   #1982  |  Link
Nintendo Maniac 64
Registered User
 
Nintendo Maniac 64's Avatar
 
Join Date: Nov 2009
Location: Northeast Ohio
Posts: 447
Phoronix benchmarked SVT-AV1 v0.8 on a variety of CPUs (i7, i9, Xeon, Ryzen, Threadripper, Epyc):

https://www.phoronix.com/scan.php?pa...0.8-Benchmarks
__________________
____HTPC____  | __Desktop PC__
2.93GHz Xeon x3470 (4c/8t Nehalem) | 4.5GHz 1.24v dual-core Haswell G3258
Radeon HD5870  | Intel iGPU      
2x2GB+2x1GB DDR3-1333 | 4x4GB DDR3-1600       
Nintendo Maniac 64 is offline   Reply With Quote
Old 27th December 2019, 09:55   #1983  |  Link
ts1
Registered User
 
Join Date: Jan 2015
Posts: 71
SmilingWolf, have you made any tweaks to PSNRHVS? AFAIK by default cweight is set to 1, that's too high, set it to 0.125.
Or just use this already tweaked and improved version from libaom (only with this version first video must be original and distorted is second).
ts1 is offline   Reply With Quote
Old 29th December 2019, 10:36   #1984  |  Link
VincAlastor
Registered User
 
Join Date: Sep 2009
Location: Berlin
Posts: 173
Quote:
Originally Posted by Mr_Khyron View Post
Hello thank you for the update notification.

I have experimented with the SVT-AV1 encoder. Except that the cmd windows in staxrip freezes after 10.000 frames (but the encoding continues), there is a big problem.

I muxed the finished elementary stream in mkv. If I now play it and jump back or forward for more than 5 min, the player (with current LAVfilters) needs a lot of time until the playback continues.

If I download a youtube AV1 video in mp4 container the player doesn't need too much time to skip.

However, mp4box does not want to mux .opus streams, so I cannot really use the mp4 container. Do you have a solution?

ffmpeg -i -nostdin -f rawvideo -pix_fmt yuv420p - | SvtAv1EncApp.exe -i stdin -fps-num 24000 -fps-denom 1001 -n 548203 -w 1920 -h 816 -enc-mode 6 -q 48 -b


edit:
if i mux youtube AV1 mp4 into mkv, i also can skip fast trough the file. So is there a SVT-AV1 bug?

Last edited by VincAlastor; 29th December 2019 at 11:30.
VincAlastor is offline   Reply With Quote
Old 29th December 2019, 16:23   #1985  |  Link
fg118942
Registered User
 
Join Date: Aug 2018
Posts: 10
Quote:
Originally Posted by VincAlastor View Post
Hello thank you for the update notification.

I have experimented with the SVT-AV1 encoder. Except that the cmd windows in staxrip freezes after 10.000 frames (but the encoding continues), there is a big problem.

I muxed the finished elementary stream in mkv. If I now play it and jump back or forward for more than 5 min, the player (with current LAVfilters) needs a lot of time until the playback continues.

If I download a youtube AV1 video in mp4 container the player doesn't need too much time to skip.

However, mp4box does not want to mux .opus streams, so I cannot really use the mp4 container. Do you have a solution?

ffmpeg -i -nostdin -f rawvideo -pix_fmt yuv420p - | SvtAv1EncApp.exe -i stdin -fps-num 24000 -fps-denom 1001 -n 548203 -w 1920 -h 816 -enc-mode 6 -q 48 -b


edit:
if i mux youtube AV1 mp4 into mkv, i also can skip fast trough the file. So is there a SVT-AV1 bug?
Try adding "-irefresh-type 2" to the command line.
__________________
I use machine translation because I am not good at English.
fg118942 is offline   Reply With Quote
Old 29th December 2019, 22:53   #1986  |  Link
VincAlastor
Registered User
 
Join Date: Sep 2009
Location: Berlin
Posts: 173
Quote:
Originally Posted by fg118942 View Post
Try adding "-irefresh-type 2" to the command line.
thank you very much. That parameter was very helpful.

But let's hope we don't need it long, because open GOPs are more efficiently pretty sure.

Last edited by VincAlastor; 29th December 2019 at 22:59.
VincAlastor is offline   Reply With Quote
Old 31st December 2019, 05:42   #1987  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,556
Quote:
Originally Posted by VincAlastor View Post
thank you very much. That parameter was very helpful.

But let's hope we don't need it long, because open GOPs are more efficiently pretty sure.
That's a decoder bug; open GOPs don't need to decode previous GOPs, but dav1d is still pretty new. Anyway, open GOPs only give you noticeable gains if you have a very short GOP. Of course, SVT-AV1 has disabled scene detection and the default (-2) is at as close to 1s as possible anyway.

Just using a longer GOP will help much more.
foxyshadis is offline   Reply With Quote
Old 31st December 2019, 14:58   #1988  |  Link
Beelzebubu
Registered User
 
Join Date: Feb 2003
Location: New York, NY (USA)
Posts: 109
Quote:
Originally Posted by foxyshadis View Post
That's a decoder bug [..] but dav1d is still pretty new
No, it's a player or file bug. The player likely selects keyframes from the index (in Mkv parleance: seekhead), and for whatever reason, the muxer didn't add invisible keyframes (non-IDR in AV1) to the index (i.e. broken file), or the demuxer (player) doesn't recognize them and ignores them.

dav1d simply decodes frames in order presented by the player (along with some reordering etc.), it cannot seek further back by itself.
Beelzebubu is offline   Reply With Quote
Old 1st January 2020, 03:00   #1989  |  Link
VincAlastor
Registered User
 
Join Date: Sep 2009
Location: Berlin
Posts: 173
Quote:
Originally Posted by Beelzebubu View Post
No, it's a player or file bug. The player likely selects keyframes from the index (in Mkv parleance: seekhead), and for whatever reason, the muxer didn't add invisible keyframes (non-IDR in AV1) to the index (i.e. broken file), or the demuxer (player) doesn't recognize them and ignores them.

dav1d simply decodes frames in order presented by the player (along with some reordering etc.), it cannot seek further back by itself.
Friends and me experimented last week with SVT-AV1 files - more than one, in many players with different versions of filters and mkvtoolnix. The only thing what solve the problem was disabling open GOPs for now. Please try, hope you can show us a solution to use SVT-AV1 with default open GOPs.
VincAlastor is offline   Reply With Quote
Old 3rd January 2020, 15:46   #1990  |  Link
Spyros
Registered User
 
Join Date: Jun 2019
Posts: 16
LG To Unveil 2020 Real 8K TV Lineup Featuring Next-Gen AI Processor At CES 2020

LG announced that their new 8K TVs will support AV1.
Quote:
Not only do LG 8K TVs deliver Real 8K, they are also future-proofed to provide customers peace of mind with multiple ways to enjoy the Real 8K experience. The new models offer the capability to play native 8K content thanks to support of the widest selection of 8K content sources from HDMI and USB digital inputs, including codecs such as HEVC, VP9 and AV1, the latter being backed by major streaming providers including YouTube. LG’s 8K TVs will support 8K content streaming at a rapid 60FPS and are certified to deliver 8K 60P over HDMI.
As far as I know these are the first TVs with AV1 hardware decoding. I wonder if they will also support Opus.
Spyros is offline   Reply With Quote
Old 6th January 2020, 10:01   #1991  |  Link
birdie
Artem S. Tashkinov
 
birdie's Avatar
 
Join Date: Dec 2006
Posts: 335
Quote:
Originally Posted by Spyros View Post
LG announced that their new 8K TVs will support AV1.


As far as I know these are the first TVs with AV1 hardware decoding. I wonder if they will also support Opus.
These TVs will feature quite powerful SoCs and Opus is a very low complexity codec, so I see no reason not to support it.

Besides, YouTube started using it years ago along with VP9 and each device which supports YT must support Opus by default.
birdie is offline   Reply With Quote
Old 6th January 2020, 14:06   #1992  |  Link
Spyros
Registered User
 
Join Date: Jun 2019
Posts: 16
Samsung Electronics Unveils 2020 QLED 8K TV at CES

Samsung will also support AV1:
Quote:
At the same time, the QLED 8K lineup is among the first in the industry to support the playback of native 8K content. In 2020, consumers will be able to enjoy and stream AV1 codec videos filmed in 8K on QLED 8K TVs. All Samsung TVs in the 2020 8K line will ship with this capability built-in.

Quote:
Originally Posted by birdie View Post
These TVs will feature quite powerful SoCs and Opus is a very low complexity codec, so I see no reason not to support it.

Besides, YouTube started using it years ago along with VP9 and each device which supports YT must support Opus by default.
That's true, thanks to Youtube (and all the other streaming services that will follow) manufacturers have reason to support both codecs.

My previous message was based on seeing Samsung support VP9 but only Vorbis, not Opus (as of 2019), but I was wrong. This is only for the .webm container, Opus is already supported in .mp4, .mkv etc. I don't know if LG has similar documentation somewhere.
Spyros is offline   Reply With Quote
Old 6th January 2020, 15:20   #1993  |  Link
soresu
Registered User
 
Join Date: May 2005
Location: Swansea, Wales, UK
Posts: 196
Quote:
Originally Posted by birdie View Post
These TVs will feature quite powerful SoCs and Opus is a very low complexity codec, so I see no reason not to support it.

Besides, YouTube started using it years ago along with VP9 and each device which supports YT must support Opus by default.
I think he may have meant ASIC support?

As you say it's low complexity so it wouldn't require an ASIC to run it in a wall powered device like a high end 4K TV, but anything that reduces thermal output is welcome, it's annoying hearing a fan coming from a TV.
soresu is offline   Reply With Quote
Old 8th January 2020, 20:23   #1994  |  Link
Atak_Snajpera
RipBot264 author
 
Atak_Snajpera's Avatar
 
Join Date: May 2006
Location: Poland
Posts: 7,803
Does aomenc really support piping from ffmpeg?

my cmd line (simplified)
Code:
ffmpeg.exe -loglevel panic -i 1.avs -strict -1 -f yuv4mpegpipe - | aomenc.exe --cq-level=20 --cpu-used=3 --skip=0 --limit=1343 --kf-min-dist=0 --kf-max-dist=240 --output=1.ivf -
No ETA and crash at the end.
Atak_Snajpera is offline   Reply With Quote
Old 8th January 2020, 21:41   #1995  |  Link
poisondeathray
Registered User
 
Join Date: Sep 2007
Posts: 5,340
Quote:
Originally Posted by Atak_Snajpera View Post
Does aomenc really support piping from ffmpeg?
2pass definitely works ok for yuv4mpegpipe or rawvideo pipe - so the pipe works

I haven't done 1pass encoding with aomenc, but that suggests something wrong with the 1pass syntax

EDIT: try adding --end-usage=cq --passes=1 , that completes works here for 1pass

Last edited by poisondeathray; 8th January 2020 at 21:49.
poisondeathray is offline   Reply With Quote
Old 9th January 2020, 10:14   #1996  |  Link
utack
Registered User
 
Join Date: Apr 2018
Posts: 63
Quote:
Originally Posted by Atak_Snajpera View Post
Does aomenc really support piping from ffmpeg?

No ETA and crash at the end.
As of recently I have also had it crash at the very end, and the last few frames of the stream never got encoded
Current git version on linux
I will try to reproduce it and see where it started
utack is offline   Reply With Quote
Old 9th January 2020, 14:05   #1997  |  Link
Atak_Snajpera
RipBot264 author
 
Atak_Snajpera's Avatar
 
Join Date: May 2006
Location: Poland
Posts: 7,803
Quote:
Originally Posted by poisondeathray View Post
2pass definitely works ok for yuv4mpegpipe or rawvideo pipe - so the pipe works

I haven't done 1pass encoding with aomenc, but that suggests something wrong with the 1pass syntax

EDIT: try adding --end-usage=cq --passes=1 , that completes works here for 1pass
Ok thanks. Those two extra switches solve my issue.
Atak_Snajpera is offline   Reply With Quote
Old 9th January 2020, 19:50   #1998  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,738
Quote:
Originally Posted by soresu View Post
I think he may have meant ASIC support?

As you say it's low complexity so it wouldn't require an ASIC to run it in a wall powered device like a high end 4K TV, but anything that reduces thermal output is welcome, it's annoying hearing a fan coming from a TV.
Vorbis is quite low complexity. The CPUs in SoCs that can do AV1 decode will have ample power to decode Opus in a small fraction of available MIPS.

A bigger challenge can be if the SW audio decoder is integrated into the DRM system, which is sometimes hinky.

That said, xHE-AAC support is growing rapidly and has or will eclipse Opus's. Since it's just a new feature of the AAC porting kit, it'll get deeper integration in many cases.

I'm really impressed that we already have TVs launching with these specs for AV1. I'd love to know how much the SoC had to get bigger for that AV1 support. The extra transistors required for AV1 support is a really important factor in AV1's market viability.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 10th January 2020, 03:54   #1999  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,986
Do you usually have to encrypt audio? I usually get a pass on that, specifically since the lower security of the audio path mandates multi key (to avoid stealing the audio key and using it for the video) and that's tricky on a lot of platforms.

I want to use Opus today, but the spec for putting it in fMP4 and using it in DASH are not finished (last I checked)! I even whined about this but sounds like there's not a ton of motivation to finish it, which is a real shame.

Last edited by Blue_MiSfit; 10th January 2020 at 03:57.
Blue_MiSfit is offline   Reply With Quote
Old 10th January 2020, 23:42   #2000  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,738
Quote:
Originally Posted by Blue_MiSfit View Post
Do you usually have to encrypt audio? I usually get a pass on that, specifically since the lower security of the audio path mandates multi key (to avoid stealing the audio key and using it for the video) and that's tricky on a lot of platforms.
The need to encrypt audio is becoming less common overall. But some platforms have had perf issues mixing encrypted and non encrypted media at the same time. And encryption would be required for high quality streaming audio.

Quote:
I want to use Opus today, but the spec for putting it in fMP4 and using it in DASH are not finished (last I checked)! I even whined about this but sounds like there's not a ton of motivation to finish it, which is a real shame.
I think xHE-AAC has stolen Opus's thunder, since the licensing and code is free for any AAC licensee. Plus it outperforms Opus some.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner 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 09:26.


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