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 February 2020, 14:19   #41  |  Link
utack
Registered User
 
Join Date: Apr 2018
Posts: 63
Quote:
Originally Posted by foxyshadis View Post
It uses the same speed on both passes, no fast-first-pass. That doesn't seem ideal, the consensus is that speed 7 or 8 is fine for generating statistics.
I figured that yes, thank you for confirming it
utack is offline   Reply With Quote
Old 4th April 2020, 05:25   #42  |  Link
stax76
Registered User
 
stax76's Avatar
 
Join Date: Jun 2002
Location: On thin ice
Posts: 6,837
In case somebody is looking for a GUI maybe try StaxRip 2.1.0.3 Beta.

https://staxrip.readthedocs.io/intro.html#download
stax76 is offline   Reply With Quote
Old 7th April 2020, 13:39   #43  |  Link
birdie
Artem S. Tashkinov
 
birdie's Avatar
 
Join Date: Dec 2006
Posts: 337
Quote:
Originally Posted by stax76 View Post
In case somebody is looking for a GUI maybe try StaxRip 2.1.0.3 Beta.

https://staxrip.readthedocs.io/intro.html#download
Not really, SVT-AV1 is still considerably worse than x265 --preset slow: at least ten times slower and blurrier (at the same bitrate).
birdie is offline   Reply With Quote
Old 10th April 2020, 05:45   #44  |  Link
stax76
Registered User
 
stax76's Avatar
 
Join Date: Jun 2002
Location: On thin ice
Posts: 6,837
Quote:
Not really, SVT-AV1 is still considerably worse than x265 --preset slow: at least ten times slower and blurrier (at the same bitrate).
I just add the features people ask for, the encoder support was completely done in 4 hours, the problem was that this encoder writes progress without new line characters and dotnet and staxrip couldn't handle it, took a full day to support because I had to do an own version of a somehow complicated piece of dotnet code, it was OK doing it because I like to learn about threading, streams and text encoding and particularly everything console related. Apparently not all codec developers have a talent for console topics, x265 shines here, it followed the excellent x264 and the talented nvenc author rigaya also followed it, for GUI authors it's a big deal but it helps users greatly as well.

This encoder might not be very interesting for everybody but it's often interesting to just take a look at something and play with it, staxrip makes it effortless to try tools, it supports over 190 tools and everything just work, without a GUI some things can be complicated to set up for users like this encoder, vapoursynth, qtgmc etc. It's very transparent what staxrip does because the command lines and scripts can be seen in the app and there is also a very clean log file.

Last edited by stax76; 10th April 2020 at 05:50.
stax76 is offline   Reply With Quote
Old 10th April 2020, 10:03   #45  |  Link
Tadanobu
Registered User
 
Join Date: Sep 2019
Posts: 37
I certainly appreciate the SVT-AV1 support. Thank you very much ! If you someday have time to add libaom too, that will be awesome.
Tadanobu is offline   Reply With Quote
Old 10th April 2020, 10:24   #46  |  Link
stax76
Registered User
 
stax76's Avatar
 
Join Date: Jun 2002
Location: On thin ice
Posts: 6,837
Quote:
If you someday have time to add libaom too, that will be awesome.
It was supported with 106 GUI options (auto generated from --help output...), not sure why it was disabled, I can take a look.

https://github.com/staxrip/staxrip/b...ding/AOMEnc.vb

Last edited by stax76; 10th April 2020 at 10:27.
stax76 is offline   Reply With Quote
Old 10th April 2020, 14:01   #47  |  Link
Tadanobu
Registered User
 
Join Date: Sep 2019
Posts: 37
Yeah ! It looks like most of the options are already here. If it's not too much trouble re-enabling it, please give it a go. No hurry tho.

Also, new options are available. I haven't checked all and every lines but the most recent options like --tune=vmaf or --enable-fwd-kf are missing. But, well, we can still pass them manually. Thanks in advance.
Tadanobu is offline   Reply With Quote
Old 19th April 2020, 22:49   #48  |  Link
MZ/X
Registered User
 
Join Date: Jan 2004
Posts: 15
Initial version of xAV1vfw v0.82 is released on http://mpxplay.sourceforge.net

The GUI is similar to the other (x264,x265) VFWs (video for windows codec).

XAV1vfw uses the svt-av1 encoder library (git 20200419), with several limitations, like
- it doesn't work properly under Windows 7 (at me), only under Win10 (wrong CPU detection)
- it's very slow, even in the fastest encoding mode (full HD encoding, 4 cores, AVX2, ~5 fps)
- AVX512 support is unofficial/experimental, disable it (on the GUI) at encoding problems (crash)
- some GUI settings have no svt-av1 support yet (encoder may rejects these), like YUV, Tuning and Profile non-default settings

Enabling the Debug log level to Full can display some additional svt-av1 errors and infos.

You can follow the updates on my sf.net page
https://sourceforge.net/p/mpxplay/discussion/219198
or on my FB page
https://www.facebook.com/mpxplay

Last edited by MZ/X; 19th April 2020 at 22:51.
MZ/X is offline   Reply With Quote
Old 20th April 2020, 02:51   #49  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
Version 0.8.2 released, download at https://github.com/OpenVisualCloud/S...ses/tag/v0.8.2

Quote:
Encoder
- Initial Super resolution support
- Mode decision rate estimation enhancements
- Altref improvements
- Manual prediction structure support
- Enhanced RDOQ support
- Improved warp motion
- Improved documentation and help menu
- New command line parameters
- Fix non-multiple of 8 resolution video corruption for 8bit
- Improved multi-stage mode decision support
- Added multi-stage motion estimation support
- Chroma mode decision optimizations
- Added 16bit pipeline support
- Added Mode decision non-square partition weights
- Enhanced qp-scaling assignment
- Memory optimizations
- Added AVX512 Optimizations
- Added AVX2 Optimizations

Decoder
- Improved multi-threading support
- Memory optimizations
- Updated documentation
- Loop filter flow optimization

Encoder and Decoder
- Encoder-Decoder-Common folder structure improvements

Build and Testing
- Bug fixes
- Improve CI
- Improve Unit Test Coverage
- Address C vs asm mismatches
- Support C only builds (for platforms other than x86)

Last edited by foxyshadis; 20th April 2020 at 02:54.
foxyshadis is offline   Reply With Quote
Old 22nd April 2020, 20:04   #50  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,750
Quote:
Originally Posted by foxyshadis View Post
Version 0.8.2 released, download at https://github.com/OpenVisualCloud/S...ses/tag/v0.8.2
Looks like some good additions!

I've wondered about SVT-AV1 having RDOQ, which is something I'm mainly familar with from x265. Is the SVT implementation derived from x265 or the HM? Does libaom have something similar by another name?
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 23rd April 2020, 03:43   #51  |  Link
TomV
VP Eng, Kaleidescape
 
Join Date: Jan 2018
Location: Mt View, CA
Posts: 51
Quote:
Originally Posted by benwaggoner View Post
Looks like some good additions!

I've wondered about SVT-AV1 having RDOQ, which is something I'm mainly familar with from x265. Is the SVT implementation derived from x265 or the HM? Does libaom have something similar by another name?
Are you asking about SVT-HEVC or SVT-AV1?
TomV is offline   Reply With Quote
Old 23rd April 2020, 10:51   #52  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
Quote:
Originally Posted by benwaggoner View Post
Looks like some good additions!

I've wondered about SVT-AV1 having RDOQ, which is something I'm mainly familar with from x265. Is the SVT implementation derived from x265 or the HM? Does libaom have something similar by another name?
I'm not exactly sure what the implementation's provenance is, although it started out almost the same as the aomenc version, but in aomenc and the older versions of SVT it goes by trellis, and was renamed in SVT in 0.8.0 to match up with SVT-HEVC and current industry norms. In libaom, it's on by default up to encmode 5, and manually turned on/off by --disable-trellis-quant.
foxyshadis is offline   Reply With Quote
Old 24th April 2020, 22:24   #53  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
Version 0.8.3 released, download at https://github.com/OpenVisualCloud/S...ses/tag/v0.8.3
Quote:
Encoder

- Presets optimization

Build and Testing

- Bug fixes
- Xcode build support
"Presets optimization" has some tweaks for preset 0, and a lot of changes for preset 1, to include slower settings and make smaller files, possibly not at the same quality. Anyone who uses preset 0 or 1 after this will want to retest their settings.
foxyshadis is offline   Reply With Quote
Old 24th April 2020, 23:44   #54  |  Link
quietvoid
Registered User
 
Join Date: Jan 2019
Location: Canada
Posts: 570
Has anyone managed to make --film-grain work yet?
There's no difference in output whether it's set to 0 or 50 for me, last tried on 0.8.2.
quietvoid is offline   Reply With Quote
Old 27th April 2020, 10:45   #55  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
Quote:
Originally Posted by quietvoid View Post
Has anyone managed to make --film-grain work yet?
There's no difference in output whether it's set to 0 or 50 for me, last tried on 0.8.2.
In the current implementation, you have to generate a noise profile, and then feed it in as a second pass. aomenc works the same way. (The implementation is almost 100% identical to aomenc.)
foxyshadis is offline   Reply With Quote
Old 27th April 2020, 12:55   #56  |  Link
quietvoid
Registered User
 
Join Date: Jan 2019
Location: Canada
Posts: 570
Quote:
Originally Posted by foxyshadis View Post
In the current implementation, you have to generate a noise profile, and then feed it in as a second pass. aomenc works the same way. (The implementation is almost 100% identical to aomenc.)
Right, but the 2 pass VBR in SVT is completely broken for target bitrates. It also looks worse than CQP.
aom allows you to pass a grain table file, I don't think SVT does.
quietvoid is offline   Reply With Quote
Old 11th May 2020, 19:23   #57  |  Link
quietvoid
Registered User
 
Join Date: Jan 2019
Location: Canada
Posts: 570
I've tried using --film-grain 10 with 2 pass CQP, unfortunately it still doesn't add any film grain on decode.
Results in a lot of
Quote:
Error in freed returnVal 0
quietvoid is offline   Reply With Quote
Old 29th June 2020, 00:08   #58  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
0.8.4 released: https://github.com/OpenVisualCloud/S...ses/tag/v0.8.4

Code:
    Bug fixes
    Improve CI
    Improve Unit Test Coverage
    Address C vs asm mismatches
    Fix static analysis warnings / errors
    Add address sanitizer
    Various ffmpeg patch fixes
Newly included is an ffmpeg build, to make it easier for ffmpeg users.

Unfortunately fixing film grain didn't make it in, that's among the priorities for the next release.
foxyshadis is offline   Reply With Quote
Old 7th September 2020, 09:19   #59  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
0.8.5 released: https://github.com/AOMediaCodec/SVT-...ses/tag/v0.8.5
Code:
Relicensing notice

    Change the outbound license from BSD+Patent to the AOM license / patent

Encoder

    Added tpl support to adaptively change lambda and quantization parameters within the frame
    Added multi staged hme support
    Quality speed trade-offs tuned to VOD use cases
    Added first level non-optimized support for 2pass VBR and CRF
    Added combined cli two pass support with options for stats being written to a memory buffer and a specified file
    Added non square partitioning optimizations
    Improved lambda generation

Build and Testing

    Bug fixes
    Improve CI
    Improve Unit Test Coverage
    Address C vs asm mismatches
    Fix static analysis warnings / errors
    Add address sanitizer
    Fix symbol conflicts with libaom and libvpx when staticly lined to ffmpeg
Quality tuning and overall improvement, fewer crashes, initial 2-pass, and fixing symbol conflicts are the highlights.

Edit: New URL, since the move to AOM is complete.

Last edited by foxyshadis; 7th September 2020 at 22:57.
foxyshadis is offline   Reply With Quote
Old 3rd December 2020, 12:50   #60  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
0.8.6 minor release: https://github.com/AOMediaCodec/SVT-...ses/tag/v0.8.6

Code:
Encoder

    Further quality-speed tradeoffs tuning for VOD use cases
    Improved TPL support within 1-pass and 2-pass CRF moode
    Continued non-optimized support for 2pass VBR and CRF
    Align kernel nomenclature to prefix svt_aom for kernels brough from libaom to avoid symbol conflicts

Build and Testing

    Bug fixes
    Improve CI
    Added CI support for gitlab
    Improve Unit Test Coverage
    Address C vs asm mismatches
    Fix static analysis warnings / errors
    Add address sanitizer
    Fix symbol conflicts with libaom and libvpx when staticly linked to ffmpeg
Missing is that this currently won't build on BSD or ARM due to a last-minute change. I imagine it'll be re-released with a hotfix soon.
foxyshadis 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 14:22.


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