AVS2 ... a new Chinese standard counterfeit?

[LigH]

The media-autobuild_suite reports to support an "AVS2" codec via pkuvcl/xavs2.git (https://github.com/pkuvcl/xavs2.git); possibly a successor of a previous xavs codec? Quite interesting, the repository reports a "AVS2-P2/IEEE1857.4 video coding standard", so it sounds official. I'm curious how it gets sorted between AVC, HEVC, VC-1, etc.

[nevcairiel]

Its a real standard, even though its mostly a rip-off of HEVC, just like AVS was a rip-off of H.264. Still fascinates me how they managed to get such a blatant rip-off to be certified by the IEEE, but some influential people is probably all it takes.

[LigH]

:o Not quite smart, the separate CLI encoder. It seems not to explicitly support Y4M, rather skip a header:

xavs2[i]: Input : Input sequence, YUV 4:2:0
xavs2[i]: InputFile : - Same as `Input`
xavs2[i]: InputHeaderLength : If the inputfile has a header, state it's length in byte here

So if you want to test, possibly prefer it as ffmpeg codec.

To let MABS build an ffmpeg with AVS2 support, add to build/ffmpeg_options.txt (if you selected ffmpegChoice=1 and a matching license model):

--enable-libxavs2
--enable-libdavs2

It's only linked to a Win64 ffmpeg build by MABS. So I guess it may not even build as Win32 target.

[hajj_3]

AVS2 isn't that new. The amlogic S905X2 chip has a hardware decoder for AVS2, it takes a 1.5yrs or so from ratification to shipping a hardware decoder. Codec comparisons have been done for a while, it is basically the same quality as x264: http://www.compression.ru/video/codec_comparison/hevc_2017/

[Monarc]

Codec comparisons have been done for a while, it is basically the same quality as x264: http://www.compression.ru/video/codec_comparison/hevc_2017/
They tested uAVS2. xavs2 is a another encoder.

[hajj_3]

AVS3 Baseline draft spec is out: http://www.avs.org.cn/English/avs3/doc/AVS3-Group%20Standard.pdf

[tnti]

AVS3 Baseline draft spec is out: http://www.avs.org.cn/English/avs3/doc/AVS3-Group%20Standard.pdf

The Chinese version of the draft was released very early.
It was released in May 2020.
It may be a variant of VVC. On this basis, new coding tools have been added.
uAVS2 and xavs2 are not products of a group.
They are developed by different developers.
The developer of xavs2 is also the developer of SVT-AVS3.

This is the code URL of the uavs team.
https://github.com/uavs3

This is the code URL of the xavs2 team.
https://github.com/pkuvcl

AVS3 encoding currently only has online test videos in Tencent Video China.

[benwaggoner]

What is the actual use case of the AVS codecs? Only in China? I don't think I've ever seen any of these files used in the wild.

[LigH]

Pretty much that, as far as I can read from the Audio Video Standard (https://en.wikipedia.org/wiki/Audio_Video_Standard) Wikipedia article. They use it mainly for digital TV, and it has a patent pool and industry alliance and IEEE standard application and ITU recognition... this article on streamingmedia.com (https://www.streamingmedia.com/Articles/Editorial/Featured-Articles/The-China-Problem-Government-Sponsored-Formats-Are-Incompatible-115128.aspx?CategoryID=429) discusses possible goals to have a separate standard, not compatible to MPEG. The obvious: Control and censorship is much easier this way. There is another paper (https://www.tandfonline.com/doi/abs/10.1080/09537325.2011.592270?src=recsys&journalCode=ctas20) but it seems to require a membership to be read.

[tnti]

What is the actual use case of the AVS codecs? Only in China? I don't think I've ever seen any of these files used in the wild.

It is rare on the Internet. Basically it is only used in China.
Even in China, it is only used on digital TV transmitted by satellite. AVS2 is basically not used for ground waves.

Pretty much that, as far as I can read from the Audio Video Standard (https://en.wikipedia.org/wiki/Audio_Video_Standard) Wikipedia article. They use it mainly for digital TV, and it has a patent pool and industry alliance and IEEE standard application and ITU recognition... this article on streamingmedia.com (https://www.streamingmedia.com/Articles/Editorial/Featured-Articles/The-China-Problem-Government-Sponsored-Formats-Are-Incompatible-115128.aspx?CategoryID=429) discusses possible goals to have a separate standard, not compatible to MPEG. The obvious: Control and censorship is much easier this way. There is another paper (https://www.tandfonline.com/doi/abs/10.1080/09537325.2011.592270?src=recsys&journalCode=ctas20) but it seems to require a membership to be read.


I do not discuss political issues.

[Selur]

btw. there's now also uavs3e and https://github.com/uavs3/uavs3e

I also found https://github.com/xatabhk/FFmpeg-avs2-avs3 which is an ffmpeg version which supports avs2 and avs3 encoding.
Their release site also includes links to modified VLC and MPC-HC versions which support avs3 and MBAS support building ffmpeg with libuavs3d,..

Cu Selur