[Cin] SVT-AV1 library upgraded to v2.2.1

Wed Aug 28 01:32:51 CEST 2024

On Tue, Aug 27, 2024 at 3:10 PM Terje J. Hanssen
<terjejhanssen at gmail.com> wrote:
>
>
> Den 27.08.2024 05:50, skrev Andrew Randrianasulu:
>
>
>
> вт, 27 авг. 2024 г., 00:28 Terje J. Hanssen <terjejhanssen at gmail.com>:
>>
>>
>>
>> Den 26.08.2024 17:49, skrev Phyllis Smith:
>>
>> Terje,
>> Not sure why it reports it as SVT-AV1 Encoder Lib v2.2.0 because I downloaded what is reported as v2.2.1 here:
>>       https://gitlab.com/AOMediaCodec/SVT-AV1/-/releases
>> and the downloaded tar file was shown as: SVT-AV1-v2.2.1.tar
>> Maybe whoever added the mods to create v2.2.1 forgot to update the version line.  But at any rate, to prevent future confusion for me, I need to leave the tar file named 2.2.1.
>>
>>
>> Seemingly this is a minor bug somewhere in this version.
>> I upgraded my system SVT-AV1 and lib to v. 2.2.1 and got the same result: 2.2.0
>>
>> What confuse me more, is that Cingg rendering seems out of proportion 2.5x faster than corresponding transcoding the same hdv file using ffmpeg directly. Possibly Cingg calculates the speed after loading the input file, and ffmpeg not(?)
>>
>> hdv09_04.m2t --> hdv09_04_m2t_svt-av1-220_pr6+opus.webm
>>
>> CinGG:  **rendered 5972 frames in 34.052 secs, 175.379 fps
>> FFmpeg: frame= 5963 fps= 69 q=35.0 Lsize=  121908KiB time=00:03:58.77 bitrate=4182.4kbits/s speed=2.77x
>>
>> Here is the FFmpeg command line I use, which also out some mpeg2 errors:
>>
>> ffmpeg -hide_banner -i hdv09_04.m2t -c:v libsvtav1 -preset 6 -crf 35 -c:a libopus hdv09_04_m2t_ffmpeg-7_svt-av1-221_pr6+opus.webm
>
>
> you can try to add same parameters to av1 profile?
>
> -preset 6 -crf 35
>
> just without "-" and as two lines ...? (commented-out params use preset 6/crf 25 - you probably can uncomment and edit second param too)
>
>
> Thank you. Yes, with parameters "preset 6" and "crf 35" the Cingg/FFmpeg rendering speed matched:
>
> Cingg: ** rendered 5972 frames in 84.437 secs, 70.727 fps
>
> At first I wondered if perhaps different CPU core utilization had been used above. I verified now with top that Cingg used CPU 1400-1500% on a 12-core (8-mt/4-st) model: 12th Gen Intel Core i7-12700KF.
>
> I also had a look back into our January conversation about SVT-AV1 parameters. As already seen above, Cingg output what I think are SVT-AV1 defaults
>
> Svt[info]: SVT [config]: BRC mode / rate factor                 : CRF / 35
>
> Therefore I suggest that the Cingg preset make use of the same SVT-AV1 defaults, to avoid confusing.

well, I tried to uncomment/edit those lines and file a bit smaller
580kb vs 600 kb, and speed not really down (4.4 fps)
so I agree with you.

>
> Another suggestion from the Simple SVT-AV1 Beginner guide, is to always use 10-bit encoding:
> https://gist.github.com/BlueSwordM/86dfcb6ab38a93a524472a0cbe4c4100
>
> I always recommend to use 10-bit (-pix_fmt yuv420p10le) no matter what. It provides a good quality boost in all cases, particularly in darker shades and noisy stuff.
>
> I tested this yuv420p10le pixel format with Cingg too (no yuv42210le pix_fmt supported by SVT-AV1 yet):
> Result: 10-bit a bit slower rendering and smaller file size, both my samples looks good in my eyes.
>
> Svt[info]: SVT [config]: bit-depth / color format : 10 / YUV420
> ** rendered 5972 frames in 97.816 secs, 61.053 fps
>
> 119M    hdv09_04_m2t_svt-av1-221_cf35_pr6+opus.webm
> 117M    hdv09_04_m2t_svt-av1-221_cf35_pr6_yuv420p10le+opus.webm
>
>
>
>
>
>> [mpeg2video @ 0x562aa8f5b4c0] Invalid frame dimensions 0x0.
>>     Last message repeated 3 times
>> [mpegts @ 0x562aa8f55e00] PES packet size mismatch
>> [mpegts @ 0x562aa8f55e00] Packet corrupt (stream = 1, dts = 258142320).
>> [mpegts @ 0x562aa8f55e00] Could not find codec parameters for stream 2 (Unknown: none ([160][0][0][0] / 0x00A0)): unknown codec
>> Consider increasing the value for the 'analyzeduration' (0) and 'probesize' (5000000) options
>> [mpegts @ 0x562aa8f55e00] Could not find codec parameters for stream 3 (Unknown: none ([161][0][0][0] / 0x00A1)): unknown codec
>> Consider increasing the value for the 'analyzeduration' (0) and 'probesize' (5000000) options
>> Input #0, mpegts, from 'hdv09_04.m2t':
>>   Duration: 00:03:59.06, start: 2629.496000, bitrate: 26110 kb/s
>>   Program 100
>>   Stream #0:0[0x810]: Video: mpeg2video (Main) ([2][0][0][0] / 0x0002), yuv420p(tv, bt709, top first), 1440x1080 [SAR 4:3 DAR 16:9], 25000 kb/s, 25 fps, 25 tbr, 90k tbn
>>       Side data:
>>         cpb: bitrate max/min/avg: 25000000/0/0 buffer size: 7340032 vbv_delay: N/A
>>   Stream #0:1[0x814]: Audio: mp2 (mp3float) ([3][0][0][0] / 0x0003), 48000 Hz, stereo, fltp, 384 kb/s
>>   Stream #0:2[0x815]: Unknown: none ([160][0][0][0] / 0x00A0)
>>   Stream #0:3[0x811]: Unknown: none ([161][0][0][0] / 0x00A1)
>> Stream mapping:
>>   Stream #0:0 -> #0:0 (mpeg2video (native) -> av1 (libsvtav1))
>>   Stream #0:1 -> #0:1 (mp2 (native) -> opus (libopus))
>> Press [q] to stop, [?] for help
>> [libopus @ 0x562aa8f8b940] No bit rate set. Defaulting to 96000 bps.
>> Svt[info]: -------------------------------------------
>> Svt[info]: SVT [version]:    SVT-AV1 Encoder Lib v2.2.0
>> .........snip
>> Output #0, webm, to 'hdv09_04_m2t_ffmpeg-7_svt-av1-221_pr6+opus.webm':
>>   Metadata:
>>     encoder         : Lavf61.1.100
>>   Stream #0:0: Video: av1, yuv420p(tv, bt709, top coded first (swapped)), 1440x1080 [SAR 4:3 DAR 16:9], q=2-31, 25 fps, 1k tbn
>>       Metadata:
>>         encoder         : Lavc61.3.100 libsvtav1
>>   Stream #0:1: Audio: opus, 48000 Hz, stereo, s16, 96 kb/s
>>       Metadata:
>>         encoder         : Lavc61.3.100 libopus
>> [mpegts @ 0x562aa8f55e00] PES packet size mismatch0:03:50.00 bitrate=4148.7kbits/s speed=2.75x
>> [mpegts @ 0x562aa8f55e00] Packet corrupt (stream = 1, dts = 258142320).
>> [mpeg2video @ 0x562aa8f86800] ac-tex damaged at 10 61
>> [mpeg2video @ 0x562aa8f86800] Warning MVs not available
>> [mpeg2video @ 0x562aa8f86800] concealing 630 DC, 630 AC, 630 MV errors in P frame
>> [vist#0:0/mpeg2video @ 0x562aa8f5de80] [dec:mpeg2video @ 0x562aa90fa840] corrupt decoded frame
>> [out#0/webm @ 0x562aa8f8d380] video:119306KiB audio:2490KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: 0.091528%
>>
>>
>>
>>> > RenderFarmClient::main_loop: client started
>>> > FFMPEG::open_decoder: some stream times estimated: /home/guest/0005.avi
>>> > Svt[info]: -------------------------------------------
>>> > Svt[info]: SVT [version]:       SVT-AV1 Encoder Lib v2.2.0
>>>
>>> I also got and noticed the output version
>>>
>>> Svt[info]: SVT [version]: SVT-AV1 Encoder Lib v2.2.0
>>>
>>>
>>> Shouldn't it be SVT-AV1 Lib 2.2.1 as mentioned above
>>>
>>> cinelerra-5.1/thirdparty/src/libsvtav1-v2.2.1.tar.xz
>>>
>>> or possibly the 2.2.1 sub-version isn't shown in the Svt info?
>>>
>>
>> Andrey, if you see this, Andrew has suggested "maybe compilefarm configure command line for new OSes need to be
>> augmented with '--enable-libsvtav1" ?".
>> libsvtav1 requires cmake to be at 3.16 (or so far has been working at 3.12) so we did not want to have failures by making it compile by default in order to not impact older operating systems where it would always fail.
>>
>>
>