[AV1 Encode Test] Sewayaki Kitsune no Senko-san - 01v2+02 [Error in the Matrix]

Category:
Date:
2019-04-23 04:41 UTC
Submitter:
Seeders:
0
Information:
No information.
Leechers:
0
File size:
136.9 MiB
Completed:
58
Info hash:
869cfb5acac11811695ee0e54dbebc4dd3641716
This is the second release of this test, I included the batch files that I used to encode it. In this case the software is: ffmpeg version N-93593-g88325f4b34 with libaom-av1 version 1.0.0-1626-g0e48db955 mkvmerge v33.1.0 ('Primrose') 64-bit I decided to remux due to firefox compatibility and that it also saves a few hundreds of kilobytes, firefox makes the release be a title late as I found another bug as if "setsar=1" isn't do during scaling it renders the video incorrectly (really small). The speed was around 4% of real time for 480p (0.96+ fps) and 0.62% of real time for 1080p (0.14+ fps). There is some banding in the episode 2 and this banding is also found in the original but less noticeable. For now my impression of Av1 that even although it have good quality it also have a poor rate control, plus "quality" and "constrained quality" modes are not consistent as at the same q factor as both give very different bitrates even in the situation of "q50" vs "q50 maxbitrate 25mbits" for sd. In 1080p the quality mode it has a great variance of bitrate, as example with q50 goes as high of 1000kbps in a sample of the movie "the golden compass", 560kbps in "kanini & kanino" or the lass that 200kbps in this case. Also, I have been surprised with the q value of 50 witch is rater low quality as it looks at least as good if not better that over the air TDT for the most part. There is also that it performs much better in HD than in SD, this is understandable as it's designed for HD and that at this very low bitrates the entropy is already making the things very difficult for the encoders. I mark it in red the due to its experimental status even if it's not needed. And I have been doing encodes for about 20 years and encoded from in order more or less chronological in: Mpeg1-VCD and variants - source VHS and DVD for sneaker net Mpeg2-SVCD and variants - source VHS and DVD for sneaker net DivX;-) 3.x - source VHS and DVD for sneaker net RealVideo G2 - source VHS and DVD, early web release attempt Mpeg2-DVD - source VHS for sneaker net OpenDivX - to sort lived to be used and forked into DivX and Xvid DivX - source VHS and DVD, also p2p wen doing hardsub, sneaker net and p2p release Xvid - source VHS and DVD, also p2p wen doing hardsub, sneaker net and p2p release VP3 - tried and liked it for anime but never used due to lack of support x264 - source VHS and DVD, p2p and dd release Theora - evaluated and never used for release but used in other context VP8 - evaluated and never used for release x265 - source VHS and DVD, p2p and dd release VP9 - evaluated, liked and using in other context. Av1 - the one that I'm evaluating now Apart of this codecs there are others but only briefly tested or only for internal use like HuffYUV, MJPEG or MS video1. I have been most active during de Xvid and Divx era and after that era I have been doing only rare to find content from time to time.

File list

  • [AV1 Encode Test] Sewayaki Kitsune no Senko-san - 01v2+02 [1080+480][Error in the Matrix]
    • [HD Test] Sewayaki Kitsune no Senko-san - 01 [Error in the Matrix][AV1 1080p 8bits q50][opus 64kbs][51BF4CA2].mkv (45.1 MiB)
    • [HD Test] Sewayaki Kitsune no Senko-san - 02 [Error in the Matrix][AV1 1080p 8bits q50][opus 64kbs][66134132].mkv (42.2 MiB)
    • [SD Test] Sewayaki Kitsune no Senko-san - 01 [Error in the Matrix][AV1 480p 8bits q50][opus 48kbs][3A1C1D10].mkv (25.4 MiB)
    • [SD Test] Sewayaki Kitsune no Senko-san - 02 [Error in the Matrix][AV1 480p 8bits q50][opus 48kbs][95EEEE32].mkv (24.2 MiB)
    • [encodec-HDa.bat (786 Bytes)
    • [encodec-SDa.bat (989 Bytes)
I know very little about encoding so question: would 10bit address the banding issue? That's why people say they use 10bit x264 even if the source is 8 bit
@anvil No, simply going to 10bit won't do a thing against the banding present in the source. You need to actually filter it for 10bit to matter.
10bit might make the file even smaller though.