[Ffmpeg-devel-irc] ffmpeg-devel.log.20180921

[burek]

[00:53:30 CEST] <cone-550> ffmpeg 03Michael Niedermayer 07master:c02ff7768113: doc/examples/vaapi_transcode: Fix the typo
[12:12:22 CEST] <cone-577> ffmpeg 03Jun Zhao 07master:04cbaae2aa50: lavu/frame: Add missing conversions from side data enum to name
[12:12:22 CEST] <cone-577> ffmpeg 03Jun Zhao 07master:50d9c85358a1: lavc/avpacket: Add missing conversions from side data enum to name.
[16:34:19 CEST] <cone-833> ffmpeg 03Pedro Arthur 07master:1cb101f60d3b: avfilter/vf_sr: Fix coverity CID 1439584
[17:36:09 CEST] <cone-833> ffmpeg 03Kagami Hiiragi 07master:cbe5c7ef386d: lavf/matroska: Allow AV1 in WebM
[18:44:28 CEST] <durandal_1707> atomnuker: how you tested? i had bug with samplerates bigger than 22050?
[18:45:47 CEST] <atomnuker> a samplerate at 44100, set noise pattern to white, maxed out the removal ratio, observed how mostly all except speech got through
[18:49:56 CEST] <durandal_1707> atomnuker: you hadn't changed noise floor?
[18:50:40 CEST] <atomnuker> no, I don't believe I did
[18:50:52 CEST] <atomnuker> I can test more after tomorrow
[18:53:07 CEST] <atomnuker> jamrial: I think that was a bit premature, the matroska spec won't be finalized until next week and they do this thing where they pile up all the changes and apply them at the end
[18:56:05 CEST] <nevcairiel> it matches the isobmff format mostly, the parts we care about  are not going to diverge
[18:57:10 CEST] <atomnuker> oh, ok, I thought it wasn't changed yet to do that as I hadn't seen commits to do that
[21:58:11 CEST] <cone-282> ffmpeg 03James Almer 07master:4755b6e6d194: avformat/matroskaenc: write CodecPrivate in WebM
[22:17:15 CEST] <durandal_1707> atomnuker: compare it with https://github.com/werman/noise-suppression-for-voice
[23:18:16 CEST] <nevcairiel> jamrial: that latest change looks odd. codec private was already being written on account of codec_id being  != webvtt
[23:18:49 CEST] <nevcairiel> i think the reason for that check was that webvtt was different between matroska and webm
[23:19:48 CEST] <jamrial> does webvtt even have extradata?
[23:20:21 CEST] <jamrial> i can revert it in any case. you're right it doesn't affect av1 codecprivate
[23:21:20 CEST] <nevcairiel> In Matroska it does, in WebM it does not
[23:21:38 CEST] <nevcairiel> WebM defined a weird mapping that the Matroska people didnt want to  use
[23:21:43 CEST] <jamrial> ok
[23:22:26 CEST] <nevcairiel> its also D_WEBVTT/SUBTITLES in WebM vs S_TEXT/WEBVTT in Matroska
[23:23:51 CEST] <jamrial> are we writting the latter? i can't find it...
[23:24:18 CEST] <nevcairiel> not sure
[23:24:28 CEST] <nevcairiel> the matroska definition is relatively  new
[23:24:46 CEST] <jamrial> there are four different D_WEBVTT tags for webm, depending on the track's "disposition"
[23:24:55 CEST] <nevcairiel> apparently we cant even demux that yet
[23:26:18 CEST] <cone-282> ffmpeg 03James Almer 07master:a068594248b7: Revert "avformat/matroskaenc: write CodecPrivate in WebM"
[00:00:00 CEST] --- Sat Sep 22 2018