[Ffmpeg-devel-irc] ffmpeg-devel.log.20160327

[burek]

[00:10:21 CET] <cone-609> ffmpeg 03Marton Balint 07master:8ff0f6ae8253: avformat/segment: set format options for all segments
[05:05:22 CEST] <cone-186> ffmpeg 03Michael Niedermayer 07master:c50be7a52bc1: avcodec/h264_slice: Check PPS more extensively when its not copied
[05:21:54 CEST] <cone-186> ffmpeg 03Kieran Kunhya 07master:e259dc86a803: avcodec: Remove libdcadec, we already have it merged internally
[05:33:47 CEST] <llogan> kierank: is libfaac next on your list?
[05:33:54 CEST] <kierank> probably
[05:34:14 CEST] <llogan> anything else?
[05:39:20 CEST] <jamrial_> wasn't there a reason we didn't remove libfaac last time we cleaned house with external aac encoders?
[05:40:18 CEST] <llogan> none that i can recall
[05:56:17 CEST] <llogan> kierank: add x264opts vs x264-params to the list
[05:56:33 CEST] <kierank> make a wiki page =p
[05:56:46 CEST] <llogan> I don't know how
[05:58:37 CEST] Action: llogan hopes the lie will allow current level of laziness/worthlessness to continue
[06:40:35 CEST] <llogan> what's the difference between prores and prores_lgpl decoders?
[06:49:53 CEST] <dinux5> How should I use tools/patcheck? Please help. :)
[06:51:05 CEST] <llogan> dinux5: https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2016-March/192192.html
[06:52:12 CEST] <llogan> or you could probably do "tools/patcheck 0001-Mlp-encoder-addition.patch"
[06:55:16 CEST] <dinux5> llogan: Alright. Thanks! :)
[06:56:17 CEST] <llogan> dinux5: also, take a look in patcheck and read the paragraph that describes it
[06:59:14 CEST] <dinux5> Sure.
[07:56:49 CEST] <llogan> dinux5: the "Frequently ignored answer" paragraph in Michael's email was just a "signature" and was not a direct message for you
[08:10:39 CEST] <dinux5> Oops. Ok :D
[11:35:38 CEST] <cbsrobot> Very off topic but maybe some anime folks like it: https://opentoonz.github.io/
[11:38:38 CEST] <JEEB> ok, so they finally did the release
[11:39:26 CEST] <JEEB> oh, dwango owns it
[11:39:29 CEST] <JEEB> funky
[11:40:12 CEST] <cbsrobot> and it's bsd licensed
[11:46:02 CEST] <ubitux> interesting stuff @ https://github.com/opentoonz/opentoonz/tree/master/toonz/sources/image i suppose
[17:47:43 CEST] <atomnuker> jamrial: will backport all the AAC patches to 3.0 tommorow (they were all relatively unobtrusive)
[17:54:49 CEST] <jamrial> atomnuker: tell michaelni so he waits before tagging the 3.0.1 release :p
[21:17:43 CEST] <durandal_170> anyone gonna review dca_core bsf?
[21:19:28 CEST] <wm4> durandal_170: it doesn't even say what it does anywhere, why would anyone review it?
[21:21:37 CEST] <durandal_170> it extracts core of dts frames
[21:28:52 CEST] <andrey_utkin> Hi! Could anybody interested please look, confirm or comment - is that a bug or not, that H.264 concatenation using MPEG TS format is broken? http://trac.ffmpeg.org/ticket/5380#comment:4 Maybe I should file a different ticket?
[21:29:40 CEST] <andrey_utkin> and what is supposed to work, and what is not. If there's a bug, what to fix.
[21:30:31 CEST] <Compn> are you talking about mpegts or mp4>mpegts ?
[21:31:07 CEST] <Daemon404> some of that bug is a bit wtf
[21:31:08 CEST] <Daemon404> > -i orig.mp4 -c copy -bsf:v h264_mp4toannexb -y bsf.mp4
[21:31:11 CEST] <Daemon404> why would yu ever do this
[21:31:15 CEST] <Daemon404> this isnt even a legal mp4
[21:31:28 CEST] <Compn> lol
[21:32:27 CEST] <Daemon404> and concating mp4 files is not legal either
[21:33:36 CEST] <andrey_utkin> Daemon404, what i want is to join videos. Initially I was going to "precache" everything (and that's a lot for out application) in MPEG TS, but then I thought, "why wouldn't we just bitstream-filter that so that extradata is duplicated in each keyframe, because MP4 -> MPEGTS -> concat (MPEGTS) -> MP4 works"
[21:34:24 CEST] <andrey_utkin> apparently at MPEGTS -> MP4 conversion different extradata at IDRs doesn't get thrown away so that merged videos from different encoders doesn't break playback
[21:34:37 CEST] <Daemon404> concatting h264 in any format which does not allow multiple sps/pps is very error prone (the clips must have the exact same properties)
[21:34:41 CEST] <Daemon404> mp4 is one of those formats
[21:35:01 CEST] <andrey_utkin> properties - like framerate and resolution?
[21:35:10 CEST] <Daemon404> the entirety of the extradata
[21:35:11 CEST] <Compn> take h264 raw
[21:35:18 CEST] <Compn> then use mp4 tools to mux it
[21:35:22 CEST] <Compn> dont use ffmpeg.
[21:35:31 CEST] <andrey_utkin> extradata - yes, tricky, but it kinda works in practice
[21:35:41 CEST] <Daemon404> it cant work if it difefrs for mp4
[21:35:47 CEST] <Daemon404> because mp4 stores it *once* for the whole clip
[21:35:54 CEST] <Daemon404> itll be broken if it differs between concat'd clips
[21:36:18 CEST] <Daemon404> watchign a concatted mpegts is fine however.
[21:36:32 CEST] <Daemon404> remuxing said mpegts is nto guaranteed to work if it has multiple diffring sps/pps.
[21:36:41 CEST] <Daemon404> to mp4, i mean.
[21:37:47 CEST] <andrey_utkin> how could i learn what is in extradata, and what isn't?
[21:37:59 CEST] <Compn> read mp4 sepc\
[21:38:00 CEST] <Compn> spec*
[21:38:13 CEST] <Compn> if you havent read mp4 spec you shouldnt be playing.
[21:38:17 CEST] <Daemon404> if you are generating your own encodes, x264 has an option to allow concatenation
[21:38:27 CEST] <Daemon404> suppposing you encode using the exact same settings.
[21:41:55 CEST] <andrey_utkin> Daemon404, could you please give a hint what's that x264 option?
[21:42:11 CEST] <Daemon404> --stitchable
[21:43:24 CEST] <andrey_utkin> would anybody of you, gentlemen, consider consulting me time to time on such questions of h264 concat? not for free.
[21:46:44 CEST] <Compn> andrey_utkin : stick around, some dev may take you up on your consulting offer.
[21:47:08 CEST] <Daemon404> keep in mind is is easter sunday
[21:47:11 CEST] Action: Daemon404 is in fatc leaving now too
[21:47:23 CEST] <andrey_utkin> sure :)
[21:48:11 CEST] <andrey_utkin> thank you Daemon404 and Compn, you helped a lot
[21:48:26 CEST] <andrey_utkin> have a nice Easter
[21:55:54 CEST] <andrey_utkin> funny, with  -x264opts stitchable=1 results of my weird test stay the same. "illegal" mode is the only which works :) all other have artifacts
[21:56:21 CEST] <andrey_utkin>  -x264opts stitchable=1 was used at creation of input video clips of course
[22:03:12 CEST] <jamrial> durandal_170: the dca_core bsf seems to work wih every sample i tried, and the output is the same as eac3to's
[22:04:16 CEST] <durandal_170> I wonder where to find le and 14 variants
[22:06:03 CEST] <jamrial> those are usually dts in wav
[22:07:12 CEST] <Daemon404> andrey_utkin, do NOT concat mo4 files directly
[22:07:22 CEST] <Daemon404> concat the elementary annexb streams, and mux that t mp4
[22:07:23 CEST] <Daemon404> to*
[22:08:31 CEST] <andrey_utkin> Daemon404, by concat the ES you mean to make .h264 container-less files?
[22:08:59 CEST] <andrey_utkin> like ffmpeg -i 1.mp4 -c copy -bsf:v h264_mp4toannexb 1.h264 ?
[22:09:14 CEST] <Daemon404> what?
[22:09:31 CEST] <andrey_utkin> could you please elaborate what exactly are you proposing to do?
[22:09:31 CEST] <Daemon404> cat 1.h264 2.h264 > out.j264
[22:09:34 CEST] <Daemon404> mux out.h264
[22:09:44 CEST] <Daemon404> why on earth would you have x264 output a mp4
[22:09:58 CEST] <andrey_utkin> ok, and in order to obtain 1.h264 i should do what i have written above?
[22:10:06 CEST] <Daemon404> ...
[22:10:15 CEST] <Daemon404> i thought you were encoding the damn files yourself
[22:10:22 CEST] <Daemon404> output to an elementary stream
[22:11:56 CEST] <andrey_utkin> fine, it's just that there are legacy 'deployments' with mp4 files. So I guess it is legal to do what i supposed for mp4 -> h264.
[22:13:27 CEST] <andrey_utkin> ok, so you think concat on containerless h264 files should work better. Thanks for the idea, will try now.
[22:19:58 CEST] <andrey_utkin> yep that works cleanly. I wonder why that doesn't happen correctly in case of using containers. MP4 (please don't kill me) should carry the same amount of media information, and concat demuxer passes all H.264 streams through h264_mp4toannexb BSF internally.
[23:42:39 CEST] <cone-965> ffmpeg 03Paul B Mahol 07master:0b9957c3019a: avfilter/vf_waveform: move mirror variable into function argument
[23:42:39 CEST] <cone-965> ffmpeg 03Paul B Mahol 07master:48be92e5b684: avfilter/vf_waveform: optimize lowpass 8bit filter
[00:00:00 CEST] --- Mon Mar 28 2016