[Ffmpeg-devel-irc] ffmpeg.log.20140919

[burek]

[00:34] <active8> http://pastebin.com/jAnmAzAa  - x11grab - Non-monotonous DTS in output stream 0:0; 4 examples with the multitude of warnings snipped to a few. 1. no audio playing. 2. VLC playing audio. 3. Audio coming from a browser video. 4. Full screencast with the browser video playing.
[00:38] <llogan> what if you use -c:a copy instead? i wonder if using libpulse instead of alsa would make any difference.
[00:40] <llogan> you would need to build with --enable-libpulse though
[00:42] <relaxed> it appears he is using pulse
[00:42] <llogan> i meant use -f pulse, not -f alsa
[00:43] <okokokno> Hmm... Even though I now did avformat_find_stream_info, I still don't see some info I need for mp3 files
[00:43] <okokokno> For example, how many packets there are in the file
[00:43] <active8> it's not in the command output so I uess I would. ffmpeg -formats would output libpulse?
[00:44] <llogan> no, because your build configuration lacks --enable-libpulse
[00:44] <llogan> http://ffmpeg.org/ffmpeg-devices.html#pulse
[00:44] <okokokno> Or can that be computed from the duration somehow?
[00:44] <active8> but also notice that in the full screencap output, the message is all about the video stream: [matroska @ 0x2224460] Non-monotonous DTS in output stream 0:1;
[00:45] <okokokno> (AVStream::nb_frames is still 0)
[00:45] <llogan> active8: all this time i was thinking of "xruns".
[00:46] <llogan> trying to do too much as once and i suck at multitasking and reading
[00:46] <active8> that's what I meant by "output" --enable-libpulse isn't there
[00:46] <relaxed> active8: you're using a 2013 build, which is pretty old
[00:48] <active8> relaxed: roughly how often do they update ffmpeg. I'm not exactly ready to rebuild. Just asking.
[00:48] <relaxed> every three months - or in git everyday
[00:49] <active8> ffmpeg version git-2013-08-04-61af627 built on Feb 19 2014 04:33:30 with gcc 4.4.7 (GCC) 20120313 (Red Hat 4.4.7-4)
[00:49] <active8> funny. i pulled down those sources in feb and they were 6 months old
[00:50] <relaxed> active8: try again with output.flac
[00:51] <active8> same command, diff extension?
[00:51] <relaxed> yes
[00:51] <relaxed> loose -threads 0
[00:51] <relaxed> lose*
[00:53] <active8> unsupported codec
[00:54] <relaxed> pastebin
[00:55] <active8> http://pastebin.com/JKx2rApg
[00:56] <relaxed> active8: ffmpeg -f alsa -ac 2 -i pulse output.flac
[00:58] <active8> now it's a red message : http://pastebin.com/6a0PK8Ji
[00:59] <relaxed> what was the exit status?
[01:00] <Hfuy> hihi
[01:01] <Hfuy> ack
[01:01] <Jan-> Problem: we have some AVC-Intra MXF files which seem to have eight streams of audio.
[01:01] <Jan-> We're converting them (successfully) to ProRes quicktimes but the output files only have one mono audio channel in them.
[01:02] <Jan-> Here's the ffmpeg analysis of the files, input first: http://pastebin.com/jwiSR6SR
[01:02] <active8> relaxed: 2 if $! is exit status - can't remember
[01:03] <active8> i mean $?
[01:03] <relaxed> exit status is $?
[01:03] <active8> linux
[01:03] <Jan-> there should presumably be some way we can create stereo output files (using map?) but I'm not sure what the syntax would be. Any suggestions
[01:03] <Jan-> ...would be most appreciated :)
[01:03] <active8> the exit is at the bottom of the paste
[01:03] <c_14> Jan-: https://trac.ffmpeg.org/wiki/AudioChannelManipulation
[01:03] <active8> exit is 2 if $? is exit status
[01:03] <kepstin-laptop> Jan-: http://www.ffmpeg.org/ffmpeg-filters.html#amerge also
[01:04] <relaxed> ok
[01:04] <pzich> Jan-: have you been through https://trac.ffmpeg.org/wiki/AudioChannelManipulation ?
[01:04] <relaxed> Jan-: you want to include all the audio streams and change them to stereo?
[01:04] <active8> IIRC. I had this, or at least a messed up seek on VLC when making a slideshow. The time in VLC would be 00:00 -  but later conversion from mpeg to mp4 or just mpeg to mpeg through c:v copy fixed it. I'd still like to not see the meaasage and don't really know what it means. Seems ffmpeg would need to guess the timestamps from the sample rate
[01:04] <Jan-> As far as we know the input should only be stereo.
[01:05] <Jan-> So there are probably only two useful channels of interest
[01:05] <relaxed> Jan-: you're build is too old and that could be the problem
[01:05] <relaxed> your*
[01:05] <Jan-> we'd like to figure out which they are (probably by spitting out wave files to analyse which input streams we want) then put those into the stereo output.
[01:06] <pzich> Jan-: sounds like you need -map_channel
[01:06] <Jan-> uhhuh
[01:07] <Jan-> The problem we found is that a lot of the examples deal with muxing wave files into videos to create multipoint surround etc
[01:07] <llogan> or perhaps play in VLC and choose each individual stream. i imagine they would be in the same order as shown by ffmpeg
[01:08] <Jan-> Oh hm you can do that
[01:08] <Jan-> wait one we'll figure it out
[01:08] <Jan-> I would be amazed if we didn't just want 0 and 1
[01:08] <Jan-> or rather 1 and 2, as 0 seems to be the video
[01:08] <Jan-> MXF is such a PITA :/
[01:09] <Jan-> yeah, it's just 1 and 2
[01:10] <Jan-> for some reason VLC is interpreting the eight soundtracks as independent mono tracks, as opposed to assuming 1 and 2 are the stereo pair.
[01:10] <Jan-> ffmpeg does the same
[01:10] <llogan> -filter_complex "[0:a:0][0:a:1]amerge" -ac 2
[01:10] <llogan> or something like that
[01:10] <relaxed> it could be a decoding issue. are you sure there are stereo streams?
[01:11] <Jan-> wait one
[01:12] <pzich> it may need to be [0:a][1:a]
[01:12] <Jan-> llogan: Perfect
[01:12] <pzich> oh nm, misread the output
[01:12] <Jan-> man MXF Is horrible
[01:12] <Jan-> also, how would we ever have figured that out :/
[01:13] <llogan> you may want to compare using -ac 2 with the pan example in https://trac.ffmpeg.org/wiki/AudioChannelManipulation#a2stereostereo
[01:13] <Jan-> Honestly it's just a simple channel assignment issue
[01:13] <Jan-> your filter_complex line does it for a wave file
[01:13] <Jan-> we just have to integrate it into a script
[01:14] <llogan> i assumed pan and -ac 2 would provide the same output, but they don't for me
[01:14] <Jan-> C:\Users\CobaltUser>c:\ffmpeg -i "E:\Roll 1\XDROOT\Clip\120_0017.MXF" -filter_co
[01:14] <Jan-> mplex "[0:a:0][0:a:1]amerge" -ac 2 d:\test.wav
[01:14] <Jan-> ...has the desired effect
[01:14] <Jan-> where does 0:a:0 come from
[01:15] <pzich> 0:a:0 would be channel 0 of the first audio stream
[01:15] <Jan-> oh, right
[01:15] <kepstin-laptop> in your ffmpeg output, there's a bunch of Stream # lines
[01:15] <Jan-> there is
[01:16] <kepstin-laptop> I think using [0:1] and [0:2], matching the stream lines in the output, would work as well?
[01:16] <Jan-> fyi this MXF is the output of a Sony PXW-X180 camcorder in AVC-Intra 1080p25 mode.
[01:16] <llogan> yes, but it's lazier to use 0:a:0, etc.
[01:17] <Jan-> What does it mean by [mxf @ 00000000024d77e0] could not resolve sub descriptor strong ref
[01:17] <Jan-> is that some weird mxf container stuff
[01:17] <llogan> i need a new video camera...
[01:17] <Jan-> don't get a pxw-x180 :)
[01:18] <llogan> maybe i'll just start renting, but that's a bitch where i live
[01:18] <Jan-> this is a review loaner.
[01:18] <Jan-> and we scratched the display on one of the radio mic receivers :/
[01:19] <llogan> it was like that when you got it
[01:19] Action: Jan- nods vigorously 
[01:20] <Jan-> sony's much-advertised "AVC Intra" just seems to be 10-bit h.264 with all I frames
[01:20] <Jan-> at craptons of bitrate
[01:21] <Jan-> what does it mean by "input channel layouts overlap" here: http://pastebin.com/Vz7xp2S9
[01:22] <kepstin-laptop> Jan-: both input channels are 'mono', rather than one being 'left' and one 'right'
[01:22] <kepstin-laptop> Jan-: in this case it does the right thing, it assumes since there's two, that you meant stereo
[01:22] <Jan-> fair enough
[01:23] <Jan-> I wasn't aware a an audio channel could be marked with a position
[01:23] <kepstin-laptop> well, an audio stream is marked with a layout
[01:23] <Jan-> does this imply that the MXF isn't properly marking its channels
[01:23] <Jan-> (and wtf is it doing with 8 channels of audio anyway?!)
[01:24] <llogan> http://ffmpeg.org/ffmpeg-filters.html#amerge describes the input layout stuff
[01:24] <Jan-> ffmpeg is a very very big piece of software isn't it...
[01:27] <Jan-> Ah. I can't have -c:a copy with this audio filtering
[01:28] <kepstin-laptop> no, but -c:a pcm_s24le will pass it through with no further encoding.
[01:28] <Jan-> will just omitting the -c:a do the same
[01:28] <kepstin-laptop> depends on the output format, some may autoselect a lossy codec.
[01:29] <relaxed> probably not, be spcific
[01:29] <Jan-> hm no I get AAC in that case :)
[01:29] <Jan-> snksnksnk a prores quicktime with 128kbps AAC audio
[01:29] <Jan-> that's hilarious
[01:29] <kepstin-laptop> using ffmpeg's not particularly good aac encoder too, I bet ;)
[01:29] <Jan-> woah, this camera uses 24bit audio!?
[01:30] <llogan> Jan-: did you ever get to mess around with a Canon EOS C100 to C500?
[01:31] <Jan-> a C300, and not me, my boyfriend
[01:31] <Jan-> why d'you ask
[01:32] <llogan> what did he think of it? i may rent one for a week for a project (and learn how to use it on the plane).
[01:32] <Jan-> he says "incredibly nice sensor, shame about the gimped 8 bit output"
[01:32] <Jan-> and "what're you gonna do for glass"
[01:34] <llogan> more rental, but i haven't decided which lens(es) yet. i need to find more about the actual project if it's mostly interviews or more outdoor.
[01:35] <Jan-> canon know how to make nice sensors
[01:36] <Jan-> this fault had us worried though
[01:36] <Jan-> suddenly both channels appear to be the same audio and neither of them is the radio mic
[01:36] <Jan-> eeeeek
[01:37] <Jan-> By the way, who's this famous Kostya guy who did the prores encoder?
[01:37] <llogan> there are 2 prores encoders.
[01:38] <Jan-> Yeah.
[01:38] <Jan-> prores_ks seems to work quite a lot better, but it's a bit slower
[01:38] <Jan-> the normal "prores" one creates files that go way over bitrate (though few things seem to care)
[01:39] <kepstin-laptop> http://codecs.multimedia.cx/?p=388 has some discussion on speed
[01:39] <Jan-> we even put some of the non-ks prores files on hardware recorders
[01:39] <Jan-> and they played them fine
[01:43] <Jan-> "probesize" tells it how much of the video to look forward at when estimating bitrate, right?
[01:44] <Soul_keeper> I use it for streams that arent detected at the begging
[01:45] <kepstin-laptop> and in some formats (mpeg ts in particular) to find streams that don't start at the start of the file
[01:45] <Jan-> oh hm
[01:45] Action: Jan- runs "convert.bat" and goes to bed
[01:45] <Jan-> is "could not resolve sub descriptor strong ref" bad?
[04:47] <fluke_> Hi every one. I want to ask how to calculate network speed while playing http video with ffplay?
[04:47] <fluke_> My thought is to calculate in http.c
[11:47] <fswings> Morning, how do I extract the DTS core from a DTS-HD MA track using ffmpeg? Or are there alternative Linux based tools which are recommended?
[11:47] <fswings> I'm running via command line.
[13:29] <okokokno> Can I somehow get the info from the XING header in an mp3 via libavformat?
[13:38] <krasnayarsk> Hi. I have an avi video and its been converted from mp3. There is no image when listening/watching the avi file. What is the correct/best way to add a still image, perhaps a series of images, to the sound in the avi file?
[13:40] <c_14> https://trac.ffmpeg.org/wiki/Create%20a%20video%20slideshow%20from%20images
[13:42] <krasnayarsk> c_14: Thank you.
[13:43] <krasnayarsk> part
[15:27] <phw> Can ffmpeg play one file with video being redirected but audio playing locally?
[15:28] <ubitux> try to map a stream to an alsa output or something
[15:28] <ubitux> and map the video stream to another output
[15:29] <ubitux> your question is weird though
[15:30] <phw> well i am using piwall to display a video, while playing the same videos audio
[15:32] <phw> This results in "ffmpeg -re -i big_buck_bunny_720p_surround.avi -vcodec copy -f avi udp://239.0.1.23:1234" to broadcast the video via stream
[15:33] <phw> but i want ffmpeg to simultaniously play the audio of the file
[18:04] <Joryn> I'm looking to get some of the functionality in the AC3 decoder from FFMPEG into a gstreamer application that I'm working on (specifically recent dialnorm and DRC changes), but not sure how compatible the current version is with gstreamer-1.0.  What do I need to know?  Thanks!
[18:19] <lipizzan> is there a way to "tee" ffmpeg output to both write to ts stream to file disk and STDOUT (or socket?) as input to mplayer?
[18:24] <kepstin-laptop> sure, have ffmpeg write to stdout, and pipe it to the 'tee' command to save a copy to disk
[18:24] <kepstin-laptop> then pipe that to mplayer.
[19:35] <c_14> lipizzan: https://trac.ffmpeg.org/wiki/Creating%20multiple%20outputs#Teepseudo-muxer
[20:30] <jelle__> I have a origin.mkv with 1 fps and I would like to keep all the frames, but play them 60 fps. How do I do this? Keep all frames, but speed them up?
[20:36] <__jack__> strange idea
[20:40] <__jack__> you can still extract all frames as image, then copy all of them 60x, and recreate the movie
[21:12] <jelle__>  have a origin.mkv with 1 fps and I would like to keep all the frames, but play them 60 fps. How do I do this? Keep all frames, but speed them up?
[21:13] <c_14> ffmpeg -i origin.mkv -r 60 out.mkv ?
[21:14] <jelle__> no filters needed?
[21:15] <c_14> shouldn't
[21:19] <jelle__> ok, Ill fiddle with that, thanks
[21:42] <Diogo> Hi this is possíble to generate  hls m3u8 with fixes extinf?
[21:49] <jjohn> Ahoy, guys. Just a quick question.
[21:50] <jjohn> I got raw pcm_s16le streams and want to put them into a lossless container format. I read that FLAC might be appropiate, but the bitrate of the converted files is lower than that of the original ones.
[21:50] <jjohn> I thought FLAC was lossless, wasn't it? Why is the bitrate lower then?
[21:50] <c_14> pcm is uncompressed
[21:50] <c_14> FLAC is compressed
[21:51] <jjohn> But still lossless? And it's the ONLY difference?
[21:51] <c_14> It's not the _only_ difference, but it is still lossless.
[21:51] <c_14> Ever take a big file and compress it in rar/zip/7z/gz/bz2/xz ?
[21:52] <c_14> Same concept.
[21:52] <jjohn> No differences in quality either?
[21:52] <jjohn> OK, I got it. Thank you very much then.
[21:52] <c_14> No difference in quality (as long as the encoder doesn't start doing weird shit)
[21:52] <jjohn> Thanks. You have been always a great help to me.
[21:52] <jjohn> Bye then.
[22:52] <ackwood> Hi everyone. What is "AVCODEC_MAX_AUDIO_FRAME_SIZE" known as these days?
[22:52] <ackwood> Last time I remember being able to use it I had ffmpeg 0.8
[23:04] <Plorkyeran> it was removed entirely, not renamed
[23:05] <ackwood> Plorkyeran, that sucks. I have a framework that depends on the existence of that variable. Is there a program way of coming up with that number?
[23:05] <ackwood> I'm not exactly an expert on multimedia encdec, I'm just trying to maintain something useful
[23:06] <Plorkyeran> it was removed because anything that used it was inherently broken and needs to be rewritten to do things correctly
[23:07] <Plorkyeran> iirc doing so was fairly easy
[23:07] <Plorkyeran> other than the utter lack of useful documentation...
[23:09] <goulard> anyone know the rule of thumb on how many AAC ADTS packets to put in one TS packet?
[23:14] <Sokolio> You mean PES packet?
[23:14] <goulard> yeah
[23:15] <goulard> I put one ADTS packet per PES, and quicktime and vlc wont play
[23:15] <Sokolio> I guess there's no rule, even a rule of thumb
[23:18] <Sokolio> PTS however must be set right according to sample rate
[23:18] <Sokolio> does vlc complain in any way?
[23:18] <goulard> it just says there is nothing to play
[23:18] <goulard> i suppose the pts is correct, as the duration in quicktime is correct
[23:19] <goulard> and media info reports properly
[23:19] <goulard> pts = frame# * 1024 / samplerate
[23:21] <Sokolio> does quicktime actually support adts packaging?
[23:21] <goulard> yes
[23:21] <goulard> you have to contain AAC in ADTS in TS
[23:21] <Sokolio> in qt there is a 2 byte config
[23:21] <Sokolio> and raw AAC
[23:21] <Sokolio> yes, in TS that;s true
[23:22] <Sokolio> well, I have no idea
[23:23] <Sokolio> I've seen a single ADTS packet per PES and it sorta played
[23:24] <goulard> do you know what buffer fullness is for adts?
[23:24] <goulard> I had wondered if that had something to do with it
[23:24] <Sokolio> you mean the header field
[23:25] <Sokolio> well, I usually hardcoded it
[23:25] <Sokolio> dunno if it wasnt all 1's in my case
[23:27] <Sokolio> 7ff to be specific
[00:00] --- Sat Sep 20 2014