[Ffmpeg-devel-irc] ffmpeg.log.20140317

burek burek021 at gmail.com
Tue Mar 18 02:05:01 CET 2014


[00:59] <thangalin> Hey folks. I'm getting Xext not found when building ffmpeg.
[00:59] <thangalin> http://pastebin.com/raw.php?i=axhNMhJ9
[01:00] <thangalin> x11proto-xext-dev is installed.
[01:00] <thangalin> Any ideas how to fix this to get a static build?
[10:10] <Aiena> HOw do I tell ffmpeg to exit cleany in a shell script ? ctrl+c seems to not be the best method as it can tend to corrupt videos if trigged at the wrong time.
[10:11] <Aiena> it corrupts videos on occasion
[10:15] <relaxed> Aiena: killall -INT ffmpeg
[10:15] <Aiena> relaxed but will ffmpeg encode properly and exit like that
[10:15] <Aiena> that will just kill ffmpeg on the spot
[10:16] <Aiena> *or that will just kill ffmpeg on the spot
[10:16] <Aiena> I had read online that this video corruption could be due to a memory leak but those reports are so old that they are no longer relevant
[10:19] <relaxed> It will make ffmpeg exit cleanly. Test it.
[10:22] <Aiena> relaxed and I need to specify it as -INT or an integer number ?
[10:24] <relaxed> I gave you that exact command.
[10:25] <relaxed> s/that/the/
[10:26] <Aiena> relaxed thanks
[10:26] <Aiena> I did not know what -INT was
[10:26] <Aiena> now I know
[10:26] <Aiena> i thought it was integer but it is a type of signal
[10:26] <Aiena> relaxed but now I am stuck again
[10:27] <Aiena> once ffmpeg is recording it will only stop recording after I tell it to but at the moment I can tell it to stop recording only with ctrl+c
[10:27] <Aiena> I do not know how I would tell the script to use your command instead
[10:28] <Aiena> i wonder if there is a way to overload ctrl + c to do something else
[10:30] <pron> Aiena -t
[10:30] <Aiena> pron an example
[10:30] <pron> t duration'
[10:30] <pron>     Restrict the transcoded/captured video sequence to the duration specified in seconds. hh:mm:ss[.xxx] syntax is also supported.
[10:30] <pron> i dont have one
[10:31] <pron> but basicly add -t 30 for 30 seconds long video
[10:31] <Aiena> ok
[10:32] <Aiena> pron is it possible to tell ffmpeg to encode for 10 more seconds after ctrl+c is pressed ?
[10:32] <pron> idk
[10:32] <pron> dont think so
[10:36] <Aiena> ok
[10:55] <relaxed> Aiena: you can run killall -INT ffmpeg from another terminal
[10:55] <Aiena> Ah ok
[12:22] <q_bit> I need to make an .exe for windows and it seems that this cross compile process is the easiest. Are there other cross compile processes I should look at?  https://github.com/rdp/ffmpeg-windows-build-helpers
[12:25] <JEEB> have a sane enough mingw-w64 compiler or build with MSVC 2013
[12:26] <JEEB> and how simple or not simple it is to compile depends wholly on what third party libraries you want to link to FFmpeg
[12:33] <q_bit> thanks JEEB that makes sense with what i'm seeing starting with this helper script.
[12:35] <JEEB> basically third party libraries are almost always encoders, so note what you are going to be doing with your binary and set things up accordingly
[12:35] <JEEB> one exception is libopus which has a decoder as well as an encoder, but if you've never heard of opus or never seen a file with it, you're OK
[12:40] <cody_> hi
[12:41] <cody_> i want to decode and display a h264 bitstream in realtime
[12:41] <cody_> how could i do that
[12:41] <cody_> no buffering, no audio
[12:42] <cody_> http://ffmpeg.org/doxygen/trunk/h264_8h.html
[12:42] <cody_> is that the right place to look at?
[13:34] <qxt> Got a so called *.ts file here that is 1080i. The interlacing looks bad. Any recommendation on how to de-interlace and not loose to much quality?
[13:34] <qxt> btw dropped the mess into a *.mkv container already.
[13:37] <qxt> Using Debian GNU/Linux stable.
[13:37] <qxt> Only main repos
[13:41] <iive> -vf yadif should do reasonable job at good speed.
[14:30] <jarainf> qxt, -vf "fieldmatch, decimate" does the job
[14:31] <jarainf> It will deinterlace the transport stream and decimate it to 23,976 fps
[14:31] <jarainf> If you still have interlaced frames after fieldmatch you can add yadif between fieldmatch and decimate
[14:34] <jarainf> horizontally scrolling text might still look bad, but as far as I know, there is no filter to fix this in ffmpeg yet (don't quote me on that)
[15:21] <pron> is there a way to tell ffmpeg not to exit when input isnt available
[15:22] <pron> like when input is http://
[15:28] <klaxa> what else should ffmpeg do?
[15:29] <pron> idk , cook me a dinner ~.~
[16:53] <namccarty> Alright, so yesterday I was experiencing an issue where ffmpeg was outputting slightly corrupt mp3s when down sampleing correct files and collecting the output over a pipe
[16:53] <namccarty> wound up having to leave before we got anywhere
[16:54] <namccarty> i am able to reproduce the problem on both linux and windows across multiple versions of ffmpeg
[16:56] <namccarty> and the files are only corrupt when the output is to the standard out
[16:56] <namccarty> where do I go from here?
[16:56] <namccarty> also ffmpeg is consistent on which files trigger the issue
[17:22] <canci> is there a way to start ffplay in fullscreen mode?
[17:24] <jarainf> namccarty, uhm... Do you need the pipe?
[17:25] <namccarty> jarainf for what i am using ffmpeg for, yes
[17:26] <jarainf> I've just tested what you did, and indeed the piped file seems to be a bit broken, but that's nothing which would stop it from playing back (atleast in mplayer/mpv that is)
[17:26] <namccarty> yeah, its only specific files that it breaks on for me, and its slight enough that vlc can play it
[17:26] <namccarty> but the media player on my phone cant
[17:26] <jarainf> namccarty, well, I don't really get what you need the pipe is for...
[17:26] <jarainf> -us
[17:27] <jarainf> -is
[17:27] <namccarty> im using a piece of software called madsonic to automically downsample and stream to my phone
[17:27] <namccarty> it requries that its transcoders output to standard out
[17:30] <namccarty> it works perfectly for maybe 99% of the songs in my library
[17:30] <namccarty> but there are a few albums that can reliably trigger this problem
[17:32] <jarainf> Soooo...
[17:33] <jarainf> You are not doing a reencode at all?
[17:34] <namccarty> normally i stream to my phone at a set 128kbps to save data, and pretty much all of my music is encoded above that, so usually i end up downsampleing
[17:37] <namccarty> manual testing has confirmed that it is only specific files that this issue shows up for when reencoding
[17:39] <jarainf> isn't madsonic supposed to do the downsampling?
[17:39] <namccarty> it passes the downsampling off to the command you have set to do so
[17:39] <namccarty> which is by default, ffmpeg
[17:39] <namccarty> and collects the output from standard out and streams it off
[17:42] <namccarty> This is what transcoding settings look like
[17:42] <namccarty> http://puu.sh/7yVM0.png
[17:43] <jarainf> you don't happen to have an example file which is cc?
[17:44] <namccarty> not that i am aware of
[17:45] <jarainf> Is that a default setting?
[17:45] <namccarty> mostly default, yes
[17:45] <namccarty> the defualt for downsampleing had the pipe:1 replaced with -
[17:47] <namccarty> this is the subsonic page on reccomended settings, subsonic being the software madsonic was forked off of
[17:47] <namccarty> http://www.subsonic.org/pages/transcoding.jsp
[19:25] <namccarty> so is this anything we can think of a sloution for or devise a hack to get around, or is it time for me to go file a bug report
[19:56] <sisco> Hello all, is there any way to do timeshifting with ffmpeg? like 6 hours
[19:57] <llogan> canci: -fs
[19:59] <llogan> namccarty: you should provide your command and the complete ffmpeg console output and any required samples so we can attempt to duplicate the issue
[20:00] <sisco> well no idea how to do timeshift thats why im asking here how to do so
[20:00] <llogan> sisco: what do you mean by "timeshift"?
[20:00] <sisco> like iptv, stream live and stream timeshift back 6 hours
[20:01] <namccarty> llogan alright, let me go fetch those
[20:01] <namccarty> made those yesterday, just need to go find where i put them
[20:03] <sisco> or to be more sample ffmpeg will record an rtmp stream but keeps always only the last 10 mins
[20:05] <llogan> sisco: you can try the segment muxer http://ffmpeg.org/ffmpeg-formats.html#segment_002c-stream_005fsegment_002c-ssegment
[20:06] <llogan> -segment_time and -segment_wrap being options you will want to investigate
[20:08] <sisco> yes i saw it before , its gonna create segment file with specific time or size
[20:08] <sisco> but im gonna be stuck to restream them to create one rtmp
[20:11] <namccarty> llogan here is the console output
[20:11] <namccarty> http://pastebin.com/Zq6Mz9U6
[20:12] <namccarty> the only samples i have that trigger the issue are copyrighted
[20:13] <namccarty> not sure how to procede, I honstly don't care about the copyright issues here since no profit is being made, but i don't know how you feel or what the procedures are for that
[20:14] <llogan> namccarty: if the input is needed to duplicate the issue then please provide it
[20:16] <namccarty> llogan: here is a sample that triggers the output accompanied with the outputs i get when piping and not piping
[20:16] <namccarty> https://dl.dropboxusercontent.com/u/42936303/samples.tar
[20:17] <llogan> namccarty: thanks. which players does the output not play correctly?
[20:17] <namccarty> the apollo and ultrasonic media players on my phone, it plays correctly on vlc on my desktop
[20:19] <namccarty> i've used a couple of tools that claim to check the validty of mp3 files, and they all claim that the pipe version has some sort of mpeg stream error
[20:27] <hi117> does x265 have a irc channel?
[20:28] <hi117> my package manager is warning that the libs have writable and executable sections
[20:28] <llogan> namccarty: WMP 12 doesnt like your piped.mp3 either
[20:30] <namccarty> llogan: any idea what is casuing me to get a bad output when its via standard out?
[20:33] <llogan> not yet. but i did encounter a seg fault that i neglected to investigate last week
[20:34] <namccarty> this issue has been plauging me for several months and I have only just now gotten around to realizing that the problem was with ffmpeg and not the program consuming its output
[20:34] <llogan> if you exclude the video stream (album art) it may work.
[20:34] <namccarty> how would i go about doing that?
[20:34] <llogan> -vn or -map 0:a
[20:35] <namccarty> let me give it a shot
[20:36] <BlackBishop> any ideas on how I could get the subtitle from a file then make it hardcoded in the output ?
[20:36] <pyBlob> I've managed to capture audio from my microphone on windows using ffmpeg and dshow, the only problem is that the audio is delayed by ~1s, but I want something about 0-50ms
[20:36] <klaxa> BlackBishop: http://trac.ffmpeg.org/wiki/How%20to%20burn%20subtitles%20into%20the%20video
[20:37] <BlackBishop> the subtitle is a stream in the initial input file ( mkv container )
[20:37] <BlackBishop> :/
[20:37] <klaxa> just specify the .mkv in -vf subtitle=somefile.mkv
[20:38] <klaxa> .mkv can be handled as a file containing subtitles
[20:38] <klaxa> because it is
[20:38] <klaxa> font attachments could become an issue
[20:38] <namccarty> llogan: looks like it did it windows media player is accepting the piped file now and it looks like i can now stream to my phone without anything breaking
[20:38] <namccarty> thank you
[20:41] <llogan> namccarty: i guess that narrows it down. you should report it as a bug. make sure to include the command, complete console output, and the input file. you can omit "-loglevel verbose". ...but first test with a more recent build if you can
[20:42] <llogan> also, why re-encode instead of stream copy?
[20:42] <klaxa> BlackBishop: here is a shellscript that extracts fonts from a matorska file, it is pretty bad, but it works: https://gist.github.com/klaxa/5651164
[20:42] <namccarty> llogan: the reencoding is to save bandwidth streaming to my phone, its done by the madsonic media server using ffmpeg
[20:42] <namccarty> ive had the same issue with the git version i pulled down yesterday testing
[20:43] <namccarty> the logs i just put together were using the version that ships with the windows version of madsonic because i forgot where i put the git verion
[20:44] <namccarty> i guess im off to go report a bug then
[20:44] <llogan> you should use the new version for your report. also test to see if stream copying instead of re-encoding can reproduce the issue because then you can possibly omit libavcodec as a possible culprit
[20:44] <BlackBishop> klaxa: thanks .. but if there's a special font, would there be any problems if I'm not using them ?
[20:44] <klaxa> it would render the subtitles incorrectly
[20:44] <klaxa> gotta go afk, back whenever
[20:45] <BlackBishop> [AVFilterGraph @ 0x20f6c70] No such filter: 'subtitles''
[20:46] <BlackBishop> :-(
[20:46] <BlackBishop> altough I have it compiled with libass
[20:53] <Kasper^> can you specify keyframes when you encode to vp9 ?
[21:18] <pzich> I'm using `ffmpeg -i <in> -vcodec h264 -preset slow -pix_fmt yuv420p -profile:v main -level 3.1 -movflags +faststart <out>` to encode some videos. Does faststart have any quality implications for the file, or does it just help getting the video to start playing faster?
[21:26] <BlackBishop> damn, ffmpeg is finishing faster than the client can view via the rtmpstream :|
[21:27] <sacarasc> There's a setting to make ffmpeg encode at real time.
[21:33] <BlackBishop> re ?
[21:33] <BlackBishop> -re *
[21:33] <BlackBishop> I think I just discovered it :)
[21:33] <sacarasc> Yeah.
[22:43] <RAZ0REDGE> hey i have install libmp3lame 3.9.5 or something
[22:43] <RAZ0REDGE> but ffmpeg git source is asking for something different
[22:43] <RAZ0REDGE> ERROR: libmp3lame >= 3.98.3 not found
[22:43] <RAZ0REDGE> why do i need a older lame for this ffmpeg
[22:43] <llogan> pzich: no quality implications
[22:44] <llogan> although note that -level does not change refs
[22:45] <jarainf> RAZ0REDGE, '>=' means greater than or equal to 3.98.3
[22:45] <llogan> RAZ0REDGE: it says greater than or equal to 3.98.3
[22:45] <jarainf> heh, llogan
[22:45] <llogan> stereo
[22:50] <RAZ0REDGE> LAME 64bits version 3.99.5 (http://lame.sf.net)
[22:50] <sacarasc> Did you also install the dev files?
[22:51] <RAZ0REDGE> no
[22:52] <RAZ0REDGE> hmm dev files ?
[22:52] <jangle> greetings all, attempting to use avcodec_video_decode2.  Its my understanding that the AVPacket struct needs to be loaded with complete frames each call
[22:52] <sacarasc> I am assuming that you just installed Lame through your package manager. Most of them split the binaries and stuff from what is needed to compile using them.
[22:53] <jangle> so I maintain a buffer that has an h264 bitstream consisting of the sps, pps, idr nals
[22:53] <RAZ0REDGE> yeah openbsd installs these lame and ffmpeg that i dont want to use
[22:53] <RAZ0REDGE> perhaps i will delete them
[22:53] <sacarasc> You need the dev files.
[22:53] <sacarasc> Of the current version only.
[22:54] <jangle> when i call avcodec_decode_video on that buffer, I get a successful decode, but when I append a non-idr, type 1, nal, and call decode, I get Missing reference picture, default is 0
[22:54] <RAZ0REDGE> where are the dev files ?
[22:55] <jangle> as I continue to append p frames, I eventually get a successful decode call, but it is my understanding that the addition of the type 1 nal should still constitute a complete frame
[22:55] <sacarasc> If you're going to uninstall things, use your package manager whenever possible, RAZ0REDGE.
[22:56] <RAZ0REDGE> ok, as i know that, but what do you mean by developer files
[22:56] <RAZ0REDGE> is this on the ffmpeg website ?
[22:56] <llogan> RAZ0REDGE: "dev" files being debian/RHEL talk for whatever package that installs lame.h
[22:57] <llogan> or you can compile lame yourself
[22:57] <RAZ0REDGE> i have install lame
[22:57] <RAZ0REDGE> by source yes
[22:57] <RAZ0REDGE> oh i'm not using linux
[22:57] <RAZ0REDGE> as i mention that i'm using openbsd
[22:57] <sacarasc> You still need them on BSD.
[22:58] <llogan> so the question then becomes why ffmpeg does not see your lame
[22:58] <RAZ0REDGE> not sure really
[22:58] <RAZ0REDGE> i type in lame
[22:58] <sacarasc> RAZ0REDGE: If you do `which lame` what does it output?
[22:58] <RAZ0REDGE> and it seems to use the 3.99.5
[22:59] <RAZ0REDGE> perhaps i have to link it
[23:00] <pzich> llogan: thanks, any idea how I can up the quality of the video in general? from what I've read online the h264 encoder is pretty automatic in quality settings, but I'm trying to go with "very best quality" and am still getting more artifacting than I'd like
[23:04] <RAZ0REDGE> aaa
[23:04] <RAZ0REDGE> this work
[23:04] <RAZ0REDGE> i have done lame as ./configure --prefix=/usr
[23:04] <RAZ0REDGE> i suppose its linking in /usr instead of /usr/local
[23:05] <jangle> so in summary, i'm presenting buffers that look like this [7 8 5] [7 8 5 1] [7 8 5 1 1] [7 8 5 1 1 1] and so on, is this correct?  The library seems to tell me, no.  so, what is correct?
[23:07] <RAZ0REDGE> now re-building ffmpeg with lame support
[23:07] <RAZ0REDGE> so far so good :)
[23:13] <llogan> pzich: right now you're using the default setting for -crf (since you did not include this option), which is -crf 23. you can use a lower value.
[23:14] <llogan> see https://trac.ffmpeg.org/wiki/x264EncodingGuide
[23:16] <pzich> llogan: thanks, I'll take a look
[23:19] <llogan> jangle: you might get more help for library usage at libav-user mailing list
[23:19] <jangle> 1logan: thanks
[23:21] <RAZ0REDGE> hmm odd
[23:22] <RAZ0REDGE> thanks for everything
[23:22] <RAZ0REDGE> cheers
[23:31] <pzich> llogan: this is exactly what I needed, thanks, much appreciated
[23:35] <pzich> since you've known all the answers thus far: do you know (or have an article that states) the compatibility of yuv444? the settings I got were optimized for iPad compatability a few gens back, so is using 3.1 and yuv420
[23:36] <pzich> looks like the last answer in the FAQ is saying to go yuv420 for quicktime
[23:37] <llogan> pzich: yes, use yuv420 for non-FFmpeg based players
[23:37] <pzich> thanks
[23:38] <llogan> which is the oldest iPad you're supporting?
[23:38] <pzich> it's actually possible that this won't need to support iPad or any iOS device, but I'm precautinarily supporting back to iPad 2/iOS 6
[23:39] <pzich> I recall there being a table listing iOS devices and the h264 level settings, so I'll take a look at that again
[23:40] <pzich> oh yes, further down on that very page
[23:41] <llogan> i should update that by mentioning refs
[23:56] <pzich> is there a way I can specify frame numbers using -ss instead of HH:MM:SS.xxx? or at a higher level, is there a good way to export a single still of a particular frame from a video?
[23:58] <pzich> looking like the 'select' video filter might be able to do it
[00:00] --- Tue Mar 18 2014


More information about the Ffmpeg-devel-irc mailing list