[Ffmpeg-devel-irc] ffmpeg.log.20130225

burek burek021 at gmail.com
Tue Feb 26 02:05:01 CET 2013


[00:29] <ezekiel> I spoke too soon - seems that one way the audio is always ahead of video, and the other way audio is always behind
[03:29] <p4plus2> Hey, quick problem, when recording with ffmpeg once I kill ffmpeg to stop the recording, I lose around 1 to 2 seconds of recording it seems to be lost
[03:29] <p4plus2> placing "sleep 1" at the beginning of my kill script fixes it for the most part, but is there a better method?
[06:57] <grepper> I wonder if it has to do with keyframes
[10:11] <Guest1692> hi
[14:37] <bgmarete> Hello all. On GIT Tip, ffprobe ignore the "-ar" command line parameter while ffmpeg uses it.
[16:41] <lightbring3r> hi everyone
[16:41] <lightbring3r> quick question: any idea how i set muxdelay programmatically?
[17:02] <hughmanwho> How can I find the analyze duration?  I'm getting an error message telling me that I should consider increasing the value for 'analyzeduration' and 'probesize' options
[17:05] <klaxa> just add them as parameters i.e. ffmpeg -analyzeduration 9001 -probesize 4096 -i <input> <output>
[17:06] <klaxa> hughmanwho: ^
[17:07] <jeje2> Hi to all
[17:12] <jeje2> I use FFMPEG libraries (last released 1.1.2 I have recompiled) to decompress H264 video stream from IP camera. I'm facing a cpu usage problem. FFMPEG can decompress video stream like I want. But I have a limitation of simultaneous decompression. I can decompress 3 streams at 0-2% of CPU usage. But if I try to decompress an fourth at the same time, my CPU usage grows up to 20%. I really can't understand why...
[17:12] <jeje2>  I someone have an explanation (and a solution). Regards.
[17:13] <jeje2> I can make a pastebin code if necessary
[17:25] <hughmanwho> What is a good probesize?  When it's at 5,000,000 which it was set to as default in code, then it takes 75 seconds to run 'avformat_find_stream_info', when it's twice that it takes 123 seconds, when it's at 4096 it crashes because it it's long enough.. I can definitely keep playing and narrow in on a good value but anyone have any experience there?
[17:30] <hughmanwho> Actually that was probably faster to find than it was to ask
[17:31] <hughmanwho> I went with 131072 (2 to the 17th), enough that it didn't crash after a few tries and also executed within 2 seconds
[17:57] <jeje2> in fact I make a class in c++ (using Visual C++) to integrate the FFMPEG libraries on windows XP OS. I use dynamic pointers on the dll files. This class is used by my application to initialize the library, calling avregister once (I also call av_lockmgr_register because I'm in a multithreaded application). And after When receiving an RTP packet (I formatted from RTP H264 stream to understanding packet for FFMPEG).
[17:57] <jeje2>  Only the first time (by stream) I find the decoder  m_lpCodec = lpfnavcodec_find_decoder(CODEC_ID_H264), I do my m_lpCodecCtx = lpfnavcodec_alloc_context3(m_lpCodec) , I open the decoder avcodec_open2( m_lpCodecCtx,  m_lpCodec, &optionsDict); with optionsDict=NULL, I alocate my frame m_lpFrame = lpfnavcodec_alloc_frame(); After all these initialization, at each packet, I do lpfnav_init_packet(&m_lpPacket);
[17:57] <jeje2>  m_lpPacket.data = lpImageData;m_lpPacket.size = dwImageSize;  then iRes = lpfnavcodec_decode_video2( m_lpCodecCtx, m_lpFrame, &iFrameFinished, &m_lpPacket); and lpfnav_free_packet(&m_lpPacket);
[17:57] <jeje2>  after decompression.
[17:58] <jeje2> But I can't understand why CPU usage is so different between 3 and 4 simultaneous decompressions. I don't set any flags to the codeccontext and codec.
[18:13] <jaffa4> hi
[18:14] <jaffa4> How do I compile ffmpeg versoin that has mp3 enconding enabled that can be used by vlc to encode mp3?
[18:31] <ddhahn> hey all.. using ffmpeg 1.1.2 on RHEL. Compiled from source with these options --extra-cflags=-O2 --enable-bzlib --disable-devices --enable-libfaac --enable-libgsm --enable-libmp3lame --enable-libschroedinger --enable-libtheora --enable-libvorbis --enable-libx264 --enable-libxvid --enable-avfilter --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libvpx --enable-libspeex --enable-postproc --enable-pthreads --disable-static 
[18:31] <ddhahn> nable-gpl --disable-debug --disable-optimizations --disable-stripping --extra-cflags=-fPIC --extra-ldflags=-fPIC --enable-nonfree --enable-version3 --libdir=/usr/local/lib
[18:32] <ddhahn> some conversions log a lot of "que input is backward in time" from the audio encoder, and PTS, DTS issues.. The output files work, but I can't really tell if i should care about the messages.
[18:36] <ddhahn> Here's a complete log from a conversion.. http://pastebin.com/n0mwUtqB
[18:38] <ddhahn> Although the output works, in that in plays in a media player, there seems to be small jumps in the audio.
[18:45] <ddhahn> command line was: ffmpeg -i /opt/kaltura/web//content/entry/data/4/267/0_6isiwutm_0_luyg7xky_2.wmv -vcodec  libx264 -subq 2 -qcomp 0.6 -qmin 10 -qmax 50 -qdiff 4 -b 750k -s 528x352 -r 29.97 -g 60 -acodec libfaac -ab 96k -ar 44100 -f mp4 -y /opt/kaltura/tmp/convert/tmp_convert_51262dc767d03 >> "/opt/kaltura/tmp/convert/convert_0_6isiwutm_676be.log" 2>&1
[18:47] <ddhahn> for a little more background, I am using ffmpeg in the context of Kaltura 5.0, CE (kaltura.org).. Needed to update ffmpeg from the included version to support the wmalossless codec.
[19:10] <relaxed> ddhahn: why would you add these --disable-optimizations --extra-cflags=-O2
[19:11] <relaxed> use one of the libx264 presets
[19:14] <Catoptromancy> moo
[19:18] <ddhahn> I was mirroring how the ffmpeg was compiled that was distributed with kaltura is the only reason
[19:18] <ddhahn> not really understang fully what all the options were, i tried to stick to that..
[19:22] <ddhahn> would you suggest removing those options and recompiling if that changes the behavior at all?
[19:23] <ddhahn> as far as the command line options, I need to dig in a bit to kaltura to see where those options are set.. The system automatically generates a command line to use with ffmpeg.
[19:27] <ddhahn> I am wondering if the issue lies somewhere in the wmalossless decoding.. I am testing some other videos that vary in codecs used to see if the same issue arises.
[19:27] <relaxed> Amazing! I wouldn't have much faith in a company shipping out ffmpeg with --disable-optimizations.
[19:28] <relaxed> ddhahn: probably, check out the latest source using git and see if you have still have the problem.
[19:28] <durandal_1707> ddhahn: it could be bug in demuxer/decoder so please provide input sample
[19:28] <relaxed> and, of course, remove --disable-optimizations --extra-cflags=-O2
[19:29] <ddhahn> relaxed: thanks, I'll do that
[19:29] <ddhahn> durandal_1707: is there a preferred way to share input files?
[19:30] <Catoptromancy> I am trying to get libavformat.so but compiling only seems to make libavformat.a
[19:30] <Catoptromancy> i must be doing something wrong
[19:30] <Catoptromancy> seems to compile flawlessly
[19:31] <durandal_1707> ddhahn: just upload part(if it is very big) of it  somewhere (make sure that part is playable firs)
[19:31] <durandal_1707> Catoptromancy: --enable-shared
[19:31] <relaxed> Catoptromancy: --enable-shared
[19:31] <relaxed> I won
[19:31] <ddhahn> Catoptromancy: did you configur with --enable--shared?
[19:31] <Catoptromancy> thax!
[19:31] <ddhahn> er.. --enable-shared
[19:51] <ddhahn> durandal_1707: test file --> http://jabba.servebeer.com:51234/store/testfile.wmv
[19:51] <ddhahn> that's the whole thing..~200MB or so.. not too bad
[19:54] <durandal_1707> ddhahn: and when you get clips in audio?
[19:58] <durandal_1707> does audio othewise plays fine wit ffplay/ or if you transcode it to flac?
[19:58] <durandal_1707> from log errors it probably means dts/pts is causing libfaac encoder to misbehave
[19:58] <ddhahn> have not tried transcoding to flac
[19:59] <ddhahn> it plays mostly ok, but in vlc I noticed there is a small a/v sync problem
[19:59] <ddhahn> So with that, and the errors, I am not sure if something is amiss or not
[20:03] <ddhahn> durandal_1707: yea, small clips in the FLV output
[20:04] <ddhahn> and in mp4, sorry
[20:53] <ddhahn> Just tried compiling from fresh git clone.. same issue, though someone changed "que" to "queue" in the log files :)
[21:19] <HorizonXP> hi there, I'm trying to use avconv on Ubuntu to capture frames from a video. I have this working
[21:19] <Fjorgynn> and?
[21:19] <HorizonXP> However, I'd like to name the output files such that they're named "screen_xxxx.png"
[21:19] <HorizonXP> where xxxx = the timestamp of the frame
[21:20] <HorizonXP> I'm wondering if this is possible via avconv
[21:20] <HorizonXP> or perhaps I'd have to script ffmpeg via a python script
[21:20] <Fjorgynn> this is ffmpeg
[21:20] <durandal_1707> resistance is futile
[21:21] <saste> HorizonXP, not at the moment, WIP and you should find a related ticket on trac
[21:21] <saste> (check for words image2)
[21:21] <saste> come back in a month and maybe it will be implemented
[21:21] <durandal_1707> why so long?
[21:22] <zoktar>  Hello, im trying to caputure desktop and some games i play, but rather than showing the poor fps in a game for example, it speeds up to try and match the FPS, how do i record with a fixed fps regardless of the real fps ?. thanks.
[21:22] <HorizonXP> saste: WIP to output timestamps?
[21:22] <saste> HorizonXP, WIP to put the timestamp in the image filename
[21:23] <saste> you have several tools to inspect the frames timestamps (showinfo filter, ffprobe, ffmpeg timestamp options)
[21:23] <HorizonXP> saste: can you tell me what to search for in trac?
[21:23] <HorizonXP> maybe I can implement it
[21:23] <saste> HorizonXP, there is a discussion on ffmpeg-devel
[21:24] <saste> i'm currently stuf since i have too much beef on my fire, and little time
[21:24] <HorizonXP> saste: http://ffmpeg.org/pipermail/ffmpeg-devel/2012-January/118549.html ?
[21:24] <HorizonXP> someone added a patch in January?
[21:24] <saste> but the idea is that we should support some form of output pattern specification
[21:26] <HorizonXP> saste: I would agree with that
[21:26] <HorizonXP> sequential file naming and timestamps isn't the whole story
[21:26] <HorizonXP> some kind of pattern would be nice
[21:26] <HorizonXP> but it is a start
[21:26] <HorizonXP> I'll try that patch out for now, since it will accomplish what I need
[21:26] <saste> http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/158092/focus=158105
[21:26] <saste> ^^
[21:31] <llogan> relaxed: how often do you update the ffmpeg source on your static builds?
[21:31] <HorizonXP> saste: thanks :)
[00:00] --- Tue Feb 26 2013


More information about the Ffmpeg-devel-irc mailing list