[Ffmpeg-devel-irc] ffmpeg.log.20150110

burek burek021 at gmail.com
Sun Jan 11 02:05:01 CET 2015


[00:18] <grepper> damn, the fast seek doesn't work with avconv 11.1 , I get black frames from mpg. Too bad I decided to support both avconv and ffmpeg
[00:18] <c_14> check the version header and fallback?
[00:19] <c_14> If it says 'Copyright the FFmpeg Developers' it's FFmpeg
[00:19] <c_14> I mean, you have the fallback anyway for old versions.
[00:24] <grepper> yeah, I might just have to code yet another work-around
[00:31] <grepper> 3 fallbacks needed really, 1. use transcode 2. use 'select' filter if new enough 3. use -ss -i and loops if not avconv 4. slow seek using loops
[00:31] <grepper> heh, no wonder this script has become so long
[00:41] <JEEB> fast seek as in container side seek?
[00:41] <JEEB> that should work just fine in avconv as well, you just have to put it before -i instead of after IIRC
[00:42] <c_14> JEEB: <grepper> damn, the fast seek doesn't work with avconv 11.1 , I get black frames from mpg. Too bad I decided to support both avconv and ffmpeg
[00:42] <JEEB> yeah, mpeg-ts/ps can be problematic depending on what exactly you're doing.
[00:43] <JEEB> if you want to get it fixed on the other side I recommend you report it there if you have the time
[00:44] <JEEB> c_14, I mostly missed the "black frames with mpg (I guess mpeg-ps)" part
[00:44] <JEEB> I think VFR Maniac or so added some timestamp stuff to raw video streams which might be helping with a container like that (which doesn't by itself contain hints for seeking)
[00:45] <grepper> maybe I should stick with the select filter for mpg then ?
[05:33] <bofh> Hello there! I've created the MP3 file using ffmpeg, and now I can see that "ffprobe file.mp3" returns: "start: 0.025057"
[05:34] <bofh> what us "start" and why it is > 0?
[05:46] <bofh> Duration: 00:00:54.00, start: 0.025057, bitrate: 128 kb/s to be more precise
[05:46] <bofh> I don't really get what is that "start"
[05:52] <BtbN> propably when the audio starts
[05:53] <BtbN> before that, there's just silence
[05:53] <bofh> well, but there's nothing like silence
[05:53] <BtbN> 0.02 seconds isn't that long
[05:54] <bofh> yeah, but when merging with video
[11:52] <meorom> hi all!
[11:52] <meorom> now I can compile ffmpeg for Windows success
[11:53] <meorom> but in lib folder, I only see *.a, it doesn't create *.lib
[11:53] <meorom> if I compile dynamic lib, it still create *.dll
[11:54] <meorom> I don't know what I am missing ?
[11:54] <meorom> I'm also thanks to c_14
[11:55] <meorom> I use config: --toolchain=msvc --target-os=win64 --arch=x86_64 --prefix=build
[12:44] <Fjorgynn> :)
[12:44] <Fjorgynn> merzo:
[12:44] <Fjorgynn> wrong
[12:48] <meorom> ?
[12:48] <meorom> could you pls tell me more ?
[12:49] <meorom> ah, sorry
[12:49] <meorom> wrong person :)
[12:51] <Mavrik> so what exactly are you trying to do?
[12:52] <meorom> me ?
[12:54] <Mavrik> yes.
[12:55] <meorom> I follow http://ffmpeg.zeranoe.com/forum/viewtopic.php?f=5&t=1309
[12:56] <meorom> now I want to build static lib, but it only create *.a, not *.lib
[12:57] <meorom> I'm also check https://trac.ffmpeg.org/wiki/CompilationGuide/MSVC
[12:57] <Mavrik> 1.) Those instructions are horribly out of date.
[12:57] <Mavrik> 2.) You still didn't tell what your end goal is. Do you need the binary? Static lib? Dynamic lib?
[12:57] <meorom> my goal is build static lib
[12:57] <Mavrik> 3.) You pasted different guides and you didn't say which compiler do you want to use :)
[12:58] <Mavrik> meorom, well, check if you have MSVC2013 first
[12:58] <Mavrik> then add "--enable-static" to configure right? :P
[13:00] <meorom> yes, actually I'm using MSVS 2013, and compiler MSYS MingW
[13:01] <Mavrik> ugh.
[13:01] <Mavrik> No, either you're using one or the other.
[13:01] <Mavrik> Anything else will cause you a huge pain.
[13:01] <Mavrik> so either mingw/gcc or msvc compiler
[13:02] <meorom> I mean I follow exactly in https://trac.ffmpeg.org/wiki/CompilationGuide/MSVC, only different thing I'm using MSVC 2013
[13:03] <meorom> I've just finished build with --enable-static, but it's still not create *.lib, it's only has *.a
[13:04] <Mavrik> .a is a static lib
[13:04] <Mavrik> it's created by gcc compiler.
[13:04] <Mavrik> if you get .a instead of .lib you're using the wrong compiler.
[13:05] <Popara> Hello, why with -f mpegts ... the bitrate is visible like bitrate=2400kbps under stats and when i output it via HLS it says bitrate=N/A
[13:05] <meorom> so, can I check which compiler is using ?
[13:06] <meorom> and you said that this document is out of date, could you show me newer doc ?
[13:13] <Mavrik> Popara, probably because HLS muxer doesn't have that implemented.
[13:13] <Mavrik> meorom, the MSVC compilation on Wiki or FFmpeg on homepage should be up to date
[13:13] <Mavrik> not the zeranoe link
[13:26] <meorom> Mavrik: but if I use "--enable-shared", it can create *.dll and *.lib
[13:27] <Mavrik> because gcc creates .dlls when using shared configuration and the glue library.
[13:28] <meorom> okey, so could you show me a way to check this ?
[13:29] <meorom> in doc "FFmpeg can be built with MSVC 2012 or earlier using a C99-to-C89 conversion utility and wrapper, or with MSVC 2013 natively."
[13:29] <meorom> so, how can I build "MSVC 2013 natively", could you tell me more ?
[13:30] <meorom> maybe here I'm using MSVC 2013, so fail result.
[13:32] <meorom> I'm a noob, so you shouldn't angry :)
[13:35] <Mavrik> Ok, again. MSVC Linker will create a .lib static library, gcc will create a .a
[13:35] <Mavrik> if you get a .a, your ffmpeg is being compiled with gcc not msvc
[13:35] <Mavrik> I can't really help you more than that because I'm not at your machine.
[13:35] <Mavrik> If you're getting a .dll you're compiling in shared library configuration.
[13:39] <Popara> Mavrik how hard it would be to add the bitrate into the hls? :/
[14:23] <ribasushi> greetings
[14:24] <ribasushi> is there something like "matroska for streaming" protocol? the goal is to be able to take a random source and -c copy it to something that will take any stream and wrap it apropriately
[14:24] <ribasushi> mpeg-ts and rtp are both limited in what they will take
[14:33] <ribasushi> the silence is unnerving, hopefully it's simply too early, instead of "no such thing" :)
[14:38] <Mavrik> hrmf
[14:38] <Mavrik> don't remember any, MPEG-TS is the standard for streaming
[14:41] <sfan5> ribasushi: it looks like matroska does support streaming http://matroska.org/technical/streaming/index.html
[14:41] <ribasushi> let me step back a bit for a bigger picture
[14:42] <ribasushi> the current setup I have is: ffmpeg -i <some hardware> -c copy -f matroska <diskstore.mkv> -c copy -f matroska /dev/stdout | ffmpeg -i /dev/stdin -f rtp rtp://127.0.0.1:30321
[14:43] <ribasushi> the rationale is to be able to "tune into" what is currently being recorded, while using virtually 0 cpu otherwise (hence the copy's)
[14:43] <ribasushi> the problem comes when <some hardware> spews out non strict-conformant streams
[14:44] <ribasushi> what I am looking for is a reliable "tune-in" mechanism
[14:44] <ribasushi> neither rtp nor mpeg-ts give me that
[14:45] <Mavrik> well.
[14:45] <Mavrik> Nothing much you can do unless you tighten up your requirements.
[14:47] <ribasushi> hm hm hm... but that's so uncool ;)
[14:47] <ribasushi> there got to be some generic format to consume "random frames from the aether"
[14:48] <Mavrik> you're not streaming random frames
[14:48] <Mavrik> you're streaming encoded data which can be delimited in bunch of ways and multiplexed in bunch of ways
[14:49] <Mavrik> and that encoded data tends to have predispositions, like headers carrying data that's critical for decoding ;)
[14:50] <ribasushi> nod
[14:51] <ribasushi> aight will ponder more, thanks
[14:51] <Mavrik> but I'm rather curious what kind of hardware outputs stuff that's not MPEG or H.264 :)
[15:33] <bofh> Hello again
[15:33] <sfan5> hi
[15:33] <bofh> I need some advice of how to remove that non-zero "start" something in generated mp3
[15:34] <bofh> ffprobe 1.mp3
[15:34] <bofh> Duration: 00:00:54.00, start: 0.025057, bitrate: 128 kb/s
[15:34] <bofh> what is that "start" and why it is non-zero?
[15:36] <shadowchancellor> Hey guys I've just managed to convert some LPCM to WAV, however, it seems the LPCM is muxed with some kind of video. How would one go about extracting and encoding the video portion?
[15:37] <relaxed> bofh: ffmpeg -i input.mp3 -ss 0.025057 -c copy out.mp3
[15:38] <relaxed> shadowchancellor: why not encode the video and copy the audio without extracting anything?
[15:38] <sfan5> shadowchancellor: ffmpeg -i input_file -an -c:v video_codec output_file
[15:38] <sfan5> video_codec can be "copy" if you just want to extract the video
[15:38] <bofh> relaxed: already tried that
[15:39] <bofh> the resulting file has exactly the same start
[15:39] <bofh> relaxed: not really the same, Duration: 00:00:53.94, start: 0.011995, bitrate: 128 kb/s
[15:43] <relaxed> can you put it up somewhere?
[15:44] <bofh> sure
[15:47] <bofh> relaxed: http://s3.amazonaws.com/animatron-snapshots-dev/8db3b0540b6f95cdef44ae07.mp3
[15:48] <Popara> Hey guys can i ask something, for a bug that WASN'T exists in FFmpeg 2.4.x (it exists in FFmpeg 2.5.x) how much time it will take to be fixed in the new version? It should be something simple. just asking nothing more :)
[15:48] <rcombs> link to the ticket?
[15:49] <Popara> https://trac.ffmpeg.org/ticket/4247#ticket
[15:49] <rcombs> it's not really possible to provide an estimate on when something will be fixed without knowing anything about the issue
[15:49] <Popara> and this one
[15:49] <Popara> https://trac.ffmpeg.org/ticket/4247
[15:49] <Popara> i made them now
[15:49] <Popara> yes i'm just saying it because in ffmpeg 2.4.x it was fine
[15:51] <rcombs> it'll be kinda hard to work on that without a media sample
[15:52] <Popara> A sorry wrong ticket ( the second one is this: https://trac.ffmpeg.org/ticket/4243 )
[15:52] <Popara> yeahi write in the ticket so its fine
[15:52] <Popara> About the sample, i can't make it, because when i get the sample with the new version it closes, when i get it with ffmpeg 2.4.x it seems that the sample is already fixed so it works on ffmpeg 2.5.x as well
[15:53] <Popara> Also, for me it appears to be a problem with the -re argument mostly
[15:54] <Popara> I also tried with FFmpeg 2.5.0 sam results. So something changed between 2.4.5 and 2.5.0 and i don't know what
[15:55] <rcombs> why are you passing -analyzeduration 15000000?
[15:55] <Popara> it just general solution because i use it for all of my streams
[15:55] <Popara> some streams require larger analyze duration
[15:55] <Popara> but as i said even in this case, in ffmpeg 2.4.x was fine
[15:55] <rcombs> provide logs with -v verbose
[15:55] <Popara> ok
[15:56] <Popara> i will do that now
[15:57] <rcombs> and yeah, you need to provide a sample that reproduces the issue
[15:57] <rcombs> otherwise nobody will be able to do anything about it
[15:57] <relaxed> bofh: hmm, I'm not sure how to set it to zero. Are you editing something with it?
[15:58] <Popara> i will try maybe to do something using wget to get the stream since it's over HTTP
[15:58] <Popara> and then parsing into ffmpeg i will see it now
[15:59] <bofh> relaxed: I'm adding it to the video stream
[15:59] <bofh> and somehow it is not in sync
[16:03] <xreal> Is it good to store anything mkv, if mkv is readable from my player?
[16:03] <xreal> anything in*
[16:03] <relaxed> bofh: what about, ffmpeg -i 8db3b0540b6f95cdef44ae07.mp3 -ss 0.025057 -y output.flac
[16:04] <Popara> rcombs that is really so strange. I grabbed a sample using WGET, i add it into FFmpeg to re-parsing it into MPEGTS. With OR Without -re it works (because i guess it's a file) . SO, what i understand is that happens when the connections is remotly and when i use -re as well
[16:05] <relaxed> bofh: without the -ss it will remove the start delay, too
[16:05] <Popara> So somthing happens with the connection when i use -re and messes the things up. Now the only way to provide the sample is to give the direct link
[16:06] <bofh> hm
[16:06] <bofh> let me try something
[16:07] <anshul_mahe> in ffmpeg hls_wrap is not working
[16:08] <c_14> anshul_mahe: what is it doing, what are you expecting it to do, what version are you running, full commandline and output on a pastebin please
[16:10] <xreal> relaxed: do you know, if mp3->flac is done in 32-bit mode? mp3 is bitless, flac can handle up to 32 "only".
[16:11] <anshul_mahe> I here is http://pastebin.com/JhFT7HHc
[16:11] <bofh> relaxed: do you know if I could play videostream from mp4 and audiostream from external file, and start playing of audio from some time?
[16:12] <anshul__> sry for noise I wanted hls_list_size i was using hls_wrap
[16:18] <relaxed> bofh: where is the delay happening with the audio/video? does the audio start late?
[16:19] <relaxed> xreal: mp3 and flac support s16
[16:20] <relaxed> (and s32 too)
[16:28] <xreal> relaxed: MP3 is floating, you can push 32 bit into it without a problem.
[16:31] <bofh> relaxed: audio seem to start late
[16:31] <bofh> so I just want to cut off that initial delay of 0.02 seconds and see what will happen
[16:31] <xreal> The FLAC format supports from 4-bit to 32-bit sample sizes.
[16:31] <xreal> libFLAC does not support encoding or decoding anything above 24-bit.
[16:33] <relaxed> ffmpeg's flac encoder supports 32bit
[16:33] <xreal> relaxed: but 32-bit decoding is only supported ifmid/side is disabled due to bitstream reader limitations.
[16:34] <xreal> relaxed: so, mp3->flac will be done in 32bit? nice.
[16:36] <relaxed> bofh: ffmpeg -i video.mp4 -itsoffset -0.02 -i audio.mp3 -map 0:v -map 1:a output
[16:36] <relaxed> that might work
[16:39] <xreal> Damn, my reported bug is reproduced by developers, but open for 9 months now!
[16:39] <relaxed> xreal: it's done in 16 bits.
[16:40] <xreal> relaxed: that really suxx. I'll better use sox for this or do it with piping.
[16:40] <relaxed> I don't know what you're going on about, it's completely lossless
[16:41] <xreal> relaxed: I talked about it on hydrogenaudio about 10 years ago. They suggested to do it in 24 or 32 bit.
[16:43] <relaxed> I still don't know what you're saying. If there's an issue with ffmpeg you should file a bug report, or point to one if you've already filed it.
[16:44] <xreal> relaxed: I don't file bugs anymore. They don't get resolved after all.
[16:47] <relaxed> If you're against filing them then there's no point in discussing them here.
[16:48] <durandal_1707> xreal: which bug?
[16:48] <relaxed> there is no bug
[16:48] <bofh> relaxed: hmm, it says something about the AAC encoder
[16:48] <bofh> which isn't my case
[16:49] <relaxed> pastebin
[16:51] <xreal> durandal_1707: https://trac.ffmpeg.org/ticket/3576
[16:53] <bofh> relaxed: http://pastie.org/private/oxmmgebszp8ygtyarisdq
[16:55] <relaxed> ok, do you want the video from the first input and the audio from the mp3? If so, -map 0:v -map 1:a only use those streams and omit the aac stream.
[16:55] <bofh> it seems that mp4 contains the AAC stream
[16:55] <bofh> may it happen that the AAC stream is different from mp3 in terms of timing?
[16:56] <relaxed> I have no idea, play it and see
[16:57] <relaxed> but you should be able to sync the mp3 with itsoffset
[16:58] <bofh> yeah
[16:59] <bofh> ffmpeg -i 993db15476e8b2cd1674d064.mp4 -itsoffset -0.02 -i 993db15476e8b2cd1674d064.mp3 -map 0:v -map 1:a test.mp4
[17:00] <bofh> [aac @ 0x7f26b37bbb40] The encoder 'aac' is experimental but experimental codecs are not enabled, add '-strict -2' if you want to use it.
[17:00] <xreal> Now durandal is gone :D
[17:00] <bofh> weird
[17:01] <xreal> bofh: -c:a copy
[17:01] <relaxed> oh, it's converting the mp3 to aac. Use -c copy to avoid that.
[17:01] <xreal> relaxed: -c copy? isn't it -c:a copy?
[17:01] <relaxed> well, -c copy if he want to copy both the audio and video
[17:01] <relaxed> wants*
[17:01] <xreal> relaxed: does -c also do -c:s ?
[17:02] <relaxed> -c copy will copy everything you have mapped from the input
[17:02] <xreal> ah ok
[17:06] <bofh> okay, that worked
[17:06] <bofh> still there's some sync problem, but perhaps not related to ffmpeg
[17:08] <relaxed> try playing with the itsoffset number. it's time for me to hit the sack.
[17:56] <bofh> relaxed: ok, jfyi - I've found the problem and it is not related to ffmpeg anyhow :)
[17:57] <bofh> anyway, thanks for the help!
[17:57] <bofh> it helped me to narrow the scope of the issue and eventually nail down the real issue
[18:00] <allengates> hi
[18:56] <xreal> Children are like Windows Vista: not that far developed and hard to handle.
[19:01] <shadowchancellor> Anyone know what the 'muxing overhead' info that gets logged when converting pcm to wav is?
[19:01] <shadowchancellor> For instance, mine logs 'muxing overhead: 0.026929%'
[19:02] <justinX> it is how much extra space a file takes compared to the involved video and sound
[19:03] <shadowchancellor> so I can assume that ~.02% is extraneous and could be removed?
[19:03] <justinX> space consumed by headers and the way it mixes the parts togheter so sound is near the video for the same time in the file etc...
[19:04] <justinX> it can't be removed unless you want to play the video separately from the sound :-D
[19:05] <justinX> in videos I mean. for pure sound files... I guess it is just headerS?
[19:05] <justinX> telling what sound file format it is I suppose?
[19:07] Action: justinX imagines a world without that info.. where everything would have to be packed into the filename instead
[19:30] <fvjfadskj> hi
[19:30] <fvjfadskj> i'm piping an image stream to ffmpeg using image2pipe
[19:30] <fvjfadskj> and taking timestamps with wall clock option
[19:30] <fvjfadskj> and meanwhile taking sound from directshow
[19:30] <fvjfadskj> sound and video are out of sync
[19:30] <fvjfadskj> how to fix?
[20:09] <fvjfadskj> nobody knows?
[00:00] --- Sun Jan 11 2015


More information about the Ffmpeg-devel-irc mailing list