[Ffmpeg-devel-irc] ffmpeg.log.20150203

[burek]

[00:04] <hetii> hmm
[00:04] <hetii> ha no windows:(
[00:04] <hetii> *have
[00:05] <klaxa> like i said, you could try wine
[02:04] <fixxxermet> Is ffserver still maintained?
[06:29] <teckris> hi guys
[06:29] <teckris> can someone help me with ffmpeg libraries
[06:30] <teckris> i am using FFMPEG with QtAV library
[06:30] <teckris> in which i use streaming
[06:30] <teckris> I need to have buffer used in streaming
[06:32] <teckris> in functions AVFORMAT_CONTEXT is used by player
[06:32] <teckris> so i need to control how the structure is filled for which i would like to get byte buffer
[06:50] <Diaz> hi all,
[06:50] <Diaz> I use libav 0.8.6
[06:50] <Diaz> for ape demux and ape decode
[06:50] <Diaz> I got a big trouble
[06:51] <Diaz> since ape decode error when some ape frame input
[06:51] <Diaz> does any one have ideas?
[06:51] <Diaz> about the layout of ape format?
[06:51] <Diaz> HELP!
[07:15] <Diaz>  does anyone know where I can find the doc that describe ape header format?
[07:15] <Diaz> HELP?
[07:19] <Diaz> does anyone know where I can find the doc that describes monkey's audio format?
[07:48] <pentanol> what's ape?
[07:48] <pentanol> Diaz:
[07:51] <Diaz> monkey's audio
[07:51] <Diaz> do you ever use that?
[07:51] <pentanol> no
[07:52] <pentanol> I use audacity
[07:52] <Diaz> i have big trouble about ape decode error for a long time when I use libav 0.8.16
[07:53] <Diaz> I do not find any doc that describe monkey's audio format
[07:54] <pentanol> try ffprobe .ape file
[07:54] <pentanol> I'm not sure this format is supported
[07:55] <Diaz> ffprobe for what?
[08:00] <pentanol> that tool from ffmpeg
[08:02] <pentanol> yes that supported by libav  D.A..S ape                  Monkey's Audio
[08:03] <pentanol> only for reading those .ape files
[08:08] <Diaz> just parse the ape file , right?
[08:08] <Diaz> pentanol
[08:13] <pentanol> I mean you can use ffmpeg for that in that programm or ffmpeg directly in cli ffmpeg -i file.ape -acodec mp2 -ac 2 -ar 192kb file.mp2
[08:17] <Diaz> ok, you mean I can use ffmpeg for compare
[08:27] <rcombs> Diaz: this isn't #libav
[08:28] <rcombs> topic
[08:28] <rcombs> erm
[08:28] <rcombs> well the topic doesn't actually mention that
[08:35] <Diaz> OOOO, I can not write the question here?
[16:14] <jfmcarreira> Heyy guys
[16:15] <jfmcarreira> there is any error concealment support in ffmpeg video decoders? is this a interesting topic for ffmpeg developers?
[17:02] <shevy> anyone happens to know some kind of audio filter, perhaps even in ffmpeg, that can filter away background noise? it is a .mp3 file recording of a speaker/lecture and there is some constant background noise (always the same noise) that I'd love to somehow get rid off
[17:02] <fixxxermet> Could anyone comment on the state of ffserver?  Is it maintained?
[17:03] <kepstin-laptop> shevy: you'd probably want to look into using audacity to do that, it's a dedicated audio editor that, among other things, has filters designed to remove background noise.
[17:07] <shevy> but I like ffmpeg more :(
[17:39] <RobertNagy> When seeking is it possible to set the seek direction? i.e. I have a huge flv file 100GB and I need to seek to a position near the end, which currently takes a very long time. Alternatively is there some way to speed up seeking by e.g. seeking 99GB into the file before doing a slow seek?
[17:39] <RobertNagy> note, I am using the -ss before -i option
[17:55] <BtbN> RobertNagy, the problem is, ffmpeg has no information where in the file a specific timestamp is. So all it can do is read through the whole thing.
[17:56] <RobertNagy> hm, and there is no way to give it a hint?
[17:56] <RobertNagy> would creating a segmented file help?
[17:56] <RobertNagy> i.e. is it possible to seek segments?
[17:57] <BtbN> a container with a global header might be faster, but i have no idea how much information for example the mp4 header contains, and if it helps with seeking.
[17:57] <RobertNagy> I'm seeking in the file while it is being written to
[17:59] <BtbN> What ffmpeg version are you using?
[17:59] <RobertNagy> latest
[17:59] <RobertNagy> 2.5.3
[17:59] <BtbN> The -ss input option(in front of -i) is the fastest you can get then, unless someone knows some magic.
[18:00] <BtbN> You can of course just cut a random section of of the flv file
[18:00] <BtbN> Not entirely sure if that works with flv, but it certainly does with mpegts
[18:00] <RobertNagy> well, not while it's being written to...
[18:00] <BtbN> Why not?
[18:00] <RobertNagy> how would I do that? the file is locked
[18:00] <BtbN> locked?
[18:01] <RobertNagy> the process writing to it...
[18:01] <BtbN> Why would that stop another process from opening it?
[18:01] <RobertNagy> nothing
[18:01] <RobertNagy> but from modifying it
[18:01] <RobertNagy> or maybe I'm misunderstanding something
[18:01] <BtbN> Don't modify it then
[18:02] <RobertNagy> what do you mean by "cut a random section"
[18:02] <RobertNagy> btw, does any format support seeking from the back?
[18:02] <BtbN> tail -c +50G somefile.flv > someotherfile.flv
[18:02] <RobertNagy> oh, I see
[18:03] <RobertNagy> hm, I don't think that will work in either flv or mpegs
[18:03] <RobertNagy> will probably complain about invalid data
[18:03] <BtbN> It certainly will for mpegts
[18:03] <BtbN> not too sure about flv, but might work
[18:03] <BtbN> If the media repeats the SPS/PPS sometimes
[18:03] <BtbN> If there is only one IDR frame at the very beginning, you have to parse the entire thing
[18:04] <BtbN> Just try it and see
[18:04] <RobertNagy> yea, even if it works it won't work for what I need it
[18:05] <RobertNagy> since I want to start reading the file before the final section
[18:05] <BtbN> ...?
[18:05] <RobertNagy> I'm doing a hack with -re + a small initial delay
[18:05] <RobertNagy> to read from a  growing file
[18:05] <BtbN> i don't think ffmpeg will read past the end of the while from the moment it opened it
[18:06] <RobertNagy> it does
[18:06] <RobertNagy> atleast with flv
[18:06] <RobertNagy> and mpegts, however I had problems with seek accuracy with mpegts
[18:07] <BtbN> since ffmpeg 2.1 the -ss input option should be frame-accurate.
[18:07] <BtbN> Also, i don't see your problem.
[18:07] <RobertNagy> what part of it don't you see?
[18:07] <RobertNagy> I can't cut a section that doesn't yet exist
[18:07] <BtbN> tail -f -c +50G somefile.flv | ffmpeg -f flv -i - .....
[18:07] <RobertNagy> aah
[18:07] <RobertNagy> I see
[18:08] <RobertNagy> if tail works with a growing file that is
[18:08] <BtbN> That's what -f is for.
[18:08] <BtbN> seeking to an exact timestamp might get complicated with that method though
[18:08] <RobertNagy> hmm
[18:08] <RobertNagy> ok, what if I instead of recording to a huge file, recording into segments using the -f segment
[18:08] <RobertNagy> would that help?
[18:09] <RobertNagy> i.e. can I do frame accurate seek and playback of segments?
[18:09] <RobertNagy> I guess I could figure out the frame accurate part by checking the segment start times and picking the right segment
[18:09] <BtbN> sure, if you know the start timestamp of a segment, just seek the time-offset into the segment
[18:09] <RobertNagy> however, will ffmpeg play segments?
[18:10] <RobertNagy> however, will ffmpeg play segments flv?
[18:10] <BtbN> If you use mpegts, you can just concat the files
[18:11] <RobertNagy> this is getting complicated
[18:11] <BtbN> mpegts is a very primitive format
[18:11] <RobertNagy> would seeking in a large file possibly be faster with mpegts?
[18:12] <BtbN> it's just a long line of 188 byte segments.
[18:12] <BtbN> no, it has to parse the entire thing anyway
[18:12] <RobertNagy> hm
[18:12] <BtbN> but you can freely cut and concat mpegts files
[18:12] <BtbN> it has no headers at all
[18:12] <BtbN> so if you have a dir of 1h segments, you just concat the ones you want to play, and pipe them into ffmpeg
[18:17] <RobertNagy> and what about the timestamps
[18:17] <RobertNagy> will they always be 0 from the first segment?
[18:17] <BtbN> They will be 0 from the segment you start at.
[18:17] <RobertNagy> ic
[19:55] <Schrostfutz> can I use ffmpeg to rename audiostreams in files?
[20:13] <llogan> Schrostfutz: can you explain in more detail?
[20:14] <Schrostfutz> llogan: I have video files with two languages, both lack labels so my playback device does not automatically select the correct one
[20:15] <llogan> ffmpeg -i input -c copy -map 0 -metadata:s:a:0 language=sme -metadata:s:a:1 language=nor output
[20:29] <Schrostfutz> llogan: thank you!
[21:12] <Schrostfutz> llogan: can I savely use the same in and output file?
[21:15] <llogan> Schrostfutz: you can't use the same file as input and output
[21:16] <Schrostfutz> llogan: too bad
[22:02] <vlt> Schrostfutz: Why would you want to do this?! Just remove your input file afterwards and rename the new one.
[22:48] <courtarro> Hi all. I'm playing with ffserver for the first time and I'm unable to use the PixelFormat config variable. Am I doing something wrong? I simply have "PixelFormat yuv420p" in my <Feed> tag. The ffserver is version 1.2.6-7:1.2.6-1~trusty1 from an Ubuntu PPA
[22:49] <courtarro> I get the error "Incorrect keyword: 'PixelFormat' "
[23:52] <fixxxermet> Is it required that ffmpeg stream to a server, or can ffmpeg itself act as the server (from which a client, such as vlc, can directly connect)?
[23:53] <c_14> As long as you only have 1 client you can open a listening TCP socket.
[23:53] <fixxxermet> What if I have more than one client?
[23:53] <c_14> Fixed number or dynamic?
[23:54] <fixxxermet> dynamic, but I know the number ahead of time
[23:55] <c_14> You _can_ use the tee pseudo muxer to create n tcp sockets, but it'll be easier to just use a dedicated server for that.
[23:55] <fixxxermet> OK, so what about the actual syntax?
[23:55] <fixxxermet> I seem to be missing it
[23:55] <fixxxermet> (assume 1 client here)
[23:56] <c_14> tcp://{iporhostname}:{port}?listen
[00:00] --- Wed Feb  4 2015