[Ffmpeg-devel-irc] ffmpeg.log.20130927

[burek]

[00:56] <captainastro> hello everyone...please excuse any protocol violations, but it's been many years since I used IRC
[00:57] <captainastro> I've been working with point to point streaming and I've had some success with UDP, but it's flakey - choppy, drop frames, freezes, etc.
[00:57] <klaxa> well it's UDP
[00:57] <captainastro> I've tried using TCP and RTSP, but I am not getting anywhere
[00:57] <klaxa> it's unreliable
[00:57] <captainastro> yea, udp
[00:58] <captainastro> here's what works: sending side
[00:58] <captainastro> ffmpeg -v debug -fflags nobuffer -f v4l2 -i /dev/video0 -c:v libx264 -tune zerolatency -b:v 100k -pix_fmt yuv420p -c:a libmp3lame -f mpegts udp://trajan.local:1234?pkt_size=188?buffer_size=65535
[00:58] <captainastro> ffplay -fflags nobuffer -i udp://localhost:1234?listen?pkt_size=188?buffer_size=65535
[00:58] <captainastro> on the receiving side
[00:59] <captainastro> when I've substituted TCP, for instance (sending) ffmpeg -v debug -fflags nobuffer -f v4l2 -i /dev/video0 -c:v libx264 -tune zerolatency -b:v 100k -pix_fmt yuv420p -c:a libmp3lame -f mpegts tcp://trajan.local:1234
[00:59] <captainastro> (receiving) ffplay -fflags nobuffer -i tcp://localhost:1234?listen
[01:00] <captainastro> it just hangs.  Same if I do something like rtsp://localhost:1234/live.sdp
[01:00] <captainastro> wiresharc sees the sending machine send out packets to port 1234, but there is no response.
[01:00] <captainastro> I'm stumped
[01:01] <captainastro> I've tried just about everything I can think of, and I think I've looked at everything that google can dig up.  By the ffmpeg docs and the streaming guide both should be working.
[01:03] <captainastro> has anyone had any success with point to point streaming with anything other than udp?
[01:30] <trem0r> Hey guys.  I'm trying to create some music on hold.  It needs to be in 16kbps PCM linear.  I've got an existing wav file that I need to convert.  how can I go about that?
[01:30] <trem0r> when I put -ab <desired rate> the bitrate never changes
[01:35] <klaxa> PCM is raw
[01:35] <klaxa> the bitrate depends on the format
[01:37] <trem0r> Can you explain that to me?  I'm not really an audio guy =)
[01:39] <klaxa> http://en.wikipedia.org/wiki/Pulse-code_modulation
[01:39] <captainastro> You need to encode the audio into something like mp3.  The encoding compresses the stream and lowers the quality when you specify the bitrate
[01:40] <captainastro> PCM streams at whatever rate it was captured at
[01:40] <trem0r> I see.
[01:40] <trem0r> Thank you
[01:43] <trem0r> Funny
[01:43] <trem0r> Just converted it to mp3..lowered the bitrate
[01:43] <trem0r> when I convert it back to a wav file, the bitrate is like 700kbps
[01:44] <klaxa> pcm is raw
[01:45] <trem0r> So are you saying that this isn't possible?
[01:45] <klaxa> yes
[01:45] <klaxa> well
[01:45] <klaxa> no
[01:45] <klaxa> it is possible, but it will sound bad
[01:45] <trem0r> Yeah
[01:45] <klaxa> because 16 kbps is very little
[01:45] <trem0r> It's for a music on hold system
[01:46] <trem0r> Pretty annoying requirement
[01:46] <klaxa> maybe you are talking about 16 khz?
[01:46] <trem0r> Here's the requirement from our voip host.  Perhaps they got it wrong
[01:46] <trem0r> v
[01:46] <trem0r> Please make sure that the file is a wave file with either 16 kpbs linear PCM encoding or 8kbps PCMU encoding.
[01:46] <trem0r> They even mistyped kbps
[01:47] <captainastro> you might need to try different encoding technologies - some are better than others
[01:47] Action: stump thinks that probably means 8 kHz 16-bit mono
[01:47] <trem0r> Lemme test your theory
[01:47] <klaxa> yeah 16 kbps is just a weird requirement for pcm
[06:25] <t4nk022> When I try to use ffserver + ffmpeg to output a live h.264 rtp stream, I got the problem
[06:25] <t4nk022> the timestamp in rtp header may get overflow.
[06:26] <t4nk022> since the pkt.pts, pkt.dts in ffmpeg are stored in 64bits length
[06:26] <t4nk022> timestamp field in RTP header is stored in 32bits length
[06:26] <t4nk022> it may get overflow when long time test. does any one have idea about how to fix the issue?
[06:33] <skifreak> having a problem with ffmpeg recognizing the codec for a right bluray, passthrough, video file. mediainfo shows very similar information between the right and left video files, with only the left containing audio. Tried increasing probesize and analyzeduration to no avail. ffmpeg output: http://dpaste.com/1397661/ mediainfo output: http://dpaste.com/1397663/ I have a sample file (10 MB) of each ready on dropbox if needed
[06:49] <SirCmpwn> I have an ass file and an mkv file
[06:49] <SirCmpwn> I want to include the ass file in the mkv, but I still want softsubs
[06:49] <SirCmpwn> how?
[06:57] <skifreak> Sircmpwn: use the map option. ffmpeg -i <video> -i <subtitle> -map 0:0 -map 1:1 <other options> <output_video>
[06:58] <skifreak> maphttps://trac.ffmpeg.org/wiki/How%20to%20use%20-map%20option
[06:58] <SirCmpwn> the numbers there make me worry that I'll lose any other streams in the video file
[06:58] <skifreak> typo: https://trac.ffmpeg.org/wiki/How%20to%20use%20-map%20option
[06:59] <SirCmpwn> oh, I see
[06:59] <SirCmpwn> thank you
[06:59] <skifreak> np
[09:46] <CentRookie> If one changes the x264 source code slightly, turns a few settings off and recompiles it, do I have to recompile ffmpeg as well?
[09:47] <JEEB> if you use shared libraries then you just have to make sure both compiles are API/ABI compatible.
[09:48] <JEEB> and you can just install/ldconfig and run
[09:48] <JEEB> if static, then ffmpeg has to be rebuilt as well
[09:48] <CentRookie> ic
[09:48] <CentRookie> yes im using shared
[09:49] <CentRookie> Is there a way to add a line into the file header, user data ? I would like to add some infos like original resolution
[09:49] <CentRookie> and camera settings
[09:50] <CentRookie> am looking for the header info in the source code but havent found it yet
[09:51] <JEEB> I'd rather put stuff like that into the container's possible freetext metadata fields
[09:52] <JEEB> but I think you can use some kinds of SEI messages for it?
[09:52] <CentRookie> sei message?
[09:52] <JEEB> I think libx264 had support for it, no idea if you can use them via ffmpeg's wrapper
[09:58] <durandal_1707> -x264-params
[10:01] <CentRookie> how does it help me durandal
[10:13] <simonsimcity> I have a mov-file without MOOV atom ...
[10:13] <simonsimcity> Anyone having an idea on how to repair it using ffmpeg? Or another app running on mac ;)
[10:21] <JEEB> too bad that's the indedx of the file
[10:21] <JEEB> *index
[10:21] <JEEB> which contains the offsets of everything in the file, basically
[10:22] <JEEB> at best you can do some guesswork, and if it's from some specific hardware thing see if it always outputs packets of a similar size
[10:22] <JEEB> there are some tools that claim they can do this guesswork automated, but none of them are free as far as I know
[10:23] <JEEB> and they naturally do not release their heuristics
[10:23] <JEEB> in some cases such things partially work, in other cases they fail badly
[10:23] Action: JEEB shrugs
[10:24] <simonsimcity> JEEB, I took a review using http://aeroquartet.com/movierepair/repair# and it told me, the movie would be repairable by 99% ... but it would cost up to 200¬ for us to do it ...
[10:24] <simonsimcity> Do you think something like this could help? Have you heard of Untrunc before? http://superuser.com/questions/417100/how-to-open-and-repair-an-m4v-or-mp4-video-file
[10:25] <JEEB> oh
[10:25] <JEEB> someone actually started doing something open source?
[10:25] <JEEB> nice
[10:26] <simonsimcity> Oh! sounds good!
[10:26] <simonsimcity> I wrote a small program using QT, ffmpeg and libfaad which rebuilds the index and, given a complete video as an example, recreates the moov atom, thus rendering the file playable again.
[10:26] <simonsimcity> quote-end
[10:27] <JEEB> well, basically good luck
[10:29] <simonsimcity> JEEB, I read that f.e. VLC is able to guess all indexes ... but since the video- and audio-codec also is stored in the missing piece, VLC can't do much.
[10:30] <simonsimcity> Is it possible to just add information about the codecs?
[10:30] <simonsimcity> I have another file, taken with the same equipment ...
[10:31] <durandal_1707> i guess so, but i never compared those bits between files
[10:31] <JEEB> well, the problem is that you have to make the offsets match too
[10:31] <JEEB> so it will be hex editing and abiding by the specification and so forth
[10:32] <JEEB> tools like atomicparsley and L-SMASH's boxdumper probably will come in handy
[10:53] <getsome> hey guys, i recently upgraded to a newer ffmpeg version (1.2.3) and now got the problem that generated thumbnails of videos are really huge (thumbnail of 10mb video is 100mb+)...
[10:53] <getsome> here is the command i am using
[10:53] <getsome> ffmpeg -y -ss 00:00:05 -r 1 -vframes 1 -i <INPUT FILE> -f mjpeg <OUTPUT FILE>
[10:54] <getsome> can anyone point me in the right direction to fix this issue?
[10:55] <durandal_1707> getsome: upgraded from what?
[10:56] <durandal_1707> shouldn't that command generate single file?
[10:56] <durandal_1707> what is your source?
[10:58] <CentRookie> does ffmpeg support QuickSync from intel?
[11:00] <durandal_1707> that question is asked here over and over again
[11:00] <durandal_1707> currently not, but there is something quicksync related cooking iirc
[11:00] <viric> mittens: ok!
[11:02] <CentRookie> too bad
[11:03] <CentRookie> thanks though
[11:03] <CentRookie> just realized that my server got quicksync
[11:15] <getsome> durandal_1707: yes it generates a single image, but it's huge...everything worked fine with FFmpeg version SVN-r26402
[11:20] <durandal_1707> getsome: how big is your source?
[11:21] <durandal_1707> i mean resolution of source image
[11:21] <durandal_1707> just pastebin whole output
[11:23] <getsome> yea sure, just give me a minute
[11:27] <getsome> seems like ffmpeg tries to output a video, not sure though: http://pastebin.com/GtZkYK5b
[11:29] <durandal_1707> hmm than your -vframes 1 should be after -f mjpeg
[11:31] <durandal_1707> getsome: and with 2.0 your command would fail and give big red message
[11:36] <getsome> durandal_1707: just tried it with -vframes 1 after the -f and it seems to work now, gonna do some further testing....it's ages ago when this feature was implemented, seems like no one really cared about upgrading the ffmpeg version, still kinda weird that it worked like this with this ancient ffmpeg version
[11:36] <getsome> oh and ofc a big thank you for your help
[11:52] <caraculo> hi
[11:52] <caraculo> does FFMPEG support the IMC4 pixel format?
[11:53] <simonsimcity> JEEB, you mentioned L-SMASH's boxdumper ... What is L-SMASH? I can just find some text in ... looks like Japanese ...
[11:59] <durandal_1707> caraculo: you have sample?
[12:00] <caraculo> http://www.fourcc.org/yuv.php#IMC4
[12:01] <durandal_1707> that is not sample
[12:01] <durandal_1707> and without sample it is hard to know code that actually adds support for it is correct
[12:03] <caraculo> it is a YUV planar like YV420, but the UV planes are interleaved in a row basis, instead of concatenated.
[12:05] <caraculo> something like this in memory:   Y_R1 Y_R2 U_R 1 V_R1
[12:06] <durandal_1707> whatever
[12:06] <durandal_1707> no sample - no support
[12:07] <caraculo> what do you mean exactly by sample?
[12:08] <caraculo> look, at FOURCC you have the pixel format described in great detail, ar shall we all just spean in "C"?
[12:09] <durandal_1707> well i will just ignore you
[12:10] <caraculo> yes, with so many typos in my last sentence I'd do the same.
[12:17] <durandal_1707> and uv planes are not interleaved at all
[12:18] <caraculo> ?
[12:23] <saste> caraculo, do you have a file (sample) using that format?
[12:23] <durandal_1707> do you want IMC support in swscale or you just need to decode it?
[12:25] <caraculo> I should state my problem better: I need to encode a buffer which is in IMC4 format
[12:28] <durandal11707> and where would you send that buffer?
[12:31] <caraculo> to disk, basicaly encode it (a sequence of those) and then just save em into a file
[12:31] <durandal11707> raw one?
[12:31] <saste> caraculo, IMC2 - this looks more like a codec format than a pixel format
[12:31] <caraculo> lossless FFv1
[12:32] <saste> in other words it can't be represented through a pixel format
[12:32] <durandal11707> saste: it is similar to nv21/12
[12:32] <saste> durandal11707, yes, but you have - one line U, one line V, one line U, ...
[12:33] <durandal11707> in nv it is: UVUVUVUVUVUVUVU and here it is UUUUUUUUVVVVVVVVV
[12:34] <durandal11707> i see no codec there
[12:34] <durandal11707> it could be supported in swscale
[12:35] <caraculo> IMC4 is a planar pixel format.
[12:36] <caraculo> it is just that the planes are not simply concatenated like in YV420
[12:36] <saste> durandal11707, I don't see how you can code the U/V components with pixdesc.h:AVComponentDescriptor
[12:36] <caraculo> now, my input buffer is in such format, I just do not want to transform it to YV240 before encoding it into FFV1.
[12:37] <saste> caraculo, how are your coding skills?
[12:38] <saste> it is trivial to write a decoder for such a raw format, check for example v410dec.c
[12:39] <saste> but i'm still curious, do you have samples using this format?
[12:43] <saste> caraculo, who/what is creating the buffer? why don't you use a more standard "pixel format"?
[12:46] <caraculo> the buffer comes from a camera... the drives is spiting frames in such format
[12:46] <caraculo> driver*
[12:47] <caraculo> so I can access the frame buffer, and I just do not want my application to have too re-format it. So I was wondering if FFMPEG can work witch such an input format.
[12:48] <saste> caraculo, so the short answer is: no
[12:48] <saste> OTOH it should be really trivial to hack your loop for converting the buffer to any of the formats directly supported by ffmpeg
[12:49] <caraculo> thanks saste
[12:50] <caraculo> yes, it is trivial andknow how to do it, I will just introduce some"latency" and lines of code to the application
[15:23] <lei> hi, I have a file test.mp4, I want to convert it to mpeg2ts(contains h.264 and aac), how to write command line ?
[15:25] <lei> hi, fellows.
[15:29] <viric> lei: it's good to show that you already did some effort on your side, if you want to receive help :)
[15:30] <lei> ffmpeg -i srcfile -y -f mpegts dstfile
[15:30] <lei> I asm googleing
[15:30] Last message repeated 1 time(s).
[15:30] <lei> and if you guys can help me, thanks a lot.
[15:32] <lei> hi, viric.
[15:33] <lei> ffmpeg -i srcfile -y -f mpeg2ts -acodec aac -vcodec h264 dstfile
[15:34] <lei> I don't know if this is right.
[15:35] <viric> test it :)
[15:40] <lei> C:\tc\1\input>ffmpeg -strict -2 -i video.mp4 -y -f mpegts -acodec aac -vcodec h2
[15:40] <lei> 64 out.ts
[15:40] <lei> [aac @ 04bcb980] The encoder 'aac' is experimental but experimental codecs are n
[15:40] <lei> ot enabled, add '-strict -2' if you want to use it.
[15:42] <lei> I try  ffmpeg -strict -2 -i video.mp4 -y -f mpegts -acodec aac -vcodec h264 out.ts
[15:42] <lei> but it tells me something wrong.
[15:44] <lei> viric
[15:45] <viric> ah, well, you may need another aac encoder.
[15:53] <blacky> hello
[15:54] <blacky> how can i read video input from declink
[16:00] <blacky> unknown input format show
[16:00] <blacky> how to solve this any idea ?
[16:00] <blacky> dshow*
[16:03] <zap0> declink?
[16:18] <blacky> zap0: decklink
[16:18] <blacky> sorry
[16:19] <zap0> same question    decklink?    what is it?
[16:21] <zap0> oh.. black magic capture card
[16:21] <zap0> they have there own API i think
[16:24] <viric> what tool can add tags to webm files?
[16:24] <viric> author, title, ...
[16:25] <saste> viric, ffmpeg -i ... -codec copy -metadata author=... -metadata title=... out.webm
[16:26] <saste> or you can use the ffmeta demuxer
[16:26] <viric> ah ok. nothing working on a single file, right?
[16:26] <viric> all goes through copying the whole file
[16:26] <viric> ffmeta?
[16:27] <saste> viric, yes you need to create a new file
[16:27] <viric> ok
[16:28] <viric> saste: I guess there are some restrictions on those tags. usual tag names, format of date field... how may I know them?
[16:29] <saste> viric, dunno, try&fail, or check some docs online
[16:29] <viric> ah http://wiki.multimedia.cx/index.php?title=FFmpeg_Metadata#Matroska
[16:29] <viric> that may apply.
[16:29] <viric> thank you saste !
[16:29] <viric> very helpful
[16:30] <saste> viric, you owe me a beer
[16:30] <viric> yes
[17:11] <voltin> Has anybody had any luck taking frame grabs of a rawvideo (Y160) rtsp stream? For some reason FFMPEG is having problems determining the information about the stream including pixel format and size. When I try to specify either option is complains that the option doesn't exist. Here is a pastebin with 3 attempts and their output with debug enabled: http://pastebin.com/JJdQY91d
[17:16] <mittens> viric: did you get vp9 working?
[17:16] <viric> I didn't have the oportunity to test yet
[17:19] <bunniefoofoo> for CBR encoding mpeg2video, is the bitrate constraint enforced per-frame, per-gop or something else?
[17:48] <bunniefoofoo> ok I see it is recalculated every frame
[19:20] <Franklin> running "gst-launch filesrc location="/home/franklin/Downloads/test.avi" ! codebin2 ! audioconvert ! audioresample ! autoaudiosink" only gives me audio, video isn't showing up, am I missing something? please help.
[21:05] <chrisballinger> hey the files i'm creating with a direct stream copy w/ libavformat play back at 3600 fps instead of 24 fps, does anyone have any clues as to where to look more closely?
[21:07] <chrisballinger> i'm not using the command line, im using the libavformat api
[21:09] <chrisballinger> but i have some stuff i prepared yesterday
[21:10] <klaxa> ah right... sorry, not too familiar with the api
[21:10] <chrisballinger> https://gist.github.com/chrisballinger/6733678
[21:10] <chrisballinger> that is the output of ffprobe for the broken file
[21:11] <chrisballinger> the main thing is that r_frame_rate=1/1, duration_ts=2999, duration=0.033322
[21:12] <chrisballinger> a proper file looks more like r_frame_rate=24/1 duration_ts=449850 duration=4.998333
[21:12] <chrisballinger> https://github.com/OpenWatch/FFmpegWrapper/blob/master/FFmpegWrapper/FFmpegWrapper.m#L349
[21:12] <chrisballinger> that's where i do the setup for the direct stream copy
[21:50] <kni> hey guys
[21:50] <kni> I'm trying to set stuff up to I can stream to an external server
[21:50] <kni> and I'm confused as to how I configure it
[21:51] <kni> is the "ACL allow" stuff for the input feeds or the actual output stream?
[22:19] <bunniefoofoo> hi all, looking for some way to get bitrate "more constant", that is to say I have a 5.5 Mbps MPEG2 stream and some of the gops are as high as 7.5 Mbps even though I have bitrate tolerance set to 0.5 mbps
[22:19] <bunniefoofoo> I am using the standard approach max_bitrate==min_bitrate==bitrate and setting rc_buffer_size
[22:20] <bunniefoofoo> I tried setting bitrate_tolerance to the minimum possible value and ended up with corrupted p-frames (or so it seems)
[22:28] <Niatross> Will the dca codec ever come out of the "experimental" phase ("-strict -2")?
[23:01] <shevy> is there a way to apply some kind of audio de-noise filter through ffmpeg? on a .mp3 file
[23:05] <klaxa> i don't think so, you can use sox though
[23:19] <shevy> nooooo!
[23:19] <shevy> :(
[23:19] <shevy> I hate sox
[23:20] <klaxa> how can you hate sox?
[23:20] <shevy> I once modified mortal kombat theme song, years ago with sox
[23:20] <shevy> I kept that .mp3 file for years... I removed it only a few years ago. I used to think it had a great effect, but it sucked!
[23:21] <shevy> the time before that I was using virtualdub + avisynth. that was beautiful
[23:21] <shevy> that's the only thing I miss from my windows days
[23:22] <klaxa> wait, so because you messed up with the configuration sox is bad?
[23:22] <chrisballinger> Here is a stackoverflowized version of my question if that helps: http://stackoverflow.com/questions/19060253/files-created-with-a-direct-stream-copy-using-ffmpegs-libavformat-api-play-back
[23:22] <klaxa> you can also just use audacity if you have a GUI available
[23:22] <shevy> klaxa configuration? no. it's a completely useless app
[23:22] <klaxa> wow
[23:22] <klaxa> seriously?
[23:23] <shevy> yes
[23:23] <klaxa> and what makes you think so?
[23:23] <shevy> klaxa I used it
[23:23] <klaxa> sox is a great open source project
[23:23] <klaxa> like what... once?
[23:23] <shevy> klaxa like many times
[23:23] <klaxa> so you can't use it and therefore it's bad?
[23:23] <shevy> ffmpeg + mplayer have things well covered
[23:23] <klaxa> then don't touch ffmpeg
[23:24] <shevy> why
[23:24] <klaxa> don't touch any software whatsoever
[23:24] <shevy> what does ffmpeg have to do with sox?
[23:24] <shevy> ffmpeg is very, very good
[23:24] <klaxa> what does you not being able to use the program have to do with its quality?
[23:24] <shevy> klaxa if something is shit, it's shit
[23:24] <klaxa> you are not providing any evidence
[23:24] <shevy> klaxa I gave you one specific example so far
[23:25] <klaxa> >I used to think it had a great effect, but it sucked!
[23:25] <klaxa> well what did you do to the file?
[23:25] <shevy> yes, the fancy effect
[23:25] <shevy> It was some morph effect, some "fade" related action
[23:25] <klaxa> actually i don't want to discuss this
[23:26] <bffmpeg> http://pastebin.com/hGSZtker  ffmpeg generated NTSC-DVD.mpg file chokes dvdauthor - any insights?    Primary error is WARN: Discontinuity of nnn in audio channel 0; please remultiplex input.
[23:51] <skifreak> having a problem with ffmpeg recognizing the codec for a right bluray, passthrough, video file. mediainfo shows very similar information between the right and left video files, with only the left containing audio. Tried increasing probesize and analyzeduration to no avail. ffmpeg output: http://dpaste.com/1397661/ mediainfo output: http://dpaste.com/1397663/ I have a sample file (10 MB) of each ready on dropbox if needed
[00:00] --- Sat Sep 28 2013