[Ffmpeg-devel-irc] ffmpeg.log.20130314

burek burek021 at gmail.com
Fri Mar 15 02:05:01 CET 2013


[00:00] <saste> dmonjo, yes there is a leak which still needs to be plugged
[00:00] <dmonjo> saste: i am specifying a list size
[00:01] <dmonjo> 9999
[00:01] <dmonjo> but i mean what is the problem with my encoding? is it a bug to be reported?
[00:01] <saste> also try it *locally*
[00:01] <saste> dmonjo, i can't reproduce any issue, as for my tests it is working fine
[00:02] <dmonjo> saste: did you try ffplay http://www2.hipernation.com/wp-content/livefeeds/out4.m3u8   ??
[00:03] <dmonjo> it only plays from 200 to 229
[00:03] <saste> dmonjo, I/O error on my end
[00:04] <saste> that's why i'm telling to test playback on the same machine
[00:04] <dmonjo> saste: what about ffplay http://www2.hipernation.com/wp-content/livefeeds/out5.m3u8
[00:04] <dmonjo> also IO Error?!
[00:05] <saste> dmonjo, starts from 370
[00:05] <dmonjo> yes
[00:05] <dmonjo> it wil end on 399
[00:06] <dmonjo> even though i have 50 ts +
[00:06] <dmonjo> it exited on 399?
[00:11] <dmonjo> saste: if i try to play the file locally on the server i get this error: http://pastebin.com/04tAxBK5
[00:12] <dmonjo> did it exit on your end at 399 please?
[00:12] <saste> dmonjo, the weird thing is that if you look at the m3u8 it starts with segment 40
[00:13] <saste> that's not expected
[00:13] <saste> also on a server ffplay usually won't work (no alsa, no X11), so you should try on your desktop
[00:17] <dmonjo> saste: did it exit on your end at 399?
[00:18] <dmonjo> only played for some 20 sec?
[00:25] <saste> dmonjo, yes
[00:48] <jones_> Hello. Im trying to use ffmpeg to stream my whole desktop, but im having some problems. The thing is ffmpeg captures an image of my desktop with all windows, but doest capture any changes from it. It just tracks mouse movement, nothing else. Ive tried several scripts and variation for streaming, but i use this code line now to test the actual screen capture. 'ffmpeg -f x11grab -s 1920x1080 -r 25 -i :0.0 -sameq ~/screencast.mp4'
[00:49] <burek> jones_, are you trying to capture a movie or something
[00:50] <jones_> burek: no, im trying to capture the whole screen. Like everything that happens on the screen i want to capture.
[00:51] <burek> what exactly are you trying to capture, that is not being captured
[00:51] <fris> jones http://www.commandlinefu.com/commands/view/5230/capture-screen-and-mic-input-using-ffmpeg-and-alsa
[00:51] <burek> fris, why "-itsoffset" ?
[00:53] <klaxa> probably because the system of the person writing that was a bit slow
[00:53] <burek> fris there is a bit more updated tutorial on our wiki: http://ffmpeg.org/trac/ffmpeg/wiki/How%20to%20grab%20the%20desktop%20(screen)%20with%20FFmpeg
[00:53] <jones_> okay so i tried the code line fris linked to. Here is full output http://pastebin.com/y2mi1NXB
[00:54] <burek> jones_, please answer my question
[00:54] <jones_> still the same problem i can see the whole desktop, but it does not track any changes except for mouse movement.
[00:54] <burek> or provide a screenshot
[00:54] <klaxa> (you are also not using ffmpeg but avconv)
[00:56] Action: llogan goes insane
[00:57] <fris> converting mp3 to ogg, is ffmpeg -i file.mp3 -acodec vorbis -aq 60 file.ogg the proper way?
[00:57] <jones_> burek, I want to capture everything on the screen. Like the whole xorg-output you could say, maybe? And when i i play the captures movie, i can see everything, but the video did not record changes. When i moved i windows it didnt show up, when i wrote somthing in terminal it did show up. Only thing that showed up was the movement of the mouse pointer arround the screen.
[00:57] <llogan> fris: no. the native Vorbis encoder is poor compared to libvorbis.
[00:58] <burek> jones_, that's the problem
[00:58] <burek> some media players play the video on screen directly (or using some other direct methods) that avoid xorg's video buffer
[00:58] <burek> so you won't capture anything but the black screen
[00:59] <llogan> fris: Range is 010, where 10 is highest quality. 36 is a good range to try. Default is -qscale:a 3.
[00:59] <fris> qa 10?
[00:59] <llogan> see the sparse guide at https://ffmpeg.org/trac/ffmpeg/wiki/TheoraVorbisEncodingGuide
[00:59] <fris> aq rather
[00:59] <llogan> -q:a/-aq/-qscale:a same thing basically
[00:59] <fris> ffmpeg -i file.mp3 -acodec libvorbis -aq 60 file.ogg
[00:59] <fris> ffmpeg -i file.mp3 -acodec libvorbis -aq 10 file.ogg rather
[00:59] <llogan> yes
[01:01] <jones_> burek, videooutput.  http://youtu.be/dIbZxvLtg90 You can see me moving windows, but only by the mouse pointer moving.
[01:02] <llogan> burek: CamelCase wiki page names are nicer looking.
[01:02] <jones_> i was under the impression the code i tried would give the complete xorg/x11/whatever output, did i misunderstand?
[01:04] <llogan> jones_: the main issue is that we do not support libav products here.
[01:04] <llogan> i recommend using ffmpeg from FFmpeg, but if that is not possible you must go to the correct help channels.
[01:07] <jones_> llogan, not sure if im following, but im really using avconv not ffmpeg, and therefor kinda in the wrong channel?
[01:08] <burek> jones_ yes
[01:08] <burek> this is ffmpeg channel, obviously :)
[01:08] <burek> llogan, is there a quick and easy way of converting current urls to camel case
[01:09] <ubitux> jones_: check this ^
[01:09] <jones_> this is my first encounter with ffmpeg/libvav or what im using, so i really dont know the difference, but yeah i saw it was a difference when i googled libav
[01:10] <ubitux> tl;dr: the ffmpeg you're using is an old broken copy distributed by the fork
[01:10] <ubitux> and avconv is their main tool, trying to be the ffmpeg equivalent
[01:10] <ubitux> we are maintaining the ffmpeg tool, not the copy you are using
[01:11] <llogan> burek: i can rename pages, or make a "this page has moved" but that might be overkill. don't worry about it for existing pages.
[01:11] <llogan> jones_: it doesn't help that they are misrepresenting our work and confusing the users and causing us to spend time explaining instead of helping/coding.
[01:12] <jones_> Aha, i see, well thanks for the help anyways
[01:13] <burek> llogan, it might be better to setup redirects, just like on wikipedia
[01:13] <burek> leaving old urls active also, because people might have already included them in their blogs
[01:13] <jones_> yeah is seems like a dumb idea to still maintain the name ffmpeg in any way..
[01:13] <burek> i'll take a look how is it done and try to convert it
[01:13] <llogan> burek: yeah, keep old urls good. i'll take a look later...but now i'm late for a video recording job
[03:01] <Hiso-android> hello
[03:02] <Hiso-android> anyone here?
[03:02] <klaxa> no
[03:03] <klaxa> just ask your question, don't ask to ask
[03:05] <Hiso-android> ok....i have a problem about ffmpeg's compiling
[03:06] <Hiso-android> about dts codec
[03:08] <Hiso-android> do i need external lib to support that?
[03:10] <Hiso-android> 
[03:10] <klaxa> not sure, can you use a pastebin like site to post some logs?
[03:13] <Hiso-android> sorry,what's your mean?im not good at English...
[04:36] <Mista_D> Hiso-android: encoder or decoder?
[10:57] <t4nk016> is it possible to generate stream if webcam support mpegts format by ffserver + ffmpeg?
[10:59] <durandal_1707> webcam records in mpegts?
[11:01] <durandal_1707> t4nk016: ^
[11:03] <durandal_1707> mpegts can be read/demuxer, but you did not specified what output stream format to use
[11:04] <durandal_1707> and i don't think one can stream with avi
[11:31] <t4nk016> yes, because when I plug webcam , it says format mpeg2 ts found
[11:32] <Mavrik> t4nk016, you can re-stream mpegts yes
[11:32] <Mavrik> you'll have to remux if you want to do it with ffmpeg
[11:32] <Mavrik> but MPEG-TS is made for streaming so use that if you can :)
[11:32] <t4nk016> how to re-mux?
[11:33] <t4nk016> re-mux to .ts stream, right?
[11:33] <t4nk016> if it is true, i do not know how to set in ffserver.conf
[11:34] <t4nk016> an another question
[11:35] <t4nk016> when I try to generate a .avi from the mjpeg webcam
[11:35] <dv_> hello. I want to stress test a hardware h264 decoder chip, to see how hot it gets. I want to generate test videos for that. is a white, tv-like noise a good pick ? to me, it seems as if the unpredictability of the white noise would maximize the number of computations needed, compared to something like a bouncing white ball against a purely black background, for example.
[11:35] <t4nk016> I got the error Too large number of skipped frames 817969319958 > 6000 Error writing frame to output
[11:36] <t4nk016> in ffserver.conf, format is avi, VideoCodec is mjpeg
[11:36] <t4nk016> and ffmpeg command line is :ffmpeg -loglevel debug -f video4linux2 -input_format mjpeg -r 15 -s 640x480 -i /dev/video0 -vcodec copy http://localhost:8090/feed1.ffm
[11:36] <t4nk016> is there any idea?
[11:45] Last message repeated 1 time(s).
[11:45] <durandal_1707> replace avi with something better
[11:48] <t4nk016> which one is better? asf?
[11:49] <durandal_1707> for mjpeg only you can use just raw mjpeg
[11:50] <t4nk016> yes, that case I have test
[11:50] <t4nk016> it is workable
[11:50] <t4nk016> And I would like to try another format with mjpeg source
[11:51] <t4nk016> and in advance , I will do the test with a new webcam,
[11:51] <t4nk016> the new web cam does not support mjpeg
[11:52] <t4nk016> I would like to do the verification if the format is changed
[14:03] <bk76_> hello
[14:03] <bk76_> I recorded my desktop with ffmpeg -f x11grab -r 10 -s 1920x1080 -i :0.0 -f alsa -i hw:0,2 -acodec pcm_s16le -vcodec libx264 -crf 0 -preset ultrafast -y recorddesktop.mkv
[14:04] <bk76_> but when uploaded to youtube there is only sound
[14:05] <bk76_> how can I convert this video to be supported by youtube?
[14:06] <sacarasc> YouTube will pretty much take anything.
[14:06] <sacarasc> Maybe it doesn't like lossless H264, though...
[14:07] <bk76_> it looks like it doesn't
[14:07] <JEEB> nah
[14:08] <JEEB> it should support lossless H.264
[14:08] <sacarasc> Try ffmpeg -i recorddesktop.mkv -c:a flac -c:v libx264 -crf 20 -preset slow output.mkv
[14:08] <JEEB> bk76_, what colorspace is your encoded output?
[14:08] <bk76_> so what is wrong?
[14:08] <JEEB> ffmpeg -i recorddesktop.mkv
[14:08] <JEEB> pastebin that
[14:08] <JEEB> sacarasc, I'm pretty sure what youtube is failing at and if his pastebin confirms it I shall say it :)
[14:09] <JEEB> and it is not failing at lossless, as that has been supported from '09-'10 or so
[14:09] <sacarasc> Okay.
[14:09] <bk76_> http://pastebin.com/Q6RhQ1uC
[14:09] <JEEB> yuv444p
[14:09] <JEEB> yup
[14:10] <JEEB> exactly as I thought :)
[14:10] <bk76_> ?
[14:10] <JEEB> you encoded in 4:4:4 YCbCr
[14:10] <JEEB> the 'tube H.264 decoder doesn't support that
[14:10] <bk76_> I always encode like that
[14:10] <bk76_> I think
[14:10] <JEEB> nope, if 'tube took it then you haven't
[14:11] <bk76_> or maybe new version of ffmpeg chcnge default behavior
[14:11] <JEEB> do note that High 4:4:4 Predictive profile contains lossless for all colorspaces
[14:11] <JEEB> so that has nothing to do with it
[14:11] <JEEB> basically ffmpeg -i recorddesktop.mkv -c:a copy -c:v libx264 -crf 0 -preset ultrafast -pix_fmt yuv420p out.mkv
[14:11] <JEEB> this should fix it :)
[14:11] <bk76_> so what I can do with this video now?
[14:11] <bk76_> ok
[14:12] <JEEB> add -pix_fmt yuv420p to your recording lines in the future :)
[14:12] <JEEB> the encode will also be faster in this case, as there is less data
[14:12] <JEEB> sacarasc, tl;dr 'tube hasn't updated their libavcodec to late 2011 :)
[14:12] <sacarasc> Heh.
[14:12] <bk76_> JEEB: great :D
[14:13] <JEEB> there actually /was/ a long point of time after x264 switched to the new lossless profile when 'tube didn't take it in
[14:13] <JEEB> but this was fixed somewhere in between of '09 to '11
[14:43] <bk76_> JEEB: it is working with youtube now
[14:43] <bk76_> JEEB: thank you
[14:43] <JEEB> yeah, it just uses an old libavcodec decoder so 4:2:2 and 4:4:4 H.264 won't work
[14:43] <JEEB> no problem
[14:46] <Guest31608> hello guys, any idea if/how I can overlay image to video with fade in /fade out to the image only? best case scenario the image fades in/out a few times throughout the duration of the video..
[14:53] <marmin> hello
[14:54] <marmin> I have a question using ffmpeg command
[14:54] <marmin> on windows
[14:54] <marmin> Is there anybody to help me?
[14:56] <Mavrik> well you COULD just ask what your problem is.
[14:56] <marmin> Ok
[14:57] <marmin> I want to encode a raw video sequence  by MPEG4-part2-SP
[14:57] <marmin> I want to turn the option of halfpel motion estimation on and off
[14:58] <marmin> In fact I want to obtain two different encoded sequences, one with halfpel one and the other with halfpel off
[14:58] <marmin> how should I map this option into ffmpeg command?
[15:02] <marmin> ???
[15:02] <marmin> any idea?
[15:07] <Mavrik> well
[15:07] <Mavrik> you can always look at source of mpeg4 decoder and see which parameters are suppoorted
[15:07] <Mavrik> to get multiple output you just pass multiple sets of output parameters one after another
[15:07] <marmin> I know they are supported
[15:08] <marmin> but I dont know how to tell ffmpeg to do that?
[15:08] <dmonjo> hello community
[15:08] <dmonjo> what are the recommended hardware specs of a system to convert an ogv live stream into an m3u8 using HLS ?
[15:08] <dmonjo> what is the thing to worry about the most? CPU and memory ?
[15:08] <dmonjo> bitrate is 128K
[15:09] <marmin> mavrik?
[15:12] <Mavrik> marmin, no idea, read documentation
[15:12] <Mavrik> scan of source doesn't really show support for that option in the encoder
[15:12] <Mavrik> but I might be wrong
[15:15] <marmin> Ok,
[15:15] <marmin> Thank you
[15:15] <Guest31608> hello guys, any idea if/how I can overlay image to video with fade in /fade out to the image only? best case scenario the image fades in/out a few times throughout the duration of the video.
[15:55] <dmonjo> where can i downlaoderd the latest libx264-dev?
[15:59] <dmonjo>  ia this versiopn considered decent? libx264-dev_0.123.2189+git35cf912-1_amd64.deb
[16:00] <JEEB> it's not ancient, there have been speed-wise updates and bugfixes since, but it's not too old. Personally I would just compile libx264 myself, I guess?
[16:04] <dmonjo> JEEB: where is the GIT?
[16:05] <dmonjo> git clone git://git.videolan.org/x264.git ?
[16:05] <JEEB> yup
[16:05] <dmonjo> its not libx264-dev
[16:05] <JEEB> yes, if you build either the static or shared library it is /exactly/ what is in libx264-dev
[16:05] <dmonjo> so i creaqte a symbolic link?
[16:05] <JEEB> by default it doesn't enable separate library compilation/installation
[16:05] <JEEB> --enable-static or --enable-shared
[16:06] <JEEB> is what you need :P
[16:06] <dmonjo> cpz ffmpeg complains even if x264 is enabled
[16:06] <dmonjo> i mean installed
[16:06] <JEEB> then you are not doing something right, see config.log for details
[16:28] <IchGuckLive> hi imtrying to convert a movie to my aiptek using ffmpeg insted of windows but it fails on stripes and pixels  at  Y40+++ upper part is ok
[16:28] <IchGuckLive> the  output of a workking moviue is as follor -> http://pastebin.com/KFW5CVrG
[16:29] <IchGuckLive> i used commandline -> ffmpeg -i in.ts -vcodec mpeg4 -b 2000k -s 640x480 -r 25 -acodec aac -strict -2 -ab 128k out.asf
[16:30] <IchGuckLive> no error it plays at mplayer and VLC but in projektor only the upper part is ok the windoes arcsoft AMC works on this
[17:04] <memand> HeyI'm trying to transcode an .m4a file to .wav with ffmpeg and I have just installed an encoder that should be able to do it (libfdk-aac) but ffmpeg does not recognize the new encoder... What can I do?
[17:06] <memand> s/HeyI'm/Hey I'm/
[17:07] <JEEB> memand, ffmpeg does not have a system to find newly installed encoders on-the-fly, you will have to rebuild ffmpeg as well
[17:08] <memand> JEEB: Oh...
[17:08] <memand> Thanks :)
[17:09] <memand> JEEB: So optimally I should uninstall ffmpeg and install all important codecs and then reinstall/recompile ffmpeg?
[17:10] <memand> by codecs I mean encodersø
[17:10] <memand> *encoders
[17:12] <JEEB> what o_O
[17:14] <memand> lol, nvm :) thanks for the help
[19:45] <IchGuckLive> hi how can i influence the tbr tbn tbc on a libx264
[19:48] <IchGuckLive> and is there a preset main ?
[19:49] <IchGuckLive> for libx264
[19:50] <JEEB> that is not a preset, that is a profile
[19:50] <JEEB> -profile:v main
[19:50] <IchGuckLive> thanks
[19:51] <JEEB> as in, baseline/main/high are profiles
[19:51] <JEEB> of H.264
[19:51] <JEEB> in other words, they control which features the encoder can use
[19:51] <JEEB> presets are general speed vs compression settings
[19:52] <IchGuckLive> i try now for hores to get  the vidio in ubuntu  as arcsoft does in XP
[19:53] <IchGuckLive> XP does -> Stream #0:0(eng): Video: h264 (Main) (avc1 / 0x31637661), yuv420p, 640x480, 2576 kb/s, 30 fps, 30 tbr, 90k tbn, 180k tbc
[19:53] <IchGuckLive> ubuntu Stream #0:0(eng): Video: h264 (Main) (avc1 / 0x31637661), yuv420p, 640x480 [SAR 4:3 DAR 16:9], 2496 kb/s, 30 fps, 30 tbr, 30 tbn, 60 tbc
[20:00] <IchGuckLive> so tbn tbr tbc are no options
[20:10] <dmonjo_> how reliable is running ffmpeg on a virtual instance
[20:10] <dmonjo_> does it require lot of resources?
[20:14] <burek> dmonjo_, test it and see for yourself.. the question is a little bit too general, dont you think
[20:20] <IchGuckLive> no way for 4hr now to gt the movi seen in the projektor
[20:20] <IchGuckLive> it try tomorrow its late in germany
[20:24] <dmonjo_> is the hls mux provided by ffmpeg the same as the one of libconv?
[20:26] <memand> I need to convert an .m4a audio file to .wav I have been trying in a buch of different ways with a bunch of different encoders now, but so far I have only succeded to fail and google has not been much help... A nudge in the right direction would be much appreciated :)
[20:28] <LithosLaptop> pretty easy with ffmpeg
[20:28] <burek> memand, can you provide a sample of what you tried
[20:28] <burek> should be as easy as: ffmpeg -i input.m4a output.wav
[20:31] <LithosLaptop> or try: ffmpeg -i input.m4a -c:a pcm_s16le output.wav
[20:31] <LithosLaptop> ,but ffmpeg is probably smart enough when it sees a .wav output file in the command line
[20:36] <memand> burek: Seriously? I tried stuff like 'ffmpeg -i input.m4a -c:a pcm_s16le output.wav' , 'ffmpeg -i input.m4a -c:a libfdk_aac -vbr 3 output.wav' and other crazy stuff...
[20:37] <burek> oh man... why?
[20:37] <LithosLaptop> I don't think aac can fit in a WAV file, lol
[20:37] <memand> burek: Why ideeed... :/
[20:38] <dmonjo_> burek: the reason i am asking this question is because when i am running ffmpeg on a vritual instance in the cloud even though it is 4GB of RAM and 8 CPUs the conversion is not good for the m3u8 file, when i do it locally it works fine
[20:39] <dmonjo_> this is why i am asking if anyone has any good experience for a virtual instance "in the cloud"
[20:39] <burek> dmonjo_, do some profiling with the cloud tools provided
[20:39] <memand> burek: But I just tried what you said and the program that needs the wav is still complainin, so I'm starting to think that WAV and WAVE is not the same thing? (I'm not the biggest genious when it comes to encoding)
[20:39] <dmonjo_> virtual instance locally works fine
[20:39] <dmonjo_> burek: any idea if the hls mux of ffmpeg is the same as the one of libav?
[20:40] <LithosLaptop> can you playback the output .wav file with other players like VLC?
[20:40] <memand> burek: ?
[20:40] <memand> LithosLaptop: Yes
[20:40] <burek> dmonjo_, no idea, i dont use libav, ffmpeg was good enough for all the tasks i had so far..
[20:41] <memand> burek: You want the console output from the transcoding?
[20:41] <LithosLaptop> then that program probably needs something else
[20:41] <burek> memand, well yes, but I can already figure that everything went ok
[20:41] <burek> i just wanted to see if there were some obvious things/errors inside
[20:43] <dmonjo_> is there a need to have akso audio driver loaded on the ffmpeg server?
[20:43] <memand> burek: Yeah everything seemed to go smooth :) I'm starting to think that the problem is with the other program (if a WAVE file is indeed the same as .wav)
[20:43] <burek> dmonjo_ for what exactly?
[20:43] <dmonjo_> to convert audio from ogv to ts
[20:43] <burek> memand, usually it is
[20:43] <memand> :P
[20:43] <burek> wave is generally refered to .wav container afaik
[20:44] <memand> Yeah that is also my understanding
[20:44] <dmonjo_> burek: what about virtual video display it is set to 16 mb on my vritual instance , do the encoding process uses the virtual display or virtual sound card?
[20:44] <burek> but what wav container can contain inside, doesn't have to be supported by your program
[20:44] <LithosLaptop> maybe it needs different PCM format in the WAV container
[20:44] <burek> dmonjo_, ogv to ts?
[20:44] <dmonjo_> ogv to m3u8
[20:44] <burek> :D
[20:44] <dmonjo_> mpeg
[20:44] <burek> video stream to a text file?
[20:45] <burek> oh
[20:45] <dmonjo_> :)
[20:45] <dmonjo_> does this depend on the virtual devices (video/audio) ?
[20:45] <burek> dmonjo_, why would you need an "audio driver loaded on the ffmpeg server" in order to convert ogv (video) to mpeg (video) ?
[20:46] <memand> burek: Ok but as far as I have understood .wav is the "pure" wave with no compression or anything?
[20:46] <craq> anyone here using ffmpeg in an ios app?
[20:46] <burek> memand, what exactly have you read about "wave" requirement? can you give us the url to check?
[20:47] <memand> The requirement for the program?
[20:47] <burek> i guess
[20:48] <memand> I could give you a link (http://opensmile.sourceforge.net/) but it's really poorly documented. It's just something I have been screening to use on a project, but I think it will be easyer to trash it and find something else :)
[20:48] <LithosLaptop> WAV files can be any of these PCM, ADPCM, Microsoft GSM 06.10, CELP, SBC, Truespeech and MPEG Layer-3.
[20:48] <LithosLaptop> yes...you can include a MP3 stream in a WAV file.. lol
[20:49] <burek> yes, also mp3 can be put into .wav container
[20:49] <burek> i used such files with winamp long time ago
[20:50] <burek> memand, it would be the best to find the proper support channel for that software and to ask those developers about your issue
[20:51] <victusfate> hello, I have a question with regards to applying the setpts filter with mpegts files. I'm not seeing the frame time stamps I'd expect on the output file
[20:51] <memand> but if I do 'ffmpeg -i input.m4a output.wav' does it then "decode" the .m4a and make a .wav or does it pack the .m4a inside the .wav?
[20:51] <burek> probably your issue is there, unless you are using some really old version of ffmpeg which might contain some bugs or so
[20:51] <memand> I'm not :)
[20:51] <burek> m4a is a container too :)
[20:51] <victusfate> http://pastebin.com/MRsxmxD2
[20:51] <burek> that was fast :)
[20:51] <victusfate> will paste the ffrpobe output as well
[20:52] <burek> but its missing the output of course
[20:52] <relaxed> memand: that will decode to pcm
[20:52] <victusfate> -loglevel debug?
[20:52] <burek> victusfate, your pastebin is missing any kind of output :)
[20:52] <victusfate> just a moment
[20:52] <LithosLaptop> memand: it takes the lossy AAC stream inside the M4A container and decompresses it to 16 bit LPCM samples put into a WAV container
[20:52] <victusfate> will update
[20:53] <memand> Ok, that makes sense relaxed and LithosLaptop :)
[20:53] <burek> memand yes, it will create uncompressed wav audio if that's what you ask :)
[20:54] <memand> Can ffmpeg output how a soundfile is compressed?
[20:54] <burek> if you want just to change the container and not touch the audio stream inside, use: ffmpeg -i bla.m4a -c:a copy output.wav
[20:54] <victusfate> here we go , http://pastebin.com/e5dY5HuZ
[20:54] <burek> but that probably won't work, since your m4a audio stream will probably not be valid for wav container
[20:54] <victusfate> the setpts time filter run, and the following ffprobe to show frames
[20:55] <memand> The program came with some demo (.wav) files that it takes gladly, so I'm thinking that I should maybe take a look at them and see how they are compressed/packaged
[20:56] <burek> victusfate "I'm not seeing the frame time stamps I'd expect on the output file" how exactly is this true
[20:56] <burek> for example, random line: [Parsed_setpts_0 @ 0x7ff1d24118e0] N:40 PTS:300000 T:3.333333 POS:57716 INTERLACED:0 -> PTS:300120 T:3.334667
[20:56] <LithosLaptop> memand: good idea
[20:56] <burek> original pts was 300000
[20:57] <burek> and outputed one is 300120
[20:57] <burek> which IS 120+PTS right?
[20:57] <victusfate> the debug output looks correct
[20:57] <victusfate> but the ffprobe doesn't agree
[20:57] <victusfate> 1 sec
[20:57] <victusfate> let me use the proper PTS offset
[20:57] <burek> memand, type ffmpeg -i demo.wav
[20:57] <memand> LithosLaptop: Can ffmpeg give this info, or should I find another tool for that?
[20:57] <victusfate> not sure why but for mpegts files I have to use a much larger PTS
[20:57] <burek> and provide us with the pastebin
[20:57] <victusfate> (as opposed to fps * offset time)
[20:57] <memand> lol thanks burek :)
[20:57] <victusfate> will do
[20:59] <memand> audio: pcm_s16le
[21:00] <victusfate> updated shift of 900000+PTS
[21:00] <victusfate> http://pastebin.com/vYePBCSH
[21:00] <victusfate> debug output shows it dead on at 10sec
[21:00] <victusfate> but ffprobe shows values from 1.4sec to 21.483333 sec
[21:01] <memand> And the files I have been making has the same audio:
[21:01] <victusfate> would like to see 10-20.083333
[21:01] <memand> Thats it, I'll find a program that works instead of that openSMILE trash :)
[21:01] <memand> Thanks for all the help guys :D
[21:01] <LithosLaptop> memand: samplerate and channels?
[21:03] <dmonjo_> i am compiling  with --enable-gpl and --enable-libx264, is there anything i can add to make it better performance for encoding OGV stream to MPEG?
[21:03] <memand> LithosLaptop: The outputs where identical with the exeption that the one I made has metadata (and the lenght of course)
[21:03] <relaxed> dmonjo_: make sure you have a recent version of yasm instaled.
[21:03] <dmonjo_> relaxed: the latest
[21:03] <dmonjo_> GIT
[21:04] <dmonjo_> latest x264 and latest ffmpeg
[21:04] <dmonjo_> and latest yasm :)
[21:04] <dmonjo_> and still my problem is there
[21:04] <dmonjo_> hehe...
[21:06] <craq> anyone here using ffmpeg in an ios app? curious about performance. cant find much via google about performance vs what apple provides.
[21:07] <memand> craq: You want to make an app?
[21:07] <craq> i have an app already. heavy video processing using what apple provides but reached a bug in their api (bug reports filed) and im looking for an alternative to writing video files vs apple's AVAssetWriter object.
[21:10] <dmonjo_> what is better than aac strict experimental?
[21:11] <memand> craq: I'm very versed in the way of apple (lets just say that I'm not a fan) but if you are developing on my guess would be that you could do some kind of benchmarking trough some debugging function?
[21:11] <relaxed> dmonjo_: build ffmpeg with libfdk_aac support
[21:12] <LithosLaptop> dmonjo_: libfdk_aac
[21:12] <dmonjo_> relaxed: trying ac3
[21:12] <dmonjo_> good i think
[21:13] <craq> definitely memand but was looking to see if someone has already done it and reported on it. i cant find anything. also was hoping to find this out before i invested the time looking at it more..
[21:14] <craq> the other factor i'd like about using ffmpeg is having more control. I'm finding AVAssetWriter leak internal objects (abandoned memory) and the nature of my app is to always be on creating multiple recordings, so over time if a lot was made, you can eventually run out of memory and i cant control that.
[21:14] <craq> sucks.
[21:14] <relaxed> craq: I assume there are many apps that use ffmpeg's libs, why not track them down and look at their source.
[21:14] <memand> AVAssetwriter is apples transcoding util?
[21:15] <relaxed> ffmpeg is well supported on arm.
[21:15] <craq> yes memand, you use it to create recordings that you capture from the camrea.
[21:16] <craq> and yes, i know its supported. i've read many peoples tutorials on getting it setup but no one talks about performance.
[21:16] <memand> I think you have some research to do for the world craq ;)
[21:17] <craq> apparently
[21:17] <craq> hah
[21:17] <craq> i have 3 shitty bugs that i found in Apple's AVFoundation library (all the video / audio stuff).
[21:17] <craq> i gave great reports to them with a sample app too
[21:18] <craq> i got 1 useless response to 1 of the reports.
[21:18] <memand> Not to bash on apple (ok maybe a little :P) but my guess would be that most people would not even find those bugs because they blindly trust that apple knows their shit
[21:18] <craq> ya. also no one is pushing the limits of that library so it seems. no one does continuous recording like this. at least nothing that i have found yet and i've done a lot of searching.
[21:19] <craq> we do motion and face detection too. and with motion based recordings, we have a featured we call the preroll where we save sample buffers as they are captured from the camera (up to 2 seconds worth), and when we start a motion recording, we inject those in first so you can fully see where a motion based event started.
[21:20] <memand> Sounds like you are gonna push the limits of what you can squeeze out of that chip :D
[21:20] <memand> Sir I salute you !
[21:21] <craq> hahah
[21:21] <craq> :)
[21:21] <craq> its fun stuff but this is all still fairly new to me.
[21:22] <craq> i've rewritten how i handle recording like 10 times already lol
[21:22] <memand> haha
[21:22] <JEEB> <craq> i got 1 useless response to 1 of the reports. <- better than zero of those I know of that were reported to the QT team
[21:22] <craq> hahaha ya.
[21:22] <craq> true.
[21:22] <Mavrik> wow, corpos actually answer to bug reports now?
[21:23] <Mavrik> sheesh
[21:23] <memand> But maybe we should take this conversation elsewhere since we are getting a bit off the channel toppic craq ;)
[21:23] <craq> all good.
[21:31] <dmonjo_> how come i play ffmpeg -f lavfi -i testsrc for 5 sec
[21:31] <dmonjo_> but avplay keeps playing for unlimieted time?!?
[21:31] <dmonjo_> should it play only the recoreded 5 seC?
[21:34] <victusfate> heyo, never heard back about the delta time offset I was seeing in mpegts output files. Basically can't shift the time stamps without pushing them further out and introducing earlier empty frames, output example: http://pastebin.com/vYePBCSH
[21:35] <victusfate> sorry if I missed a reply, fighting off a fever today
[21:45] <dmonjo_> is there any good parameter to provide at compile time to make encoding performance better?
[21:45] <klaxa> -O3
[21:45] <dmonjo_> klaxa: is that for me?
[21:46] <dmonjo_> ./configure -03 ?
[21:46] <klaxa> yeah that's about all I know
[21:46] <dmonjo_> what does that do exactly?
[21:46] <klaxa> it's an uppercase O and you'd have to specify it as an extra CFLAG
[21:46] <klaxa> at least in gcc it means optimization-level 3
[21:46] <klaxa> but i doubt that it will help much
[21:46] <klaxa> there is no magical haxx0r tricks to make encoding faster
[21:47] <llogan> dmonjo_: avplay is not a FFmpeg product
[21:47] <dmonjo_> ffplay..
[21:48] <llogan> add -autoexit
[21:49] <dmonjo_> llogan: but i want to understand the way -f lavfi -i testsrc works
[21:49] <dmonjo_> i encode for 5 sec
[21:49] <dmonjo_> it plays infinitiely
[21:50] <llogan> or ffplay in your case
[22:02] <craq> so for the record memand, i've come to the conclusion that ffmpeg sadly wont be fast enough for my needs :( damnit. found a developer on github that did a few things with iOS and ffmpeg. reached out to him and got a super fast response. i guess im movin on.
[22:05] <memand> craq: That's too bad, and I really have so little experience with transcoders that I can't point you anywhere... But if AVwhateverthenamewas is fast enough, then why not do some "manual" memory mannagement (if that is possible in obejctive c)?
[22:06] <craq> its internal objects so thats why im stuck. oh well. just thought i'd let you know. i guess this all just my big challenge for the next while lol. i was also about to go and add more notes to my apple bug reports but now there bug tracker is broken lol. fail..
[22:07] <memand> Send a bug report on the bug tracker?
[22:07] <memand> :P
[22:14] <dmonjo_> -c:a libfdk_aac ?
[22:15] <dmonjo_> how can i use that library with ffmpeg?
[22:17] <llogan> dmonjo_: https://ffmpeg.org/trac/ffmpeg/wiki/AACEncodingGuide#fdk_aac
[22:19] <dmonjo_> llogan: error while loading shared libraries: libfdk-aac.so.0: cannot open shared object file: No such file or directory
[22:19] <dmonjo_> even though i installed the library
[22:20] <dmonjo_>  libfdk-aac 0.1.1-2
[22:20] <saste> dmonjo_, fixed the m3u8 riddle?
[22:20] <dmonjo_> saste: installed a local server on a local vm it works fine
[22:20] <dmonjo_> it looks like i have a problem with the vm loaded in the cloud
[22:20] <dmonjo_> resource thing :/
[22:20] <dmonjo_> i dunno
[22:20] <saste> uhmm.... ok
[22:21] <dmonjo_> coz its not a bandwidfth issue since i am trying to play from my pc into the cloud and it doesnt work
[22:21] <saste> that's why i asked you to test locally, with a network you never know
[22:21] <dmonjo_> i am still working on it now
[22:21] <dmonjo_> will update you
[22:21] <dmonjo_> thanks for following up :)
[22:22] <saste> thanks for the update, it's always interesting to know how a problem was fixed
[22:23] <dmonjo_> i am trying to load now the  libfdk_aac and getting this error :  error while loading shared libraries: libfdk-aac.so.0: cannot open shared object file: No such file or directory
[22:23] <dmonjo_> using  -c:a libfdk_aac :/
[22:24] <saste> dmonjo_, how did you install ffmpeg? from source or binary?
[22:25] <dmonjo_> source GIT
[22:25] <saste> also looking for LD_PATH may help in case you're using a custom path
[22:25] <saste> LD_LIBRARY_PATH
[22:26] <dmonjo_> ffmpeg version N-50937-g870e625
[22:26] <dmonjo_> i installed libfdk also from source
[22:26] <relaxed> dmonjo_: try, LD_LIBRARY_PATH=/usr/local/lib ffmpeg
[22:27] <saste> dmonjo_, the problem is *where* did you install
[22:27] <dmonjo_> make install should do it no?
[22:27] <saste> dmonjo_, check the paths, trust no one
[22:27] <relaxed> the default prefix when compiling software is /usr/local, fyi
[22:29] <dmonjo_> if i make install should it expoert it and set all these?
[22:29] <dmonjo_> ok i have libfdk.so in /usr/local/lib
[22:30] <saste> dmonjo_, force the LD_LIBRARY_PATH and see if it works
[22:30] <dmonjo_>  echo $LD_LIBRARY_PATH /usr/local/lib
[22:30] <dmonjo_> and still it doesnt work
[22:31] <saste> as a last resort you can rely on strace ffmpeg ... to read which paths it's reading
[22:31] <relaxed> RUN THIS --> "LD_LIBRARY_PATH=/usr/local/lib ffmpeg"
[22:31] <saste> echo $LD_LIBRARY_PATH /usr/local/lib <- what's this useful for?
[22:31] <dmonjo_> i assigned that variable
[22:31] <dmonjo_> ok
[22:32] <relaxed> dmonjo_: distro?
[22:32] <dmonjo_> debian 6
[22:33] <relaxed> echo "/usr/local/lib" >> /etc/ld.so.conf.d/ffmpeg.conf && sudo ldconfig
[22:33] <dmonjo_> col it works...
[22:35] <relaxed> but sometimes you don't want the libs global, but whatever.
[22:52] <dmonjo_> can someone try to play ffplay http://www3.hipernation.com/wp-content/livefeeds/qaz.m3u8 and tell me for how many seconds the stream plays nonstop?
[22:54] <dmonjo_> that will help me troubleshoot my problem
[22:54] <ubitux> seems to work
[22:55] <dmonjo_> for how many seconds plase?
[22:55] <ubitux> about 1 minute now
[22:55] <dmonjo_> is it cutting after 25 sec?
[22:55] <klaxa> timer is at 350 seconds
[22:55] <dmonjo_> which timer?
[22:55] <klaxa> ffplays timer
[22:55] <ubitux> dmonjo_: i got some little cut sometimes
[22:55] <klaxa> but it didn't start at 0
[22:55] <ubitux> but they don't last much
[22:55] <klaxa> yeah the audio kinda stops
[22:55] <ubitux> (and i believe it's due to bandwidth or something)
[22:56] <dmonjo_> hmmm is that related to encoding?
[22:56] <dmonjo_> the server is in a cloud
[22:56] <dmonjo_> liquidweb
[22:56] <dmonjo_> good stuff
[22:56] <dmonjo_> shouldnt be bandwidth
[22:56] <dmonjo_> you see video and audio?
[22:56] <ubitux> both
[22:56] <klaxa> yeah been running for over 100 seconds now
[22:57] <ubitux> the buffer is very small
[22:57] <ubitux> something like < 10KB for a & b
[22:57] <ubitux> a & v
[22:57] Action: ubitux stopping now
[22:58] <klaxa> well... with mplayer2 i get new chunks all the time but neither video nor audio
[22:58] <klaxa> or rather i get a green image and no audio might be mplayer related
[23:00] <dmonjo_> so is it an encoding problem or a client or a bandwdith?
[23:01] <klaxa> the mplayer thing is probably a client problem since ffplay plays it back fine
[23:02] <dmonjo_> do you guys see an encoding problem?
[23:02] <dmonjo_> the cuts...
[23:02] <dmonjo_> i will try a second time if possible
[23:02] <dmonjo_> on a new stream
[23:03] <dmonjo_> ffplay http://www3.hipernation.com/wp-content/livefeeds/wsx.m3u8
[23:03] <dmonjo_> please let me know if you see some cuts
[23:04] <dmonjo_> or video/audio problems..
[23:07] <dmonjo_> hello?
[23:09] <dmonjo_> klaxa: ubitux  ?
[23:09] <klaxa> doesn't even start
[23:10] <ubitux> same here
[23:10] <dmonjo_> hmm why :/
[23:11] <dmonjo_> ffplay http://www3.hipernation.com/wp-content/livefeeds/wsx.m3u8
[23:11] <dmonjo_> now plz
[23:11] <dmonjo_> looks like a bandwidth problem from the webcam to the server
[23:12] <dmonjo_> gst-launch-0.10 v4l2src ! videoscale ! video/x-raw-yuv,width=320,height=240 ! queue ! ffmpegcolorspace ! videorate ! theoraenc bitrate=64 ! queue ! oggmux name=mux pulsesrc ! audio/x-raw-int,rate=44100,channels=1,depth=16 ! queue ! audioconvert ! audiorate tolerance=40000000 ! vorbisenc ! queue ! mux. mux. ! queue ! shout2send
[23:13] <dmonjo_> do you think i can lower more the resolution :/ bitrate? optimize it?
[23:13] <dmonjo_> but the thing is
[23:13] <dmonjo_> it plays well if you point to the live stream
[23:13] <dmonjo_> so something is happening in conversion
[23:13] <dmonjo_> when i point to the stream http://wwww....stream.ogv it plays well but the m3u8 is slow
[23:14] <dmonjo_> so what could be the problem?
[23:21] <dmonjo_> saste you here?
[23:30] <fatpony> there's a typo in that warning http://bit.ly/16vwaqv "extreemly"
[23:31] <fatpony> i don't know who i should highlight though...
[23:31] <fatpony> oh there's a dev channel
[23:32] <saste> fatpony, that is already fixed in master
[23:32] <fatpony> really? ithought it was running a git version...
[23:33] <saste> fatpony: git show dae76e8c4730f60bb46ac3ab30f5231b50b5fa56
[23:34] <dmonjo_> saste: ffplay http://www3.hipernation.com/wp-content/livefeeds/zxc.m3u8
[23:34] <dmonjo_> can you please check
[23:34] <dmonjo_> if the stream is playing nonstop?
[23:35] <dmonjo_> it is cutting from my end every 10sec
[23:35] <dmonjo_> want to see another point of view
[23:35] <dmonjo_> appreciate if anyone can check form his side
[23:35] <saste> dmonjo_, i get some freeze, possibly when switching chunk
[23:36] <dmonjo_> but it continues?
[23:36] <saste> that may be related to download delay
[23:36] <dmonjo_> saste: you mean bandwidth?
[23:36] <saste> also i suspect the hls demuxer/protocol is not doing pro-active download
[23:36] <saste> vlc may be better from that POV
[23:37] <dmonjo_> saste: what can i do to the hls demuxer?
[23:38] <saste> dmonjo_, probably nothing, unless you can code
[23:38] <dmonjo_> saste: i get this error from vlc when playing the stream http://pastie.org/6501552
[23:38] <dmonjo_> interresting
[23:39] <saste> i don't know how is the implementation of the hls demuxer, so can't say
[23:39] <dmonjo_> saste: but why is it working locally but over the internet
[23:39] <dmonjo_> not over the internet
[23:39] <saste> dmonjo_, because the internet is complex beast
[23:40] <saste> and data takes time to travel
[23:40] <saste> much easier to read from your hd
[23:40] <dmonjo_> i need to talk to an HLS expert
[23:40] <dmonjo_> i am sure lot of people are suffering from that
[23:40] <SubJunk> I'm looking to convert video to mp4 that can be played in web browsers using HTML5's video tag, does anyone have any tips?
[23:42] <saste> dmonjo_, did you try with an i* thing?
[23:42] <dmonjo_> what is i thing?
[23:43] <saste> dmonjo_, ipad, iphone, ...
[23:43] <dmonjo_> yea that applies to them
[23:44] <dmonjo_> i think the HLS needs to be be injects some good options that are magic to me
[23:44] <dmonjo_> i am using this ffmpeg -v debug -i http://127.0.0.1:8000:/event1.ogv -vcodec libx264 -acodec libfdk_aac -b:v 100k  -flags -global_header -map 0:0 -map 0:1 -f hls -hls_time 20 -hls_list_size 999  /var/www/hipernation.com/wp-content/livefeeds/zxc.m3u8
[23:45] <saste> dmonjo_, but the thing is that it's probably a net thing
[23:46] <dmonjo_> i am using a tunnel
[23:46] <dmonjo_> but i dont think it has to do with it
[23:46] <dmonjo_> encrypted tunnel
[23:50] <dmonjo_> saste: i dont think its a problem with the net
[23:50] <dmonjo_> since i can play my ogv over the net without cuts
[23:50] <saste> dmonjo_, how?
[23:51] <saste> dmonjo_, HTTP != HLS
[23:52] <dmonjo_> exactly
[23:52] <saste> with HTTP you download all the file, then you play it
[23:52] <llogan> SubJunk: ffmpeg -i input -codec:v libx264 -preset medium -crf 23 -movflags faststart -codec:a libfdk_aac -vbr 3 output.mp4
[23:52] <saste> with HLS you download every single chunk, while you're still playing
[23:52] <llogan> SubJunk: https://ffmpeg.org/trac/ffmpeg/wiki/x264EncodingGuide
[23:52] <llogan> SubJunk: https://ffmpeg.org/trac/ffmpeg/wiki/AACEncodingGuide
[23:52] <saste> it's easy to get freeze if there is not enough bandwidth, or if the bandwidth is not constant and there is no robust buffering logic
[23:52] <dmonjo_> saste:  how can you download a stream if its size is never known in HTTP?
[23:53] <saste> dmonjo_, dark magik
[23:54] <klaxa> wget
[23:55] <SubJunk> llogan: Thanks I'll try it out :)
[23:56] <saste> dmonjo_, also: http://127.0.0.1:8000:/event1.ogv
[23:56] <relaxed> why wouldn't libfdk_aac use -q:a instead of -vbr ?
[23:56] <dmonjo_> ?
[23:56] <saste> this is localhost so you basically have no latency and no drops
[23:56] <dmonjo_> no its a tunnel
[23:56] <dmonjo_> 127.0.0.1-> port redirection to the server
[23:57] <relaxed> there's no place like 127.0.0.1
[23:57] <llogan> ::1
[23:57] <dmonjo_> llogan:  what do you mean?
[23:58] <llogan> a dumb joke. ::1 is localhost in ipv6, IIRC.
[00:00] --- Fri Mar 15 2013


More information about the Ffmpeg-devel-irc mailing list