[Ffmpeg-devel-irc] ffmpeg.log.20120829

[burek]

[00:09] <llogan> Lns: ffmpeg2theora is not a FFmpeg project
[00:11] <llogan> and ffmpeg can encode with libvorbis and libtheora
[00:12] <Lns> llogan, ty..i'm about to try using ffmpeg to do this then just pipe it to ffmpeg2theora =)
[00:12] <llogan> what does ffmpeg2theora do for you that ffmpeg itself can not?
[00:15] <Lns> llogan, encode to theora? Can ffmpeg do that itself and pipe it to stdout?
[00:15] <Lns> sorry, i'm kind of patching things together, that would be GREAT if it could completely replace ffmpeg2theora
[00:16] <Lns> looks like it can...heh
[00:17] <Lns> so this page is misleading: http://current.workingdirectory.net/posts/2010/video4linux-and-audio/
[00:17] <llogan> ffmpeg -i input -c:v libtheora -c:a libvorbis -q:v 6 -q:a 5 -f ogg -
[00:18] <llogan> adjust q:* as desired
[00:18] <Lns> nice.
[00:18] <Lns> i think i love you.
[00:18] <Lns> haha
[00:18] <llogan> or you can simply declare bitrates
[00:18] <llogan> -b:v and -b:a
[00:19] <Lns> that'll be a test for later...all of the icecast folks seem to swear that quality setting is superior to static bitrates
[00:19] <Lns> but i've had issues in the past with some listening clients using that
[00:20] <llogan> i think q:a range for libvorbis is -1 to 10, but i'm not sure if ffmpeg will parse -1 as expected
[00:21] <Lns> I'm going to be recording/streaming live music and i've had good luck with 2
[00:22] <llogan> hopefully that will give you a good start. i have almost no experience with live stuff
[00:22] <Lns> it did, thanks. i'll have more questions i'm sure =)
[00:23] <Lns> debian is complaining that ffmpeg (in my context anyway) is depreciated and i should use avconv... yes/no?
[00:24] <llogan> http://stackoverflow.com/questions/9477115/who-can-tell-me-the-difference-and-relation-between-ffmpeg-libav-and-avconv/9477756#9477756
[00:24] <tuxx_> hey guys are there instructions on how to crosscompile ffmpeg?
[00:24] <tuxx_> i'm trying to build it for arm
[00:25] <tuxx_> i did ./configure --arch=arm --cross-prefix=... --target-os=linux
[00:25] <tuxx_> it seems to be building so far..
[00:25] <Lns> llogan, gak.. ok thanks for that
[00:25] <llogan> Lns: so obviously we encourge ffmpeg comilation
[00:25] <llogan> *compilation
[00:26] <llogan> Lns: you can adapt this: https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide
[00:26] <llogan> or the Hardy version linked in that guide might be better for your debian
[00:27] <llogan> and there are also "static" ffmpeg binaries available at http://ffmpeg.org/download.html if oyu're lazy
[00:28] <llogan> you might want to compile anyway since there was a commit 5 days ago that, "enerates output with bitrates similar to libvorbis for a given quality
[00:28] <llogan> ...value. *generates
[00:28] Action: llogan learns how to paste
[00:31] <Lns> llogan, thanks =) i'll try the package versions first, i'm just trying to prove i can do this..but later on this will be a big project so i want the most stable version as possible
[00:31] <llogan> ffmpeg from recent git is usually "stable"
[00:33] <llogan> anyone seen burek around lately?
[00:33] <Lns> heh, and testing is at 0.8.3-6
[00:34] <llogan> Lns: that's a libav version, not FFmpeg.
[00:40] <Lns> hrm, this is strange.. " Unrecognized option 'c:v' " and then " Failed to set value 'theora' for option 'c:v' " ...
[00:40] <Lns> even though manpage specifies -c:v
[00:41] <Lns> err, s/theora/libtheora/
[00:41] <JEEB> you set it after -i, right?
[00:41] <Lns> yep
[00:41] <JEEB> also sure your build contains libtheora and is new enough?
[00:41] <Lns> that might be the prob, how do i check?
[00:42] <llogan> maybe your build uses -acodec and -vocdec instead of -c:a, -c:v
[00:42] <llogan> *-vcodec, damn it
[00:43] <Lns> http://pastebin.com/Xrwb6ram
[00:43] <llogan> ah, i didn't see you mention the man page.
[00:44] <shevy> Anyone knows whether this here is bad? [mp3float @ 0xe06e40]Header missing
[00:44] <shevy> for a .mp3 file
[00:44] <shevy> :(
[00:45] <llogan> that's usually bad AFAIK, mmmmkay.
[00:45] <JEEB> yes, it usually means it's a broken file
[00:45] <Lns> so i know it said /dev/video0 invalid device.. i used -f video4linux2 before -i /dev/video0 and it seemed to work but i was confused as to llogan's command earlier with '-f ogg'
[00:46] <JEEB> before and after -i is different
[00:46] <JEEB> before -i = input/decoders
[00:46] <JEEB> after -i = output/encoders
[00:46] <Lns> oh =)
[00:46] <llogan> i added that because if you pipe out you need to specifiy the output format
[00:47] <shevy> hmm damn, ok
[00:47] <JEEB> shevy, in case you believe the file is not broken
[00:47] <JEEB> you should check the trac for similar reports
[00:47] <JEEB> and if you don't find them, report and give a sample
[00:47] <Lns> ok here's something a bit more useful: http://pastebin.com/vcjAmHS6
[00:47] <JEEB> it will be then checked if the sample is broken
[00:48] <shevy> JEEB yeah I was just trying out mpgtx and its merge ability
[00:49] <shevy> but even during the "merge", it complained... and the resulting .mp3 is odd
[00:49] <shevy> when playing it in mplayer, it spams out 100s of those lines
[00:49] <shevy> the original file does not spam anything much at all
[00:49] <shevy> ffmpeg can merge .mp3 files too or?
[00:50] <Lns> ah, i think i got somewhere with getting rid of all the -c:N parms. using -acodec and -vcodec (i should file a bug for the manpage..)
[00:54] <shevy>   ffmpeg -i concat:start.mp3\|end.mp3 -acodec copy new_file.mp3
[00:54] <shevy> that seems to have worked
[00:54] <shevy> at least I don't get those "Header missing" things anymore
[00:54] <shevy> :)
[00:58] <llogan> Lns: you would have to report the bug to ubuntu if the fix is already in libav.
[00:59] <llogan> and many of us can't really provide reliable answers for libav products here (JEEB excluded).
[01:00] <Lns> mmk
[01:00] <JEEB> oh
[01:00] <JEEB> just switch from ffmpeg to avconv command line wise
[01:00] <JEEB> that one is updated in libav
[01:01] <JEEB> libav's elenril rewrote quite a few spots of ffmpeg, and renamed that avconv
[01:01] <JEEB> then ffmpeg was left around unupdated for a release
[01:01] <JEEB> which is what you have
[01:01] <JEEB> ffmpeg (the project) has merged most if not all of those changes into its ffmpeg app
[01:02] <JEEB> so if you use libav, you use avconv, if you use ffmpeg you use ffmpeg
[01:10] <Lns> damn..i feel like i'm getting in over my head atm.. heh.. here's what I have now (changed up my command a bit): http://pastebin.com/vA2UfTLL
[01:11] <Lns> the only diff if i s/ffmpeg/avconv is that it gives me a bunch of 'this message repeated 3 times' instead of one line of increasing value
[01:13] <JEEB> well, let's just say that within libav, avconv is the thing that's up-to-date on that release.
[01:13] <JEEB> might want to ask on #libav
[01:13] <Lns> JEEB, so you're saying commandline options for both ffmpeg+avconv are interchangable?
[01:14] <JEEB> mostly yes
[01:14] <Lns> ok
[01:14] <llogan> he's saying to use avconv if you insist on using libav.
[01:14] <Lns> ok..i think i'm going to compile ffmpeg =)
[01:15] <JEEB> you might want to ask on #libav tho, if that's a known bug or not
[01:15] <JEEB> there are some differences like the fact that ffmpeg uses x264opts, while libav has separate options f.ex.
[01:33] Action: Lns falls down the compilation black hole
[01:34] <ubitux> this might help if you're lazy
[01:36] <Lns> well i got it done but it's giving me "Unknown encoder 'libtheora' " even though i swear i saw that included right after ./configure
[01:38] <Lns> trying static
[01:39] <Lns> static works
[01:41] <llogan> did you install libtheora-dev?
[01:41] <Lns> yep
[01:41] <llogan> did you add --enable-libtheora?
[01:42] <Lns> no.. ;) but it's weird, static ffmpeg build still gives me no audio
[01:42] <llogan> of course it's all guesswork unless you provide the console output
[01:42] <llogan> and your commands
[01:42] <Lns> hold on
[01:42] <llogan> some users like the guessing game
[01:43] <Lns> http://pastebin.com/naKgUG6N
[01:44] <llogan> you have no audio input
[01:44] <Lns> oh jesus..haha
[01:44] <llogan> lol
[01:45] <Lns> well my /dev/video0 has a mic, how do i specify that w/ffmpeg?
[01:46] <llogan> https://ffmpeg.org/trac/ffmpeg/wiki/Capturing%20audio%20with%20FFmpeg%20and%20ALSA
[01:49] <Lns> " Unknown input format: 'alsa' "
[01:51] <Lns> i need a break...i'll idle here though...thanks for the huge amount of help thus far!
[01:51] <Lns> bbl
[02:12] <llogan> Lns: i don't know if the static build(s) support alsa
[02:27] <GrenBaykre> How does ffmpeg handle the fact that JPEG by nature is lossy when doing PNG => JPEG with -sameq?
[02:29] <llogan> GrenBaykre: sameq does not mean "same quality" and is not designed to be used between formats that do not share a similar quantizer scale.
[02:30] <GrenBaykre> Yes, it does mean that.
[02:30] <GrenBaykre> Unless I'm horribly wrong.
[02:30] <GrenBaykre> Quantizer scale is a term I am unfamiliar with.
[02:30] <llogan> the docs used to imply it, unfortunately. they have been updated some time ago.
[02:31] <llogan> the simplest advice i can give you regarding that option is to never use it
[02:31] <GrenBaykre> What's so bad about it? :|
[02:32] <llogan> sameq only works in a few situations
[02:33] <GrenBaykre> Does it harm in others?
[02:34] <llogan> since it was not designed to convert between quant scales it's behavior may not be desirable. i'm not sure what it does when it's used like that.
[02:35] <GrenBaykre> But...
[02:35] <GrenBaykre> I don't understand what quantizer scales even are.
[02:35] <GrenBaykre> Dangit.
[02:35] <GrenBaykre> ffmpeg is incredibly powerful. I must assume that countless software embed it internally... right?
[02:35] <GrenBaykre> For all kinds of importing video and converting it and whatnot.
[02:36] <llogan> the libraries can be used by other things, and some use the ffmpeg cli program ditrectly
[02:36] <llogan> *directly
[02:39] <GrenBaykre> Seems like any media player would want to use it.
[02:42] <GrenBaykre> To the owner of this channel: will you please ban any known log bot from here?
[02:43] <llogan> we have a useful log bot.
[02:45] <GrenBaykre> It's not useful. It violates privacy. :|
[02:45] <llogan> this is a public channel
[02:47] <llogan> your privacy concerns would be better directed at more non-benign entities such as if you are a facebook user.
[02:47] <GrenBaykre> I am not.
[02:47] <GrenBaykre> Public logging is the spawn of Satan IMO.
[02:47] <GrenBaykre> I cannot think of any valid use of it.
[02:47] <llogan> you could always not type anything.
[02:48] <GrenBaykre> Not an option.
[02:48] <llogan> this is not really the place for such a discussion. do you have any ffmpeg related questions?
[02:50] <GrenBaykre> The one I entered with is my main one.
[02:51] <llogan> i answered that already, but to summarize, "i don't know".
[04:06] <rocktop> I couldn't get ffpresets under ffmpeg directory
[04:06] <rocktop> anyidea how to get them ?
[04:18] <relaxed> rocktop: ffmpeg/presets
[04:27] <rocktop> relaxed: I am looking for libx264 presets
[04:28] <relaxed> ffmpeg uses libx264's internal presets. Look at $(x264 --fullhelp | less)
[04:29] <relaxed> ffmpeg -i blah -c:v libx264 -preset veryslow ...
[04:31] <rocktop> relaxed: File for preset 'hq' not found
[04:35] <relaxed> Look at `x264 --fullhelp | less` for a list of libx264 presets. You dig?
[04:36] <Sashmo> im encoding from a transport stream source, I keep going out of sync on my streams, any one have any ideas on what I could be doing wrong?  My source is h.264, and the output is a lower bitrate h.264
[04:52] <aspirant> can ffserver be used for live streaming?  Because what it seems to do is play the stream back from the start every time anyone connects
[05:06] <rocktop> I'm trying to play a video on JWPlayer . I was wondering what it would take to create an mp4 with ffmpeg to be able to psudostream? the problem is when i run the video file it starts buffering, waits until completion of the buffer and plays the video
[05:10] <relaxed> rocktop: run qt-faststart on the input
[05:10] <relaxed> s/input/mp4/
[05:12] <rocktop> relaxed: how ?
[05:16] <relaxed> cd ffmpeg; make tools/qt-faststart
[05:16] <relaxed> qt-faststart input.mp4 output.mp4
[05:18] <rocktop> relaxed: its OK now , Thank you very much
[05:19] <relaxed> you're welcome
[06:09] <aspirant> is ffserver even supposed to be used for live streaming (I.E. from a webcam or surveillance)?
[07:07] <aspirant> Is there a good free streaming server that I can use to stream live to HTML5?
[11:42] <coalado> I'd like to get as much details about my mediafiles as possible. I use ffprobe -i -show_streams -show_format
[11:42] <coalado> But this seems to be slightly inaccurate
[11:44] <sacarasc> Try mediainfo.
[11:49] <coalado> sacarasc:  mediainfo does not support http stream input
[11:56] <saste> coalado: "slightly inaccurate" it is inaccurate
[11:57] <coalado> at least not as detailed as mediainfo
[12:06] <saste> coalado: which information are you trying to get, what is missing
[12:07] <coalado> ffprobe maps many containers to  a formatid format_name=mov,mp4,m4a,3gp,3g2,mj2
[12:08] <coalado> profile informations are missing.
[12:08] <coalado> it cannot differ between aac and he-acc
[12:24] <saste> coalado: yes
[12:24] <saste> coalado: I'm not sure there is some way to export that info
[12:25] <saste> anyway you could file a feature request on trac
[14:38] <DX099> hello all
[14:39] <DX099> can ffmpeg perform old film colourization ?
[14:46] <DX099> ?
[15:04] <saste> DX099: do you mean to take a B/W video and add colour to it?
[15:07] <DX099> saste, yes
[15:08] <saste> DX099: you can't
[15:08] <saste> at least not without something which knows how things are colored
[15:08] <saste> e.g. a human being
[15:09] <DX099> ah ok
[16:11] <cinead> I'm having trouble converting 3gp to avi. I want to convert a cell phone video to avi and simultaneously strip the audio out to save file size.
[16:12] <cinead> When I try to convert both the video and audio streams I get an error that says "Error, Invalid timestamp=1580, last=1580. Video encoding failed."
[16:13] <cinead> I used: ffmpeg -i file.3gp -f avi file.avi
[16:13] <cinead> And I am running ffmpeg for Windows, version N-36193-gf514695 compiled with gcc 4.6.2
[16:14] <relaxed> ffmpeg -i file.3gp -map 0:v -q:v 3 output.avi
[16:17] <cinead> Interesting. This successfully strips the audio (thank you) but I'm still getting a video encoding error at timestamp=1580
[16:18] <cinead> The video is about 9 minutes long. When I try to view the resulting avi (after the error) it plays the first minute or so successfully and then stops. Is there a way to fix this timestamp error with a command-line workaround? Or is it a problem with the video?
[16:26] <relaxed> cinead: your version is pretty old, try git
[16:27] <cinead> Yeah, good idea. Thanks for your help.
[16:41] <epettrey> w/in 3
[16:41] <epettrey> whoops, sorry
[17:34] <DX099> hello, i'm building ffmpeg do I need to remove libavcodec ?
[17:40] <DX099> ?
[17:53] <ramiro> hi
[17:54] <ramiro> saste, michaelni: could you guys take a look at the dshow patches? ("More DirectShow patches" and the followup to Roger's dshow patch)
[18:01] <saste> ramiro: give me a few hours
[18:01] <ramiro> thanks
[18:02] <DX099> hello, i'm building ffmpeg do I need to remove libavcodec ?
[18:03] <ramiro> DX099: no
[18:04] <DX099> thanks
[18:11] <shevy> anyone knows off hand whether ffmpeg has more contributors than mplayer on average, per any given month of the year?
[18:12] <ramiro> back in 2010 there was much more contribution to ffmpeg than mplayer
[18:12] <ramiro> I don't know how things are now after the split
[18:12] <saste> shevy: check stats on ohloh
[18:14] <shevy> oh I see
[18:14] <shevy> ohloh says mplayer 16 current committers
[18:14] <shevy> and ffmpeg 282 current committers
[21:23] <hackeron> hey, I'm using opencv's CvCapture *input = cvCreateFileCapture("rtsp://192.168.88.11") which interally uses FFMPEG. It opens the RTSP with the UDP transport. I asked the maintainers of opencv how to pass rtsp_transport but it appears to be impossible so I'm looking for a way to tell FFMPEG to use TCP by default instead of UDP. Anyone has any suggestions where to look? - maybe this line in ffmpeg.c? < { "rtsp_transport", "RTSP transport ...
[21:23] <hackeron> ... protocols", OFFSET(lower_transport_mask), AV_OPT_TYPE_FLAGS, {0}, INT_MIN, INT_MAX, DEC|ENC, "rtsp_transport" },
[22:19] <shevy> hmm
[22:19] <shevy> are there python or perl or ruby bindings to ffmpeg directly?
[22:20] <hackeron> Any familiar with opencv? - how do I pass something like an AVDictionary or specifically do the equivalent of av_dict_set(&opts, "rtsp_transport", "tcp", 0); when using CvCapture *input = cvCreateFileCapture("rtsp://192.168.88.14/"); to let ffmpeg know that I want the tcp transport?
[22:30] <saste> hackeron: seems more a problem which should be addressed at the libopencv level
[22:30] <saste> the dictionary use in libavformat is fairly new, they may need to update their API
[22:34] <hackeron> saste: I tried asking them and filing a bug, but it's been about a month and nobody seems to take interest. I tried to make sense of the code but it's beyond me unfortunatly :( - is there anyway to just compile ffmpeg to force TCP to always be used? - where in the ffmpeg code is UDP used by default?
[22:35] <saste> hackeron: maybe you can hack the url to pass
[22:35] <saste> did you check the rtsp docs?
[22:37] <hackeron> saste: ffmpeg dropped support for adding ?tcp at the end, the docs say to pass rtsp_transport but I can't in opencv, so I would like to change the ffmpeg source to just always use tcp - I never use UDP :P - I'm looking at libavformat/rtsp.c but the code doesn't make much sense to me :( - can you point me in the right direction?
[22:37] <hackeron> I see stuff like switch(reply->transports[0].lower_transport) { case RTSP_LOWER_TRANSPORT_TCP:
[22:38] <saste> hackeron: so the question is really, why do you want to do that?
[22:38] <hackeron> but I can't seem to find where it gets udp by default :/
[22:38] <hackeron> saste: because I'm reading from IP cameras and the image is horribly broken if UDP is used - there are a bunch of threads online about this
[22:39] <saste> hackeron: what about using an old ffmpeg version, supporting the ?tcp thing?
[22:40] <saste> and no i have no special knowledge of rtsp code (never looked at it)
[22:40] <hackeron> saste: doesn't support the latest versions of x264 that work far better with the somewhat dodgy h264 coming out of most IP cameras
[00:00] --- Thu Aug 30 2012