[Ffmpeg-devel-irc] ffmpeg.log.20130730

burek burek021 at gmail.com
Wed Jul 31 02:05:01 CEST 2013


[00:01] <alexschomb> exactly. not sure whether this is a problem anyway. right now it is still compiling ;)
[00:03] <alexschomb> wait, if I understood you right you mean that I should modify the PKG_CONFIG_PATH variable? I already tried .../output/lib/pkgconfig, but that didn't remove the error
[00:05] <roxlu> hi
[00:05] <alexschomb> hi
[00:05] <roxlu> can ffmpeg get input from a named pipe? (e.g. raw rgb images or audio)
[00:05] <llogan> alexschomb: probably don't need to. it looks like the external libraries were located anyway.
[00:07] <alexschomb> roxlu: I guess that answers your question? http://stackoverflow.com/questions/7605388/named-pipes-and-can-they-stream-data
[00:07] <alexschomb> llogan: no errors while compiling so far. so I guess you're right. Thanks for your fast help! :)
[00:09] <roxlu> alexschomb: partially :-)  I'm looking for a solution where I write into a named pipe
[00:14] <alexschomb> http://ffmpeg.org/ffmpeg-protocols.html#pipe
[00:14] <alexschomb> replacing pipe:0 with your named pipe doesn't work?
[00:15] <llogan> there are some examples in the FAQ: http://ffmpeg.org/faq.html#Concatenating-using-the-concat-protocol-_0028file-level_0029
[00:23] <roxlu> alexschomb: yeah I think I can just create a named pipe and use that as an input
[00:24] <roxlu> I only need to find a way to tell ffmpeg what kind of input I'm providing (raw rgb in this case)
[00:32] <alexschomb> llogan: finally compiled and copied the contents of my arm-directory to /usr/local/ on my RPI. but libaacplus is said to be missing although I followed the guide: http://pastebin.com/AXUgSbLJ
[00:36] <llogan> alexschomb: i've never used libaacplus. if i test that guide i may switch it to libfdk-aac from libaacplus
[00:41] <alexschomb> llogan: my fault. had to add /usr/local/lib to library path, http://www.cyberciti.biz/faq/linux-setting-changing-library-path/
[03:10] <Guest83225> I'm trying to use the encoder h264 with ffmpeg. I installed x264 from source and with few, and I configured ffmpeg with ./configure --enable-gpl --enable-shared --extra-ldflags="-ldl"  --enable-libx264. Yet the config.log tells me fatal error: 'x264.h' file not found
[03:10] <Guest83225> and I get the error libx264 not found
[03:29] <klaxa> Guest83225: you need to add --extra-cflags="-I/path/to/include" and maybe replace --extra-ldflags="-ldl" with --extra-ldflags="-L/path/to/lib"
[03:29] <Guest83225> thanks! to which command?
[03:29] <klaxa> configure for ffmpeg source
[03:30] <klaxa> /path/to/include should be the path to the directory that contains x264.h and /path/to/lib should be the directory that contains libx264.so.XXX
[03:30] <klaxa> whereas libx264.so.XXX should probably be libx264.so.134
[03:30] <klaxa> *135
[03:41] <Guest83225> Thanks! The configuration seems to have worked. Yet my application still shows this error when I try to transcode a video: Unknown encoder 'h264'
[03:43] <Guest83225> How can I verify that it worked? In the output of the configure command, I see 'bzlib                   libx264' in the list of External libraries
[03:44] <Guest83225> oh, I also see libx264 in the list of enabled encoders, so yes, it worked
[03:54] <relaxed> Guest83225: the codec encoder name is "libx264", -c:v libx264
[03:54] <relaxed> -codec
[03:56] <Guest83225> now I get Unknown encoder 'libx264'
[03:59] <Guest83225> at this point, I think I can figure it out, thanks!
[04:20] <Guest83225> nope, ffmpeg -codecs | grep 264 shows only  " D.V.LS h264                 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10" which is the decoder. I need the encoder, which is libx264
[04:23] <Guest83225> I'm running this command on the ffmpeg executable in the same directory I ran the ./configure
[04:25] <relaxed> Guest83225: you need to compile ffmpeg with libx264 support, or use  http://goo.gl/DPrRY
[04:30] <Guest83225> thanks! I don't see any installation instructions or configure or setup files. How do you recommend I install it?
[04:52] <Guest83225> nvm I'm on a mac. I'm looking for another build
[05:11] <relaxed> Guest83225: look at the ffmpeg.org download page
[05:38] <Guest83225> I ran through every step of https://trac.ffmpeg.org/wiki/MacOSXCompilationGuide, with only the error that libaacplus not found, yet when I ran make & make install, I got errors, beginning with this: Makefile:2: Makefile:2: config.mak: No such file or directoryconfig.mak: No such file or directory
[06:12] <Guest83225> nvm I got it up and running
[09:54] <nlight> what behavior can I expect from swscale if the input and output formats are the same?
[10:08] <b6> the AVFrames i'm reading have pkt_pts seemingly in milliseconds. i think those frames were encoded with microsecond precision. how can i tell?
[10:18] <elkng> I have a question related using ffmpeg libraries, anyone know something about ffmpeg libraries ? I think they can be used to work directly with wideo files on a low level, am I right ?
[10:18] <nlight> elkng, you are right
[10:18] <elkng> all right, so the question is:
[10:19] <elkng> I have a program like some sort of video editor but its more like frame manager, to edit video I convert video into bunch of png images and that program then uses SDL to show them and I can use "Left" "Right" to change images left and right like in some video editor, that program is better than image viewer like xnvies because if you use xnview
[10:19] <elkng>  when you change images they blink and its very tiresome for eyes, also that program can do some other things, but now I want to know is it possible to somehowe use ffmpeg libraries or something to load frames directly from video file ? now I need to convert entire video to png images first then use something like "image=IMG_Load(path_to_file);
[10:19] <elkng>  SDL_BlitSurface(image)" to load image from file and show it on the screen, I really would like not to convert video to png files because 5 minutes of video will consume 3GB of disk space, and its also time consuming operation, so is it possible to somehow skip part of converting video to images and use some functions from ffmpeg libraries to load
[10:19] <elkng>  images from video file ?
[10:20] <nlight> yes, that's what ffmpeg is for
[10:21] <nlight> go here http://dranger.com/ffmpeg/tutorial01.html
[10:25] <elkng> nlight: but how fast it would be ? I have celeron 500, and in case I allready have converted video to png files, that program can show them at a rate about 10 frames per second because all it needs to do is to open png file, and its fast, but if I instead would use ffmpeg library how fast that library could access needed frame load it and give it
[10:25] <elkng>  to my program for it to eventually show it ?
[10:28] <elkng> also my program uses SDL function "image = IMG_Load()" and it gets that image in "SDL_Surface" format, that also could increase time processing because of conversion, isn't it ?
[10:28] <elkng> beceuse I would need to convert from ffmpeg format to SDL_Surface format
[10:35] <elkng> any ideas ?
[10:35] <elkng> worth to try ?
[10:36] <nlight> celeron 500?
[10:36] <elkng> yes
[10:37] <nlight> what?
[10:37] <nlight> really?
[10:37] <elkng> yes
[10:37] <nlight> I don't think you can decode video in real time on that but dunno
[10:37] <elkng> "on that", was that offence ?
[10:38] <nlight> it's ancient hardware, so yes :D
[10:38] <nlight> it's a 7 years old cpu
[10:42] <elkng> I got it on 20 april in 2000
[10:43] <elkng> nlight: that PC also has video: Riva TNT, 400MB RAM 133Mhz, CDROM 48X speed, 2x80GB HDD
[10:43] <elkng> impressive ? monster
[10:43] <elkng> Riva TNT 16MB
[10:43] <nlight> wow
[10:44] <elkng> and any video I need to convert to 650xXXX so it will not be too slow to watch
[10:45] <nlight> well, go ahead and try it out
[10:46] <nlight> what is the difference between AVInputFormat and AVOutputFormat?
[10:46] <nlight> and why does AVFormatContext have both?
[10:46] <nlight> should I expect only one of iformat/oformat to be available at any time?
[11:07] <hendry> anyone tried streaming to icecast using ffmpeg? A "PUT" IIUC?
[12:27] <jcath> hi, friends. does ffmpeg support the blu-ray directory as input?
[12:28] <Fjorgynn> blu-ray direcotry?
[14:13] <pyBlob> I've got an avi file that contains 2 mjpg-streams
[14:15] <Fjorgynn> aha
[14:15] <pyBlob> ffmpeg -i HNI_0001.AVI -vcodec copy -f image2 dump/out%05d.jpg
[14:15] <pyBlob> that extracts the single jpg-frames
[14:16] <pyBlob> problem is that only the first 2 frames contain a HuffmanTable
[14:16] <pyBlob> > other jpg-files can't be read by standard image viewers
[14:20] <relaxed> pyBlob: try with -bsf:v mjpeg2jpeg
[14:23] <pyBlob> relaxed: that works, thanks =)
[14:24] <relaxed> you're welcome
[14:34] <b4u> is there a way to set a crf with libx264 but also have an upper bound on the bitrate rather than an average?
[14:36] <JEEB> bitrate is always an average of something
[14:36] <JEEB> amount of bits over something
[14:36] <JEEB> anyways, you will have to set BOTH -bufsize and -maxrate
[14:36] <b4u> is it possible to tell the encoder not to encode any frames higher than a specific bitrate though?
[14:36] <b4u> at all
[14:36] <b4u> but don't I need to know what will be buffered for bufsize?
[14:36] <JEEB> do you really need per-picture VBV?
[14:37] <JEEB> you really need to have a specific use case to need per-picture VBV
[14:38] <b4u> bandwidth limited situation where I need to know it won't go over a certain bitrate but can't guarantee a certain amount will be buffered in advance either
[14:39] <JEEB> well single picture VBV still means you will have to have that picture's amount buffered at the very least
[14:40] <b4u> okay, how do I do something like that?
[14:40] <b4u> I've not heard of it before
[14:40] <JEEB> I won't tell you unless I'm sure you need it :P
[14:40] <JEEB> because it means that encoder will be really limited and thus you are pretty much derping quality that way
[14:41] <b4u> well my other way of doing it was to set an arbitrarily low bufsize
[14:41] <b4u> of like 200k
[14:42] <JEEB> anyways, it sounds like you are dealing with something you could very well fix somewhere else
[14:43] <JEEB> instead you're just opting to *heavily* derp the encoder
[14:44] <b4u> does a low bufsize do that also?
[14:44] <JEEB> yes
[14:46] <b4u> is there a way to do this even with 2 pass?
[14:47] <JEEB> uhh, what? VBV works the same way no matter what rate control you are using (except maybe constant quant, although no-one should in general be using that rate control mode except for development purposes)
[14:51] <b4u> the problem is I can't guarantee a certain amount of data will be buffered
[14:51] <b4u> I can do a certain amount of seconds
[14:51] <b4u> which may vary depending on the video
[14:53] <JEEB> it can vary of course, but in general it should be OK. Just use <maxrate> * seconds_to_buffer in that case for bufsize?
[14:53] <bencoh> wow, two vbv-days in a row :)
[14:53] <JEEB> and match them that wise
[14:53] <b4u> uhh but that won't work will it?
[14:53] <JEEB> why wouldn't it
[14:53] <b4u> or could produce bad buffer times
[14:54] <b4u> because right now people wait a number of seconds worth of buffer to view a video rather than data
[14:54] <b4u> so I guess small videos will start more slowly than normal
[14:54] <b4u> perception wise
[14:55] <JEEB> I don't follow you
[14:56] <b4u> well right now a video with 100k bitrate takes not long to start
[14:56] <b4u> because 6 seconds of that is not much
[14:56] <JEEB> and it still wouldn't be because flash would then see that not much data is going on?
[14:56] <b4u> what I mean is
[14:57] <b4u> ok so you said set bufsize to seconds * maxrate?
[14:57] <b4u> that means the encoder assumes I will buffer that much before playback
[14:57] <b4u> yes?
[14:57] <JEEB> yes
[14:57] <b4u> my player only buffers in seconds
[14:57] <JEEB> yes, most probably some kind of dynamical thing
[14:58] <JEEB> so it looks how much data is flowing over and tries to adapt to that
[14:58] <JEEB> so it should still buffer the same amount?
[14:58] <b4u> it's a constant in seconds, I think
[14:58] <JEEB> seconds of content I would guess :P
[14:59] <JEEB> or dunno
[14:59] <b4u> but I mean the first seconds of the video might not be at that maxrate
[14:59] <b4u> so surely I will not necessarily buffer enough data
[14:59] <b4u> to meet what the encoder expects
[14:59] <JEEB> as far as I know flash shouldn't be that dumb
[14:59] <JEEB> as to not be adaptive related to your use of bandwidth
[14:59] <b4u> how would it know when to start playing?
[15:00] <b4u> short of having a time or data limit
[15:00] <JEEB> anyways, tons of people are using it well
[15:00] <JEEB> so you are most probably looking at it too bad :P
[15:00] <JEEB> and yes, I hate things that don't let you set the amount of data to buffer
[15:01] <JEEB> anyways, single picture VBV totally was the wrong thing for you, as I thought
[15:01] <b4u> so what kind of buffer limit should I set do you think? I feel like normalseconds * maxrate is too much
[15:01] <JEEB> just use bufsize = <seconds to buffer> * maxrate
[15:01] <b4u> I could maybe do 1 or 2 seconds at maxrate
[15:04] <JEEB> anyways, for all of the things I've so far seen if you have to set a time on the decoder side to buffer, the bufsize to be calculated with that should be <time_to_buffer> * maxrate
[15:04] <nlight> anyone seen these kind of artifacts - http://i.imgur.com/tob4Bm5.jpg ?
[15:17] <JEEB> b4u, I just checked the blog of a certain x264-using streaming thingy, and it uses the exact same thing to calculate the matching between seconds of buffering and bufsize
[15:18] <JEEB> and this is with CRF rate control, yes
[15:25] <luc4> Hi! I noticed that libavformat needs to seek when using callbacks muxing a stream. Is this always necessary? Can I avoid this somehow or is it impossible to avoid?
[15:26] <JEEB> b4u, btw I would say that such VBV-specific questions with x264 might be better off being asked on #x264 rather than here
[15:28] <b4u> sorry, was afk
[15:30] <Sum> Afternoon chaps - I'm having trouble getting ffmpeg to install in ubuntu 13.04 64 bit using the guide on http://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide.  Running ffmpeg 2>&1 | head -n1 gives an error saying I  need to install ffmpeg via apt-get install.  Any ideas?
[15:30] <JEEB> Sum, sounds like you're just not having wherever you compiled your ffmpeg binary in your PATH
[15:32] <b4u> you didn't forget to "make install" right? :P
[15:33] <Sum> yes - i did 'make install'
[15:33] <b4u> probably what JEEB says then
[15:34] <Sum> @ubuntu:~/ffmpeg_sources/ffmpeg$ find ~ -name ffmpeg -print  gives:
[15:34] <JEEB> uhh
[15:34] <JEEB> you set a prefix as well as a bindir when configuring
[15:34] <JEEB> IIRC the guide says to do that
[15:34] <JEEB> the bindir would be the place where the binary gets installed
[15:35] <Sum> ah - i just copied and pasted the commands - will read a bit further
[15:40] <Sum> ah - i see I think.  I guess running the following will fix the path to the executable?  echo 'export PATH=$PATH:/home/sum/bin' >> ~/.bashrc
[15:40] <JEEB> basically yes
[15:40] <JEEB> in the next opened terminal that is
[15:40] <JEEB> since that file is run every time you open bash
[15:41] <JEEB> for currently open terminals you would have to run that export part in the terminal
[15:42] <Sum> brill!  thanks :-)
[17:25] <Sum> Hey folks.  I'm wondering if you can help with this: I don't appear to have the 'libfaac' encoder installed and following https://help.ubuntu.com/community/Medibuntu hasn't appeared to install that encoder.  any hints on where to get this encoder from and how to install it?
[17:26] <microchip_> Sum: did you ask in #ubuntu ?
[17:26] <JEEBsv> That's because faac was found to be nonfree, aka not distributable. Which means you will have to compile it yourself. That said, now we have something better than that, so you will most probably want to build fraunhofer's encoder instead :)
[17:27] <JEEBsv> (fdk-aac)
[17:29] <JEEBsv> http://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide there's always this, but this includes way too many encoders in it, so feel free to read through and compile only the external encoders you need
[17:38] <Sum> ok - is this another instance where the fdk-aac 'file' (executable?) isn't in the path environmental variable (or whatever it's called in unix?  As per the Ubuntu Compilation Guide I ran make, make install adn make distclean adn there is a file that looks like the one I need - /home/sum/ffmpeg_sources/fdk-aac - is this the file that should be in a path statement?
[17:41] <JEEBsv> Sum: ffmpeg -codecs:a |grep "aac"
[17:42] <JEEBsv> see if it's listed there
[17:42] <JEEBsv> if it is, you can just use its name with -c:a
[17:45] <Sum> JEEBsv: - it isn't I think - output is  DEA.L. aac                  AAC (Advanced Audio Coding) (encoders: aac libfdk_aac )  D.A.L. aac_latm             AAC LATM (Advanced Audio Coding LATM syntax)
[17:45] <Sum> do i need to recompile ffmpeg with this encoders support added?
[17:47] <klaxa> >encoders: aac libfdk_aac
[17:47] <klaxa> that looks good though
[17:48] <JEEBsv> Sum: libfdk_aac is fdk-aac :)
[17:48] <JEEBsv> so -c:a libfdk_aac will use it
[17:48] <JEEBsv> instead of -c:a libfaac for example
[18:04] <Sum> JEEBsv - fantastic!  Now transcoding at 110fps instead of 8 - I'm a very happy man.  Thanks for your patience.  :-)
[20:21] <phr3ak> how could I join several asf?
[20:28] <durandal_1707> phr3ak: with concat demuxer
[22:31] <roxlu> hi guys, I'm a bit cross posting, but I hope someone here knows this ....
[22:31] <roxlu> when I combine a video and audio file, how can I set that the resulting video takes the duration of the video, not of the audio?
[00:00] --- Wed Jul 31 2013


More information about the Ffmpeg-devel-irc mailing list