[Ffmpeg-devel-irc] ffmpeg.log.20120519

[burek]

[00:02] <xp99work> hello
[00:02] <burek> o/
[00:02] <xp99work> I'm trying to compile 0.10.3 of ffmpeg
[00:03] <xp99work> and even though I tell it to disable all decoders and only enable certain ones, when it does linking, it tries to link to vc1
[00:04] <burek> what is vcl
[00:04] <xp99work> a decoder
[00:05] <burek> can you please use pastebin.com, to show your command line and its output?
[00:06] <ironm> thank you ... just to see if I will still have artifacts like with crf=26 ... <Mavrik> ironm: um, you're trying to make a huge losless webm file?
[00:06] <ironm> thank you very much for the hint :) ... <sacarasc> ironm: You don't need -strict experimental with that line.
[00:13] <xp99work> sure
[00:14] <ironm> hello burek. Are you familiar a bit with ffmpeg options for mp4 files for Apple?
[00:16] <burek> I don't like Apple, nor support it, sorry :/
[00:16] <ironm> eg. lile:  ffmpeg -i test.mp4 -vcodec libvpx -crf 5 -acodec libvorbis -strict experimental -ab 90k test10.m4v
[00:17] <burek> it's just another flavor of linux for which you have to pay
[00:17] <ironm> I hate apple and don't use it ..
[00:17] <burek> it's just stupid
[00:17] <ironm> but mayn people do :/
[00:17] <burek> well yes, and as long as we do support those idiots, apple will be popular
[00:17] <ironm> you are right
[00:18] <ironm> I can run .mp4 video files on samsung note n7000 (android 2.x) ...
[00:18] <ironm> but they dont't work on iPhone
[00:19] <ironm> and what phone do most of women have?
[00:19] <teratorn> quicktime is a real whore
[00:19] <ironm> just not Android ...
[00:19] <ironm> yes teratorn
[00:19] <burek> :)
[00:19] <teratorn> from what i understand quicktime will break on all kinds of presets
[00:23] <ironm> burek, what should I say ... even most of my male colleagues use iPhone ... :(
[00:23] <burek> well, the point is to make things work for your pleasure and watch all your colleagues how nothing works on their "gadgets" but work on yours :)
[00:24] <ironm> :)))
[00:24] <burek> I mean, did you read about "segmented http streaming"
[00:24] <burek> those people are really crazy :)
[00:24] <ironm> I will do only small exception for some particular women then :)
[00:25] <ironm> no .. I didn't read about "segmented http streaming". Is it something new from Apple ?
[00:27] <xp99work> do you want the config.log file?
[00:27] <burek> i'm not quite sure if that's the right terminology, but it's really a shot in the head from an ambush
[00:27] <burek> xp99work, ok
[00:27] <ironm> a bit strange .. even I used crf=0 I have still artifacts I don't have in the input .mp4 file
[00:27] <burek> "suicide from an ambush"
[00:27] <burek> :)
[00:28] <ironm> ironm at sid:~$ ffmpeg -i test.mp4 -vcodec libvpx -crf 0 -acodec libvorbis -strict experimental -ab 90k test10.webm
[00:28] <burek> im not sure if crf 0 is lossless for libvpx too
[00:28] <burek> it is for libx264
[00:28] <xp99work> http://pastebin.com/MeHZKmQ6
[00:28] <ironm> burek, could I use .ogv as input file instead?
[00:30] <burek> xp99work, 'vcl not found'
[00:30] <burek> ?
[00:30] <ironm> I used this file as input: http://www.it-infrastrukturen.com/fileadmin/images/videos/xe_CLI_scripts_to_create_VMs_on_XCP%2BXenServer.mp4
[00:31] <ironm> the .ogv file is nearly three times bigger: http://www.it-infrastrukturen.com/fileadmin/images/videos/xe_CLI_scripts_to_create_VMs_on_XCP%2BXenServer.ogv
[00:31] <xp99work> http://pastebin.com/XxR3KfBr
[00:31] <xp99work> thats when it tries to install
[00:31] <burek> ironm, if you want to be sure that "-crf 0" is lossless for libvpx too, you might want to check libvpx's source code..
[00:31] <xp99work> and gets the linking error with vc1_parser.c
[00:31] <burek> I'm sure for libx264 it is lossless, but not sure for libvpx
[00:32] <ironm> burek, thank you.
[00:33] <burek> xp99work, "--enable-parsers"
[00:33] <burek> "libavcodec/libavcodec.a(vc1_parser.o): In function `vc1_parse_init':"
[00:33] <xp99work> oh when I run ./configure ?
[00:34] <burek> you configured with --enable-parsers
[00:34] <burek> and vc1 parser is giving you a linking error
[00:34] <burek> most probably you have some different version of it
[00:34] <xp99work> so disable parsers then?
[00:34] <burek> without those exports
[00:34] <burek> you can disable that parser only?
[00:35] <burek> ./configure --help
[00:35] <xp99work> ah ok
[00:35] <xp99work> I'll do that
[00:35] <xp99work> thanks :)
[00:35] <burek> :beer: :)
[00:46] <ironm> git clone http://git.chromium.org/webm/libvpx.git
[00:47] <ironm> hmm .. there is empty output of: ironm at sid:~/libvpx$ grep -R crf ./*
[00:48] <xp99work> awesome!
[00:48] <xp99work> it worked :D
[00:48] Action: xp99work takes a beer
[00:48] <burek> xp99work, \o/ :)
[00:54] <ironm> burek, it is probably somewhere in ./vp8/encoder/ratectrl.c
[00:54] <ironm> http://paste.debian.net/170058/
[00:54] <ironm> anyway there is no "crf" inside the whole source code tree
[00:56] <burek> ironm, http://ffmpeg.org/ffmpeg.html
[00:56] <burek> search for crf
[00:56] <burek> you'll see this: ‘crf’: VPX_CQ, VP8E_SET_CQ_LEVEL
[00:56] <burek> so, I'm pretty sure your -crf X maps to those 2 vpx params
[00:57] <burek> (cq and cq_level)
[00:57] <ironm> thanks a lot burek  ... in the code of libvpx there is no "crl" at all
[00:58] <ironm> "crf"
[01:00] <burek> well, I know
[01:00] <burek> ffmpeg re-uses -crf for both encoder
[01:00] <burek> just remapping that option to those encoder's native options
[01:00] <burek> like cq and cq_level
[01:00] <ironm> *_*
[01:00] <burek> when you type -crf X
[01:01] <ironm> good idea ... I will look for cq / cq_level
[01:01] <burek> ffmpeg calls libvpx with -cq X -cq_level X
[01:01] <burek> something like that
[01:01] <ironm> merci :)
[01:01] <burek> :)
[01:01] <ironm> 5th beer ?
[01:01] <ironm> :)
[01:02] <burek> ^^
[01:02] <burek> shh :)
[01:02] <ironm> whatever it is .. (ssh)
[01:02] <burek> psssst :)
[01:04] <ironm> sur bien ;)
[01:07] <ironm> ./vp8/vp8_cx_iface.c:    RANGE_CHECK(vp8_cfg, cq_level, 0, 63);
[01:07] <ironm> the question is: what value gives the best quality ... "0" or "63"
[01:08] <ironm> I should give up for today ..
[01:09] <burek> A new VP8 encoder based on the x264 framework is under development. The name of the new project is xvp8.
[01:09] <burek> meow :D
[01:09] <burek> ironm try it :)
[01:09] <burek> put -t 30
[01:09] <burek> and run 2 times :)
[01:10] <ironm> "-t 30" where ?
[01:10] <ironm> on the CLI of ffmpeg?
[01:11] <burek> yes
[01:11] <burek> just before the output name
[01:11] <burek> it will create the output of 30 seconds
[01:11] <burek> (speeds up your testings)
[01:13] <ironm> OK
[01:17] <ironm> burek, can I list somehow a list of available codes for installed ffmpeg package?
[01:17] <ironm> codecs
[01:17] <burek> ffmpeg -codecs
[01:17] <burek> :)
[01:18] <ironm>  DEV    libvpx          libvpx VP8
[01:18] <ironm>  D V D  vp8             On2 VP8
[01:18] <burek> also -formats, -bsfs, -filters, protocols ..
[01:18] <burek> vp8 is just a decoder
[01:19] <burek> libvpx is both
[01:19] <ironm> hmm ...
[01:19] <ironm> do I need to install new package for xvp8 support ?
[01:37] <pisto> can anyone help me? I want to pipe into ffmpeg some images (pngs) **but I don't want to save them on the disk**. I have them already piped out of my program. How can tell ffmpeg to accept such stream?
[01:38] <burek> pisto, that's easy
[01:38] <burek> you need -f image2pipe
[01:39] <burek> something like this:
[01:39] <burek> cat $(ls *.png) | ffmpeg -f image2pipe -vcodec ... output
[01:39] <burek> of course, you don't want to use files, but you get the point
[01:40] <burek> sorry, the example was cut
[01:40] <burek> this is the complete example:
[01:40] <pisto> I'm pretty sure that I tried -f image2pipe before, but again I'll try again
[01:40] <burek> cat $(ls *.png) | ffmpeg -f image2pipe -vcodec png -r 25 -i - <encoding options> output.avi
[01:42] <pisto> [png @ 0x2d88ec0] chunk too big
[01:42] <pisto> [png @ 0x2d88ec0] Missing png signature
[01:43] <burek> invalid png images?
[01:43] <burek> old ffmpeg?
[01:43] <burek> old libpng?
[01:43] <burek> can you please use pastebin.com, to show your command line and its output?
[01:44] <pisto> here you are burek http://pastebin.com/A80G4eBi
[01:45] <pisto> I don't know how to tell if gnuplot is outputting bad png images after the first one (which I can view)
[01:45] <pisto> the resolution is 1366*768, I'm going to try to lower that
[01:45] <burek> hmh..
[01:45] <burek> try this
[01:46] <burek> cat $(ls *.png) | ffmpeg -f image2pipe -r 25 -i - <encoding options> output.avi
[01:46] <burek> (remove -vcodec png)
[01:47] <pisto> [image2pipe @ 0x1770240] Could not find codec parameters (Video: none)
[01:47] <pisto> [image2pipe @ 0x1770240] Estimating duration from bitrate, this may be inaccurate
[01:47] <pisto> pipe:: could not find codec parameters
[01:47] <burek> hm
[01:47] <burek> try with bmps
[01:47] <burek> *.bmp and -vcodec bmp
[01:48] <burek> also, that way you don't recompress your images twice (and loose quality)
[01:48] <pisto> I won't be able to use that, because gnuplot produces crappy quality graphics with bmp. Anyway,I'm goingmto try
[01:49] <pisto> (btw, png is lossless)
[01:51] <pisto> hm, seems it works
[01:52] <pisto> shows a blank screen, but that must be a problem with my gnuplot script
[01:52] <pisto> is there a png->pbm streamer around?
[01:54] <burek> I didn't use it
[01:54] <burek> so I really don't know, sorry
[02:30] <pisto> how do I optimize x264 for monochrome input?
[03:23] <PineappleCLock> how can I get better resolution out of this: `ffmpeg -i input.avi -ar 22050 -ab 32000 -f flv -s 640x480 -vcodec flv -b 1000000 output.flv`
[04:37] <avocado> try -b 2000k
[04:37] <avocado> PineappleCLock: ^
[04:38] <PineappleCLock> thanks
[04:38] <avocado> or just keep raising the bitrate
[06:31] <aum> hi - i got files in raw h264 format - i can play them in ffplay by using the '-f h264' option - but i want to convert them. how do i tell ffmpeg that they are raw h264?
[07:47] <krion> hi guys
[07:49] <krion> i'm running ffmpeg-git and x264-git, trying to transform and avi in mp4, i'm having an error message with google doesn't help with
[07:49] <krion> [libfaac @ 0x2f03d40] Que input is backward in time
[07:51] <krion> i see in the topic you talk about -preset -tune or -profile for x264 i will have a look at that
[07:53] <krion> for the record here is my command line http://sebsauvage.net/paste/?149d7c9e4f654c28#aTMpVOZumePrkw6AogoK3axKspsRBqA61qVsQe51Dlc=
[14:10] <silverrocker> I have a question about H.264 standard but I"m not sure if it is completly related to ffmpeg. Can I ask that here?
[14:13] <ironm> try to ask
[14:15] <saste> silverrocker: don't ask to ask, just ask
[14:16] <silverrocker> okay, I'm confused how a given level is encoded into a stream. the standard describes the decoding only (from the decoders perspective). I would like to know how a level is constructed
[14:16] <silverrocker> the code that is, I have been searching for a full day now without any success. I also tried to reverse the algoritme that decodes the levels from the standard but I seem to fail with it.
[14:17] <silverrocker> other sources refer to Variable Length Coded tables which map levels to their encoding but I can't find those anywhere.
[14:18] <silverrocker> I have searched the sources of the reference encoder/decoder but couldn't find where the codes are constructed
[14:49] <casso> I am trying to compile program under qt creator, that uses FFMPEG. Upon compiling I get: undefined reference to 'av_register_all' and many more like that
[14:50] <casso> I use latest dev for ffmpeg
[14:56] <Mavrik> your bindings are old and not compatible with new ffmpeg anymore
[14:57] <casso> you mean my code, which contains function calls for ffmpeg?
[14:59] <Mavrik> yeah.
[14:59] <Mavrik> grab an older release of ffmpeg, depending on just how old API you're using
[15:01] <silverrocker> casso: av_register_all(); should be correct, I'm using a fairly recent version of ffmpeg in my project and it works
[15:02] <silverrocker> casso: make sure you add the correct libraries to your qtproject aswell
[15:02] <casso> I did
[15:03] <silverrocker> casso: have you also done this; DEFINES += __STDC_CONSTANT_MACROS
[15:03] <casso> silverrocker: in .pro file?
[15:03] <silverrocker> yes
[15:04] <silverrocker> casso: for some reason the ffmpeg source is not made extern "C" so you have to wrap your includes in that aswell to let the C++ coompiler know that you are calling C functions. other libraries have extern "C" in there headers but ffmpeg doesn't so you
[15:04] <silverrocker> casso: have to add it manually, otherwise you compiler will not be able to find the functions
[15:05] <silverrocker> extern "C" { #include <libavcode/...>; #include <..other ffmpeg stuff..> }
[15:05] <silverrocker> casso: I'm guessing that is your problem? (I had the same problem once)
[15:05] <casso> I have extern "C"... already
[15:06] <silverrocker> casso: can you pastebin your pro file?
[15:08] <casso> http://pastebin.com/vkLQGukU
[15:10] <silverrocker> casso: add -lm to the libraries
[15:11] <silverrocker> casso: still same problem?
[15:11] <casso> I added this:   LIBS += -lm
[15:11] <casso> but still the same
[15:13] <casso> wait
[15:14] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=tree;f=doc/examples;hb=HEAD
[15:14] <casso> no, still the same
[15:14] <ubitux> have a look to this
[15:14] <silverrocker> casso: I have LIBS += -lm -lavdevice -lavformat -lavfilter -lavcodec -lswscale -lavutil
[15:14] <ubitux> check the Makefile for build options
[15:14] <silverrocker> casso: I have always with the l infront
[15:15] <silverrocker> casso: that's probably the problem at your side
[15:15] <Mavrik> hm, yeah, sorry, I've mixed it up with another lib, av_register_all is indeed still in the API :)
[15:15] <casso> If I put l in front of each library, it says cannot find file
[15:16] <Mavrik> if you don't pass "-lavformat -lavcodec ... " etc
[15:16] <Mavrik> you will get missing reference errors since the program won't be able to link with libav libraries
[15:16] <Mavrik> casso: you might also need "-L<path to libav files>" for correct linking
[15:16] <casso> I am doing that on windows
[15:16] <Mavrik> (libav files being libavcodec.(a, dll, so), libavformat.(a, dll, so))
[15:17] <silverrocker> casso: I don't know exactly how to get it working on windows, I'm working on linux, sorry
[15:18] <silverrocker> casso: but it's the libraries. make sure you have them installed
[15:20] <casso> I'll try same in linux
[15:58] <sporn> hello!
[15:59] <sporn> anybody there?
[16:00] <sporn> i need some help with using AVIOInterruptCB
[16:07] <sporn> nobody here?
[16:23] <nijm> Why might a decoder not know what the the type of frame (I/P/B) the first few frames in an mpeg transport stream are?
[16:35] <pisto> is there a specification of the format of the input file name?
[16:35] <pisto> i.e. input files with %
[16:38] <Bombo> pisto: -i %%04d.jpg like this
[16:39] <pisto> I want a complete reference, I know that example
[16:40] <pisto> by the way that forces the input filename to have trailing zeroes, doesn't it?
[16:40] <Bombo> yes it reads 0000.jpg to 9999.jpg
[16:41] <pisto> http://ffmpeg.mplayerhq.hu/ffmpeg.html#toc-Video-and-Audio-file-format-conversion
[16:41] <pisto> got it.
[16:54] <sporn> Anybody know what *ctx contains when interrupt_callback fires? from AVIOInterruptCB?
[17:37] <arch_is_awesome> Hello, I was wondering who it was I talked to last time about _things_ here...
[17:37] <malinens> Hi! What does GOP mean(-g option). I know it creates keyframe after x frames but why it is needed? Do I need to set keyframes for live transcoding? I know it is good for pseudo streaming and better seeking...
[17:52] <nijm> Is it possible to get the address of an mpeg2 sequence header and the addresses of individual frames in an mpeg ts file with libavformat/libavcodec?
[17:52] <nijm> Addresses in the file that is.
[18:35] <valaminicknev> Hi all
[18:35] <valaminicknev> can i control ffmpeg if i'm starting it for example from java?
[18:36] <valaminicknev> so basically cli
[18:36] <valaminicknev> for example pause, start?
[18:49] <CTPAX> Hi, I'm having problem with building FFMPEG NaCl port. Who can I ask for more info?
[18:49] <herdingcat> I notice that ffmpeg cannot record many cameras video at the same time stream in real time, can anyone tell me why? 2 hrs recording only get ~30mins video stream file.
[18:52] <herdingcat> is there anything related w/ real time?
[18:56] <herdingcat> anyone can help me?
[19:01] <jimbodude> Anyone know anything about the meaning of this message: "[mp4 @ 0x8ee8840] pts has no value"?  I am remuxing AVI to MP4, and I get that message repeatedly.
[19:02] <jimbodude> Resulting file plays, but can't skip/fastforward past a certain point, usually around 5-7 minutes in.
[19:05] <jimbodude> command I'm using is: ffmpeg -i input.avi -vcodec copy -acodec copy output.mp4
[19:06] <jimbodude> Version is the latest from Ubuntu PPA for 10.04: ffmpeg version 0.9.1-4:0.9.1-0ubuntu1~jon1~lucid1
[19:09] <jimbodude> incoming data is XVID MPEG-4, 720x480 @ 24 fps, AC-3 5.1 @448kbps
[20:30] <khali> hi all
[20:30] <khali> is ffplay supposed to hide the mouse cursor in fullscreen mode?
[22:17] <An_Ony_Moose> is this the right channel to ask questions about avconv?
[22:18] <An_Ony_Moose> I'll just assume it is and ask: what is the option used to set the audio source? I want to use anullsrc rather than the audio included in the video file I'm converting.
[22:48] <An_Ony_Moose> never mind I didn't know the audio source counted as a filter
[22:57] <Sazpaimon> what's the status on mplayer filters in ffmpeg
[22:57] <Sazpaimon> im desperately looking for mplayer's expand filter
[22:58] <ubitux> can't you do that with the pad filter?
[22:58] <ubitux> also, isn't expand available through -vf mp=expand=...
[22:58] <ubitux> ?
[22:59] <Sazpaimon> it is not
[22:59] <ubitux> it might have been removed because we have a native one for it
[22:59] <Sazpaimon> i dont think there's an equicalent to expand=:::::16/9 for pad
[23:00] <Sazpaimon> basically in mplayer I use expand=:::::16/9,dsize=854:480 to fit a video of any resolution and aspect ratio into a 854x480 16:9 area without affecting the source video's aspect ratio
[23:00] <Sazpaimon> so a 320x480 video will show as 640x480 pillarboxed
[23:01] <ubitux>                # pad the input to get a final w/h ratio of 16:9
[23:01] <ubitux>                pad="ih*16/9:ih:(ow-iw)/2:(oh-ih)/2"
[23:01] <ubitux> i see this in the examples
[23:01] <Sazpaimon> oh shoot how did I miss that
[23:01] <Sazpaimon> will that work with dsize?
[23:02] <ubitux> isn't that a scale filter?
[23:02] <ubitux> pad="ih*16/9:ih:(ow-iw)/2:(oh-ih)/2",scale=854:480 ?
[23:03] <Sazpaimon> i believe scale screws with the aspect ratio
[23:03] <ubitux> check the scale examples
[23:03] <ubitux> it has variables too
[23:03] <ubitux> to honor sar or whatever
[23:03] <Sazpaimon> well ffmpeg support dsize through the mp filters
[23:04] <ubitux> :)
[23:04] <Sazpaimon> according to the docu, dsize doesn't do any scaling itself
[23:05] <Sazpaimon> "it just affects what later scalers (software or hardware) will do when auto-scaling to correct aspect."
[23:05] <Sazpaimon> so I guess scale would be better with pad
[23:05] <ubitux> (maybe setsar and setdar filters?)
[23:05] <Sazpaimon> the source aspect ratio is unknown
[23:06] <Sazpaimon> so i can't specify it on the command line
[23:06] <Sazpaimon> the aspect ratio and resolution can also change mid-stream
[23:06] <Sazpaimon> it's an ogg theora stream
[23:06] <Sazpaimon> does ffmpeg support this?
[23:07] <ubitux> it should
[23:09] <Sazpaimon> pad="ih*16/9:ih:(ow-iw)/2:(oh-ih)/2",mp=dsize=854:480 seems to work fine
[23:10] <Sazpaimon> also whats the best way to get ffmpeg to read a stream over HTTP? just pipe wget into it?
[23:10] <Mavrik> ffmpeg -i http://<stream>?
[23:11] <Sazpaimon> that never worked for me
[23:12] <Sazpaimon> gives me a no such file or directory message
[23:12] <Sazpaimon> oh, i guess it would help if i emerged ffmpeg with the network use flag
[23:19] <Sazpaimon> ubitux: "[ogg @ 0x743030] Changing stream parameters in multistream ogg is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented."
[23:19] <Sazpaimon> I'm using the latest git as of 10 minutes ago
[23:19] <ubitux> ah then too bad
[23:20] <ubitux> if you provide a sample and open a track issue, someone might be working on
[23:20] <ubitux> afaik it's supported in mpeg-ts so the current ffmpeg design should allow such thing
[23:21] <Sazpaimon> also I tried another stream, and that gives me "[pad @ 0xc3e110] Negative values are not acceptable."
[23:22] <Sazpaimon> the video stream shows as: "Stream #0:1: Video: theora, yuv420p, 864x464 [SAR 18400:18361 DAR 993600:532469], 23.98 fps, 23.98 tbr, 23.98 tbn, 23.98 tbc"
[23:24] <Sazpaimon> also, using a normal everyday theora file instead of a stream makes the output webm file play very fast in an html5 video tag
[23:37] <Sazpaimon> so I guess that dsize example doesnt work if one of the dimensions of the input is larger than the output
[23:39] <funkster> Looking to do some fading transitions between images like in jquery slideshows - possible to do with ffmpeg?
[23:42] <Sazpaimon> I tried an example from the mailing list: http://ffmpeg.org/pipermail/ffmpeg-user/2011-July/001746.html
[23:43] <Sazpaimon> but that also didn't work: "Input area 0:10:865:474 not within the padded area 0:0:864:486 or zero-sized"
[00:00] --- Sun May 20 2012