[Ffmpeg-devel-irc] ffmpeg.log.20111109
burek
burek021 at gmail.com
Thu Nov 10 02:05:01 CET 2011
[00:00] <redgetan> ic, i just assumed that by default, ffmpeg is gonna use the original codecs
[00:01] <b0nghitter> i'm trying to input an FLV stream from an rtmp source. if the source does not have any audio, ffmpeg will not play the stream. which params should i be using? thnks http://pastebin.com/BzaTAWyU
[00:02] <b0nghitter> when i remove the -acodec copy, it does the same thing. if i add audio to the stream, it works.
[00:03] <sacarasc> -an
[00:03] <b0nghitter> ty
[00:04] <b0nghitter> i replace the acodec copy with an? i tried and it still wont play
[00:10] <redgetan> pasteeater: your command worked
[00:10] <redgetan> putting the -ss on the output did the trick
[00:11] <redgetan> when i tried to put in the input with the vcodec copy / acodec copy, it still gave me the 149.9s
[00:11] <burek> ffmpeg has trouble with flv
[00:11] <burek> a lot
[00:11] <redgetan> so re-encoding didnt seem to affect things
[00:13] <pasteeater> redgetan: what are you using to determine the duration of the output?
[00:14] <redgetan> ffmpeg -i video
[00:14] <burek> b0nghitter can you give us the output of the command you used to feed into ffmpeg from rtmp source (use pastebin.com)
[00:17] <redgetan> im usually confused about durations though, ffmpeg -i only gives me one duration 00:02:29.02
[00:17] <redgetan> while using mediainfo, i'll get three durations: one for general, one for video and one for audio
[00:18] <redgetan> general: 2mn 29s, video: 2mn 27s, audio: 2mn 29s
[00:19] <redgetan> and i'm thinking that maybe i shouldn't be cutting videos just at some random location, i was told that video cutting should be done at IDR frames to avoid audio desync issues
[00:20] <b0nghitter> burek: i am streaming from flash player, one of adobes demo players
[00:20] <blane1> Hi all.
[00:21] <b0nghitter> it only has the video attached to the stream
[00:22] <b0nghitter> can ffmpeg play an flv rtmp stream with no audio? it appears to be the problem.
[00:22] <blane1> Anyone have it in them to school a n00b on streaming to an HTML5 <video> tag? I'm on localhost so minimal DNS to deal with.
[00:24] <b0nghitter> the rtmp source is flash media server, which i am streaming to from flash player.
[00:25] <b0nghitter> the free development server, which is very nice in conjunction with ffmpeg to split into multi-bitrate
[00:26] <b0nghitter> but ffmpeg will only play the stream if i add audio
[00:31] <b0nghitter> if you would liek to try yourself, feel free to use the interactive publishing to create a stream, i am just using the publish button, http://fms.ganjanet.org
[00:31] <b0nghitter> rename the streem to whatever you need
[00:32] <b0nghitter> it should work, i have tried all sorts of commands
[00:32] <b0nghitter> still, no luck
[00:36] <queuetip> what is -crf ? i'm debugging some code and the encoding uses -crf 30 but i can't seem to find what that is on the ffmpeg docs
[00:37] <pasteeater> b0nghitter: can you try ffmpeg from git head? that will rule out any old bugs.
[00:37] <queuetip> another one is -refs 3 i have no idea what this does :|
[00:37] <b0nghitter> i'll try thanks
[00:38] <pasteeater> crf is a "rate control method", or in other words it controls the quality of the output. it is generally used with libx264 or libvpx video encoders.
[00:39] <pasteeater> you shouldn't have to monkey with refs. the presets deal with that.
[00:39] <queuetip> pasteeater: this is already existing as part of application code that deals with a large amount of file uploads and encoding : | i'm not 100% sure why it's being used
[00:40] <pasteeater> if you don't declare a rate conrtol method with libx264 then it will use a default value of -b 200k which is terribly low for most uses.
[00:41] <queuetip> pasteeater: i see, so the crf should be set, but what exactly is refs?
[00:41] <queuetip> pasteeater: and also why can't i find -refs as a param in the docs?
[00:42] <pasteeater> ffmpeg -h | grep refs
[00:43] <pasteeater> is the application encoding with libav* libraries or via the ffmpeg binary?
[00:47] <queuetip> pateeater: via the binary
[00:49] <pasteeater> if you show the command we can give suggestions
[00:55] <WillMarshall> Hey guys! How can I get ffmpeg to spit out ogg vorbis?
[00:55] <WillMarshall> With -f ogg it gives me flac...
[00:55] <sacarasc> -acodec libvorbis
[00:56] <WillMarshall> sacarasc: Thanks! Many people on the tubes are talking about ffmpeg2theora - why would I want to use that over just ffmpeg?
[01:00] <queuetip> pasteeater: ffmpeg -i %video_path% -metadata title=%video_title% -metadata comment=www.break.com
[01:00] <queuetip> -y -vcodec libx264 -crf 30 -refs 3 -bf 16 -g 60 -keyint_min 25 -sc_threshold 40
[01:00] <queuetip> -b_strategy 1 -bframebias 0 -coder 1 -directpred 1 -deblockalpha -1 -deblockbeta -1 -trellis 1 -me_method umh
[01:00] <queuetip> -subq 6 -me_range 16 -qcomp 0.6 -qmin 10 -qmax 51 -qdiff 4 -i_qfactor 0.71 -b_qfactor 1.3 -b_strategy 1
[01:00] <queuetip> -partitions +parti8x8+parti4x4+partp8x8+partb8x8 -flags +loop-psnr -flags2 +wpred+mixed_refs+dct8x8+fastpskip
[01:00] <queuetip> -cmp +chroma -maxrate 25000 -acodec libfaac -ab 96k -s %width%x%height% -threads %thread% %output_name%
[01:00] <queuetip> sorry, lemme use pastebin
[01:01] <queuetip> http://pastebin.com/BwWXtnm9
[01:09] <pasteeater> queuetip: what a mess. can you also pastebin your ffmpeg version info?
[01:45] <blane1> Hi all.
[01:46] <hi117> hi blane1
[01:46] <blane1> I've read through the documentation and am still having a difficult time understanding the difference between the HTTP stream from ffmpeg, and what ffserver produces.
[01:46] <blane1> Any help?
[01:46] <elfMobile> so I did a -crf 25 and then a -crf 5 and both videos ended up 156M
[01:46] <elfMobile> that make sense?
[01:47] <sacarasc> elfMobile: What was the complete command line?
[01:48] <elfMobile> ffmpeg -y -i input.flv -acodec libfaac -ab 96k -vcodec libx264 -crf 5 -threads 0 output.mp4
[01:48] Action: sacarasc shrugs.
[01:49] <pasteeater> elfMobile: use a pastebin site to show your ffmpeg command and the complete terminal output
[01:49] Action: pasteeater pastes
[01:50] <elfMobile> hmmm... I'm trying my command followed by | wgetpaste
[01:51] <elfMobile> it seems like too much is going to stderr or directly to the terminal?
[01:51] <elfMobile> I'm running again with > /tmp/out 2> /tmp/err
[01:52] <pasteeater> try output.mp4 2>&1 | wgetpaste
[01:52] <pasteeater> although i'm not familiar with wget paste
[01:58] <elfMobile> http://paste.pocoo.org/show/504949/
[01:58] <elfMobile> ^^ pasteeater
[01:59] <elfMobile> whoops
[01:59] <elfMobile> disconnected... I'm back now
[01:59] <elfMobile> ... I think
[02:01] <pasteeater> elfMobile: last one. the output of "strings output.mp4 | grep x264"
[02:03] <pasteeater> nevermind. that info is in your paste.
[02:03] <pasteeater> it's actually using crf 23 which is/was the default
[02:18] <elfMobile> weird.... so -crf doesn't do anything
[02:22] <pasteeater> it should. it does for me. you're probably encountering a bug that has already been fixed in newer ffmpeg.
[02:55] <queuetip> pasteeater: sorry i was afk for a while for dinner. I don't have access to the production machines which these are running on, but i'm pretty sure the encoder lives on an ubuntu 10.10 box, so lemme see if i can tell which version comes with the apt repos for ubuntu 10.10
[02:56] <queuetip> pasteeater: FFmpeg version 0.6-4:0.6-2ubuntu6.2, Copyright (c) 2000-2010 the FFmpeg developers
[02:56] <queuetip> built on Sep 16 2011 17:04:54 with gcc 4.4.5
[02:56] <queuetip> pasteeater: i'm pretty sure that's the ancient version in use : |
[03:04] <Mista_D> there are faint grey lines whenever video is scaled and encoded, especially visible in dark areas. Any filter there that might help??
[03:10] <Mista_D> http://i40.tinypic.com/2vi01fq.jpg - see if you can spot the lines.
[03:14] <queuetip> Mista_D: to me it seems like those lines were part of the original
[03:14] <queuetip> Mista_D: I think you can try to use -refs as a parameter. (reference frames to consider for motion compensation (Snow))
[03:32] <Mista_D> queuetip: the lines were not on the original. I thought it was x264 but the I saw the artefacts in y4m (raw video) when it was scaled. http://doom10.org/index.php?topic=2014.0
[04:30] <pfloyd> sort of off topic, but does anyone know of a "dead pixel" 1080p video that not only cycles colors but also stays all one color for a while? a friend of mine notices a high pitched noise from his tv when it's all white so I'm trying to find a video that is widescreen but also shows an all-white frame for a few seconds
[04:33] <Mista_D> try -i /dev/null and then inverse it.
[04:33] <pfloyd> -f raw?
[04:33] <Mista_D> sure
[04:34] <Mista_D> -s 1920x1080 -f rawvideo -pix_fmt rgb24 -r 25 -i /dev/null
[04:34] <pfloyd> heh, /dev/null: Invalid data found when processing input
[04:34] <pfloyd> perfect, thanks
[04:35] <pfloyd> but how do I invert it?
[04:38] <Mista_D> '/dev/zero instead /dev/null works.. invert with mencoder, it has a ton of filters.
[04:39] <pfloyd> I found an easier solution :) I just made a 1920x1080 all white image, since the popcorn hour will display images. thanks for the idea though, good to know
[04:39] <pfloyd> shouldn't I be able to use tr or something to invert the \0 to 0xFF or whatever white is ?
[04:42] <pfloyd> heh /dev/zero is all green
[04:52] <dericed> after upgrade to latest git version, the -psnr options is unrecognized. Is this renamed or gone?
[10:43] <teratorn> can ffmpeg load standard image file formats? e.g. I want to convert a PNG in to a YUV420P coded AVFrame
[10:47] <ubitux> yes, ffmpeg -i in.png -pix_fmt yuv420p -f rawvideo -y out.raw should do the trick i guess
[10:48] <teratorn> OK, that's neat I'll try it - but I want to write that loader using the C API
[10:50] <ubitux> the pix format convert can be done with swscale (sws_getCachedContext + sws_scale), you can see follow what's done in ffmpeg.c and see doc/examples for the rest of the API usage
[10:52] <teratorn> yeah, I'm familiar with sws_scale
[10:52] <teratorn> but I wouldn't know how to get an rgb frame (presumably) to start with
[11:22] <patrakov> hi. which output format should I use for the rather unusual task of recording both from the webcam (video4linux2 + alsa) and from x11grab? i.e. a format that supports two video streams and one audio when written by ffmpeg?
[11:29] <Mavrik> huh
[11:29] <Mavrik> .mkv probably
[11:33] <patrakov> doesn't quite work (for my without audio, see http://pastebin.com/GEJQTNkj). it just discards the second video
[11:33] <patrakov> for my attempt*
[11:34] <burek> i think vob might do
[11:34] <burek> but just type ffmpeg -formats
[11:34] <burek> and try whichever works
[11:35] <burek> also .ts
[11:35] <burek> -f mpegts
[11:35] <burek> that can hold a lot of stuff :)
[11:37] <burek> dericed, http://ffmpeg.org/ffmpeg.html says its still there
[11:41] <burek> queuetip: http://en.wikipedia.org/wiki/Constant_rate_factor (crf)
[11:42] <burek> instead of keeping constant bitrate or constant quantization parameter (cqp), its sometimes better to keep the "constant quality"
[11:43] <burek> thats what crf is for.. also, lower crf = better quality (0 usually means lossless)
[11:50] <burek> Mista_D, you need to dither/deband that
[11:50] <burek> http://www.google.com/search?q=ffmpeg+dither
[12:34] <simon-argh> Trying to encode video to with HTML5 video tag in IE9. I encoded two videos from OGV to MP4 using the same command, one worked one failed in IE9. The command I used was:ffmpeg -i 350home.ogv -acodec libfaac -aq 100 -vcodec libx264 -preset slow -crf 22 350home.mp4
[12:34] <simon-argh>
[12:36] <simon-argh> Looking at output of "ffmpeg -i" the videos are slightly different sizes, slightly different lengths, and the failing one reports DAR and PAR values in the output, the working one doesn't. Newbie to video encoding and just wanted to establish that HTML5 video would work mostly (and mostly it does except for IE8 and IE9 sigh).
[12:38] <simon-argh> I have flash fall back for IE8, but IE9 just displays some basic video controls on right click but doesn't seem to have a video to display to me.
[12:39] <simon-argh> I retried encodng from the webm version rather than OGV. I am using the Debian Multimedia versions of x264 and related libraries in Debian Squeeze.
[12:42] <teratorn> does ffmpeg have any functions for blending frames with an alpha channel?
[12:43] <teratorn> e.g. to overlay one image on top of another with some level of transparency?
[13:24] <dericed> burek: I saw -psnr was still on that page, however with recent builds it isn't recognized. I filed a bug https://ffmpeg.org/trac/ffmpeg/ticket/629
[13:26] <burek> thanks :)
[14:43] <burek> simon-argh, if you are going to work on multimedia you _really_ need to escape from debian "stable" logic
[14:44] <burek> either move on to testing
[14:44] <burek> or use debian's multimedia repository
[14:44] <burek> (unofficial)
[14:47] <burek> teratorn, try overlay video vilter using -vf
[14:47] <burek> http://ffmpeg.org/ffmpeg.html#overlay
[14:48] <teratorn> burek: thanks, I'll check it
[15:19] <simon-argh> burek: urm - yes it was Debian Multimedia repository, I presume I'm missing an option of some sort since MIRO produced a bigger MP4 which worked from the same input. So something about how I'm trying to invoke it - presumably if I cared I could dig through the MIRO presets to see what is what.
[15:20] <burek> does ie9 have any error console output
[15:20] <simon-argh> burek: sorry should have mentioned the multimedia - but that it offers encode to h264 should be a give away it isn't Debian's own.
[15:21] <burek> it is a wise thing to compile your own x264, because its not big
[15:21] <burek> and you always stay up to date with latest speed/quality tricks
[15:27] <simon-argh> burek: no relevant errors it did log a complaint about the Google Plus 1 button - the only obvious difference aside from data rate and the like is that MIRO is Lavf 53.18.0 rather than 53.13.0 Debian Multimedia produced. I'm prepared to believe it could be simply an old bug and leave it at that
[15:27] <simon-argh> burek: thanks
[15:32] <burek> simon-argh, can you give us the output of "ffmpeg -i inputfile", for both files, using pastebin.com
[15:39] <simon-argh> burek: http://pastebin.com/buKyZ0Qz
[15:40] <simon-argh> burek: I don't think it is simply settings as I said I created one video that worked and one that didn't using the same command (file names excepted). Since the input movies were different it is hard to rule out data against software.
[15:47] <burek> simon-argh, the output contains information about video too..
[16:06] <stonie> hi.. how may i pass lame options (like -V0) when encoding with libmp3lame?
[16:09] <burek> stonie, what is -V0
[16:10] <burek> try -aq 0
[16:12] <Uriziel> Hello, how to gram the very last frame of movie
[16:12] <Uriziel> grab*
[16:13] <burek> i would use ffmpeg -i input -ss X -f image2 out%3d.png
[16:13] <burek> and would set X to a time close to an end of file
[16:13] <burek> the resulting image with the highest number would be the last frame
[17:21] <stonie> burek, -V0 is the highest quality option for lame iirc
[17:22] <burek> stonie, ok, try ffmpeg -i ... -acodec libmp3lame -ar 44100 -aq 0 -ac 2 output.mp3
[17:38] <designbybeck> Greetings all, awhile back I had a line of code that would simple allow me to broadcast my webcam (not record it) on screen so that I could be seen when doing a screen capture demo
[17:39] <designbybeck> I put this command in Compiz Commands so I coud just use a keystroke to launch my webcam feed. I just don't remember the command and can't find the site i got it off of? Can anyone help!
[17:51] <sereal-work> I can't seem to find this in the man page, is there a -n argument (the opposite of -y ) ?
[17:53] <Mavrik> not really
[17:53] <Mavrik> you'll just have to check if the file exists in some other way
[17:58] <sereal-work> doh
[17:58] <sereal-work> i'll just do it the lazy way and do -y
[18:05] <burek> what would -n do
[18:05] <burek> designbybeck http://www.google.com/search?q=ffmpeg+screen+capture
[18:05] <burek> ffmpeg -f x11grab -r 25 -s 1280x720 -i :0.0+0,48 -vcodec libx264 -vpre lossless_ultrafast -threads 0 video.mkv
[18:06] <designbybeck> Thanks burek , that shines a little more light on to it. I do have it currently broadcasting my webcam using mplayer
[18:07] <designbybeck> I'm using gtkrecordmydesktop for the screencapture itself
[18:08] <designbybeck> mplayer -cache 128 -tv driver=v4l2:width=640:height=480:outfmt=i420 -vo xv tv://
[18:08] <designbybeck> that is currently the line that is working for me
[18:08] <stonie> burek, ok.. do you know if its joint stereo then? because we dont need that
[18:09] <designbybeck> Now I'm trying to figure out how to adjust the Hue/Contrast and such of the webcam feed
[18:09] <burek> stonie, use mediainfo or winamp to see what it is?
[18:09] <burek> designbybeck, use v4l2-ctl
[18:09] <burek> apt-get install v4l-utils
[18:09] <designbybeck> isn't that commandline only?
[18:10] <burek> im not sure how to do it within ffmpeg
[18:10] <burek> maybe using -f video4linux -i /dev/video0:options
[18:11] <designbybeck> ah yes that is commandline
[18:16] <stonie> joint, so..
[18:19] <burek> try adding -async 1
[18:21] <burek> Joint stereo is used in lossless compression. Since it is lossless, there is obviously no loss of anything. Joint stereo is a technique to save storage space, nothing more.
[18:21] <burek> why would you not want joint?
[18:29] <evil_core> hi all
[18:29] <evil_core> I am searching for any2h264 script using ffmpeg
[18:30] <evil_core> I found some commands, but I dont have presets that they are using, dunno if its because they are obsolete?
[18:32] <burek> what exactly your goal is
[18:34] <evil_core> I am downloading many strange things, usually I want to convert from DivX to h264, but now I downloaded some big soundclips with Microsoft Video 1 and PCM
[18:35] <evil_core> I simply want to have something called from CLI that will to the best to convert from anything to h264, saving original aspect ratio, resolution, etc
[18:38] <evil_core> ffmpeg -y -i "$iFile" -an -vcodec libx264 -vpre medium -b 800k -threads "$threads" -f mp4 /dev/null
[18:38] <evil_core> ffmpeg -y -i "$iFile" -acodec libfaac -ab 128k -vcodec libx264 -vpre medium -b 800k -threads "$threads" "$oFile"
[18:39] <evil_core> and it doesnt find 'medium' profile, original was medium_firstpass (which also doesnt exist)
[18:40] <evil_core> so am I missing those profiles and should download them, or there were depracted in ffmpeg, and I shoulf find another commands?
[18:42] <burek> well
[18:42] <burek> you can use -profile and -preset
[18:42] <burek> which now maps directly to x264 options
[18:43] <evil_core> burek: but I dont have any profiles in x264 package also
[18:43] <burek> type x264 --help
[18:45] <evil_core> maybe should I use x264 directly? (ffmpeg supports much more formats?)
[18:45] <burek> im just saying that x264 does have its own profiles
[18:45] <burek> with ffmpeg's -profile you just choose one of those
[18:46] <burek> baseline,main,high,high10,high422,high444
[18:46] <evil_core> burek: but I see that x264 suypports --pass option, etc
[18:46] <burek> x264 --help | grep pass
[18:47] <burek> btw, whats the purpose of this command <evil_core> ffmpeg -y -i "$iFile" -an -vcodec libx264 -vpre medium -b 800k -threads "$threads" -f mp4 /dev/null
[18:47] <sereal-work> burek, -n would be assume no (where -y is assume yes)
[18:47] <burek> sereal-work, ffmpeg by default assumes no
[18:47] <Mavrik> evil_core, profiles were moved to x264 itself now, no files needed
[18:48] <Mavrik> evil_core, that's also why options changed to "-preset medium" "-profile high" or w.e.
[18:48] <evil_core> burek: to convert any movie file to x264
[18:48] <burek> but /dev/null ?
[18:49] <evil_core> burek: for two pass encoding
[18:49] <evil_core> 1st output were used only to calculate bitrate
[18:49] <burek> but you didnt save that info
[18:49] <burek> where is your passlog file
[18:49] <evil_core> probably, not sure, in the past I used mencoder many years ago, but it was for DivX
[18:50] <evil_core> x264 will copy soundstream also?
[18:50] <burek> no
[18:50] <burek> i dont think so
[18:50] <Mavrik> evil_core, also, there's no need for _firstpass presets anymore, the encoder now detects the pass and automatically selects "firstpass" settings
[18:50] <burek> evil_core http://itbroadcastanddigitalcinema.com/ffmpeg_howto.html#Encoding_H.264_Long_GOP
[18:52] <burek> or using a pass log file, like here: http://forum.videohelp.com/threads/336447-FFMPEG-2-pass-syntax-help-needed
[18:53] <burek> anyway, Mavrik, is it now possible all with 1 command
[18:53] <burek> since x264 now gets ffmpeg's -pass option forwarded
[18:53] <Mavrik> what's possible with 1 command?
[18:53] <burek> 2 pass encoding with x264
[18:53] <burek> x264 --pass 2 --bitrate 1000 -o <output> <input>
[18:53] <burek> thats how you would do it with x264
[18:54] <evil_core> but theres bitrate not automatically calculated
[18:54] <burek> is it possible with ffmpeg to type, for example: ffmpeg -i input -pass 2 -vcodec libx264 output
[18:54] <Mavrik> which version?
[18:54] <burek> latest
[18:54] <evil_core> burek: - 2: Last pass, does not overwrite stats file
[18:55] <evil_core> burek: this are similar opts to ffmpeg, and it looks like you need to call pass 1 and later pass 2
[18:55] <burek> oh i see
[18:55] <tchmiel> hey all
[18:55] <burek> makes sense
[18:55] <Mavrik> burek, em, it definately doesn't work with 0.8.6
[18:55] <Mavrik> you have to do two passes separate
[18:56] <burek> yes yes, 1st creates stats, 2nd creates usable output
[18:56] <burek> just you have to change -pass 1 to -pass 2
[18:56] <burek> :beer: :)
[18:57] <burek> now, i've found somewhere on the google :) how to prefix the input file name with some prefix and set it as an output for ffmpeg, but ive closed that page..
[18:57] <tchmiel> what could "[h264 @ 0x17597d0]missing picture in access unit" mean? I'm streaming from an rtsp feed from a camera to an ffserver on another machine, and just after this message ffmpeg exits.
[18:58] <burek> is it the latest git ffmpeg
[18:58] <evil_core> should I do 1 or 2pass if I dont have target file size?
[18:58] <tchmiel> and it happends a little less than a minute after the stream starts
[18:58] <tchmiel> burek: erm... no, it isn't :/
[18:58] <burek> evil_core, use -vb <bitspersecond>
[18:58] <burek> to calc output file size
[18:58] <burek> or use -crf X if you want constant quality, regardless of file size
[18:59] <burek> i would suggest the latter, because you never know what your input will be
[18:59] <burek> so its not that wise to target the same file size for various inputs
[18:59] <burek> tchmiel, compile the latest
[18:59] <burek> maybe its a bug thats been fixed
[19:00] <tchmiel> burek: will try, thanks
[19:02] <sereal-work> burek, sorryb about the late reply, it will prompt me for y/n if it wants to over write a file.
[19:02] <sereal-work> I want to always do no for overwriting a file.
[19:04] <burek> stat file && ffmpeg ...
[19:04] <burek> ?
[19:04] <sereal-work> so there isn't a way to just do 'no' with ffmpeg?
[19:04] <sereal-work> i'm just gonna be lazy and do -y until someone complains
[19:05] <burek> you can always set output to `date ...`
[19:05] <burek> and always be sure there is no such file
[19:05] <sereal-work> (it's a script that takes a folder of images and uses mencoder to create a time lapse avi, then uses ffmpeg to convert that to ogg and webm)
[19:06] <sereal-work> I use mencoder since the naming scheme with ffmpeg is kinda annoying.
[19:07] <sereal-work> in some respect it's good that it reencodes the videos since the images *could* change.
[19:21] <evil_core> thx all for help
[19:21] <evil_core> bye
[19:30] <tchmiel> the latest ffmpeg from git gives me a segfault :(
[20:06] <burek> tchmiel, what command
[20:06] <burek> pastebin?
[20:13] <tchmiel> burek: hold on, I'm trying to recompile with the "correct" configuration...
[20:19] <kvnn> I'm using the same command to encode two different flv's to h264 base for mobile. One works for iPhone, and the other doesn't. Does anyone know how I can proceed? I'm confused.
[20:42] <pasteeater> kvnn: use a pastebin site to show your ffmpeg command(s) and the complete terminal output(s)
[20:47] <kvnn> pasteeater : I just read "I found out that if your .mp4 file is larger in dimension than 640*360 then the iPhone (iPad, iPod) won't even give the user the option to attempt to play it." and am testing against that
[20:48] <pasteeater> the outputs will still be useful so we can possibly make suggestions for your commands
[20:57] <tchmiel> burek: I re-compiled the most recent version from git with good configuration options, a I keep getting a segfault
[20:57] <tchmiel> pastebin coming up
[21:00] <tchmiel> burek: http://pastebin.com/cDhpbssm
[21:01] <burek> wow..
[21:01] <burek> tchmiel
[21:02] <burek> can you please read this http://ffmpeg.org/bugreports.html
[21:02] <burek> and try to run gdb with trace
[21:02] <burek> and report the results as a bug report
[21:02] <burek> this is obviously a bug..
[21:02] <burek> :S
[21:03] <burek> did you ./configure your ffmpeg with --enable-debug
[21:03] <tchmiel> nope, will have to recompile
[21:03] <burek> thanks
[21:04] <tchmiel> I just need to complete a working solution first, tonight. I'm on a tight schedule :(
[21:04] <tchmiel> will try one of the latest tags from the git repository first, see if I can get something working
[21:04] <tchmiel> then I'll submit the bug, I promise ;)
[21:10] <pasteeater> tchmiel: did this command work with a previous ffmpeg version?
[21:11] <tchmiel> pasteeater: yes, well almost... it sent the stream for a minutes, then said "[h264 @ [...]]missing picture in access unit" and stopped
[21:19] <pasteeater> try this to see if the issue is with decoding http://pastebin.com/EJJWYJhB
[21:22] <tchmiel> mh... no segfault this time, it just keeps turning and turning
[21:31] <pasteeater> you can mention that in the bug report if you decide to do that
[21:31] <pasteeater> you may also try a file output, like output.mpg
[21:50] <tchmiel> well, with the 0.8.6 release, if I add -vsync 2, it seems to be working.
[21:51] <tchmiel> I just can't seem to make it use the -r and -b options, the output bitrate and fps are completely off... :(
[23:41] <teratorn> I've got a really weird problem with doing a YUV overlay by hand... I overlay the image just fine, but when I play back the movie the watermark "flashes"... the luminocity cycles from dark to bright and back again. AFAIK my YUV watermark data is not changing, and the background area of the video I'm overlay on top of isn't changing much either
[23:42] <teratorn> if you look at "out.mpg" here, you can see what I mean: http://teratorn.org/code/mpg_overlay/
[23:42] <teratorn> the code and input files are all there too, if anyone would like to have a poke at it...
[00:00] --- Thu Nov 10 2011
More information about the Ffmpeg-devel-irc
mailing list