[Ffmpeg-devel-irc] ffmpeg.log.20120716

burek burek021 at gmail.com
Tue Jul 17 02:05:02 CEST 2012


[00:06] <tdj-br> arpu: Thanks, -vno worked... now i can't listen audio on rtp, what i did is:
[00:07] <tdj-br> publisher: ffmpeg -fflags +genpts -i c:\temp2\bla\sing.mp3 -vn -threads 0 -f rtp rtp://127.0.0.1:54546 > output.sdp
[00:07] <tdj-br> player: ffplay.exe -i config.sdp
[00:08] <tdj-br> anybody have an idea how to get it working?
[00:13] <Slackyman> bye
[00:14] <tdj-br> is there any option to make ffmpeg stream at original file timming?
[00:14] <tdj-br> i mean, if i have a file with 1 minute, i want that the stream to RTP take 1 minute.
[00:15] <tdj-br> -re
[00:15] <tdj-br> :D
[00:59] <tdj-br> Is there any way to specify number of samples in each packet?
[01:01] <tdj-br> the audio frame size
[02:30] <tdj-br> any specialist on pcm_mulaw?
[02:30] <tdj-br> i am streaming from ffmpeg to x-lite
[02:30] <tdj-br> and audio is modified (stronger voice)
[04:14] <juanmabc> hi, so i keep getting: [swscaler @ 0x8a84ee0] Warning: data is not aligned! This can lead to a speedloss
[04:15] <juanmabc> but what i think is my data is allocated with av_malloc so, what gives?
[04:16] <juanmabc> there is some indirection: this is the sws_scale call: sws_scale(media->video.sc, (const uint8_t * const *) media->video.frm->data, media->video.frm->linesize, 0, media->video.cc->height, media->video.convfrm->data, media->video.convfrm->linesize);
[04:17] <juanmabc> convfrm = avcodec_alloc_frame(); buffer = av_malloc(...); avpicture_fill((AV_Picture*) convfrm, buffer, ...);
[04:17] <juanmabc> that's all the info, i guess
[04:17] <juanmabc> all plays nicely, apart from the speedloss
[04:18] <juanmabc> as i thin, avpicture_fill puts buffer in convrm. and buffer is av_malloc so it should be aligned, shouldn't it?
[04:18] <juanmabc> but still the message says convfrm.data is not aligned?
[04:19] <juanmabc> thanks
[04:22] <juanmabc> since it is not SDL based (OpenGL) ffplay.c is of no help
[05:03] <juanmabc> could it be a bug as pointed by: http://ffmpeg.org/pipermail/libav-user/2012-January/001149.html ?
[05:20] <juanmabc> bye for today
[05:21] <juanmabc> i keep the question open :D
[06:59] <lake> i have an mpeg that i am editing (home video) and there are long blue frames at certain times throughout the video.
[06:59] <lake> what can i do to automatically slice/remove the blue frames from the tape?
[07:03] <lake> can i ask each frame for an average color?
[07:03] <lake> and then for ones that match the blue, output a timestamp to a file?
[07:05] <lake> maybe silence detection is better
[07:22] <zap0> unpack the video into static image files...  .PNGs     then just look at the frames in your OS' file browser  thumb view.
[07:23] <zap0> lake, delete all teh blue.    then re-encode the PNGs into a movie.
[07:24] <lake> zap0: what about the audio?
[07:24] <lake> zap0: interesting idea
[07:26] <zap0> aaahh...  never thought of that bit.  that make this plan useless.
[07:26] <zap0> lake,  i often work without audio; hence i forget to consider it.
[07:27] <zap0> how many spans of blue are you having to deal with?
[07:27] <zap0> surely a video editor would be the quickest approach?
[07:28] <lake> zap0: just a couple per video, between 2 and 5 probably
[07:28] <lake> i want to avoid opening an editor
[07:28] <lake> lol
[07:30] <zap0> editor would be _way_ faster than any other method
[07:33] <lake> zap0: i am trying to automate the process
[07:33] <lake> detecting the silence proves to be working very well, however, with everything, there are edge cases.
[07:34] <lake> silence filter gives silence start, silence end, and elapsed time between start and end. that is extremely useful in CLI editing. so i'm quite happy now.
[07:36] <zap0> you confident any content isn't silence?
[07:37] <lake> not fully, for instance, there are some camera made fade-outs which silence the audio during the fade.
[07:37] <lake> however, it is much easier for me to verify by hand now.
[07:38] <lake> just finding the scene changes took a lot of time before. now i run the silence filter and it tells me exactly what the timestamps are. so cool!
[07:38] <lake> even if i do have to verify, it took the hardest part out of the process.
[07:38] <lake> it would still be cool to have a bluefilter similar to a blackfilter.
[07:51] <WannaBeGeekster> Maybe write your own filter to search for silence and a plain blue screen?  http://wiki.multimedia.cx/index.php?title=FFmpeg_filter_HOWTO
[07:51] <WannaBeGeekster> If the code is Open you should be able to modify the blackfilter code pretty easily to do your blue instead.  I would almost say that it would be cool to be able to provide an image of the frames  you want to remove and have it remove any frame that matched the image.
[07:55] <WannaBeGeekster> Maybe you should check out AVISynth instead?  Looks like there are several filters available there also.
[07:55] <lake> WannaBeGeekster: that would be great. i am reading over the black filter now.
[07:58] <lake> WannaBeGeekster: i imagine i can compose filters on one command.
[07:59] <WannaBeGeekster> lake: combine or compose?
[07:59] <lake> so i would only need to find/create a blue filter
[07:59] <lake> sorry, combine
[07:59] <WannaBeGeekster> Yes, I would image that to be the case.  I imagine it would also increase the processing time exponentially some how, but that is probably not one of your concerns here.
[08:00] <lake> exactly.
[08:02] <WannaBeGeekster> Another suggestion might be to figure out a way to process the file first and determine the start and stop times of  the blue frames then actually just cut the frames out using the time stamps you found during the first part of the command.
[08:03] <lake> right silence filter won't be needed if i can find the blue.
[08:03] <lake> good point
[08:04] <WannaBeGeekster> exactly.  I am looking at this here now:  http://avisynth.org/mediawiki/External_filters#Meta-Filters  and they have a cutframes filter that will cut video/audio out.
[08:08] <lake> i wonder if opencv could help
[08:08] <WannaBeGeekster> good question...
[08:08] <WannaBeGeekster> Will be curious to know what you figure out that is for sure
[08:09] <lake> my gut tells me to create a blueframe filter, but reading that code, even at 170 some lines, is cryptic
[08:10] <WannaBeGeekster> lol
[08:10] <WannaBeGeekster> where did you find the code at?
[08:10] <lake> i cloned the git repo
[08:11] <lake> one sec
[08:13] <lake> WannaBeGeekster:http://ffmpeg.org/doxygen/trunk/vf__blackframe_8c-source.html
[08:13] <WannaBeGeekster> k, let me see if I can understand any of it.
[08:14] <lake> i'd like to know where the color definition is, lol
[08:19] <WannaBeGeekster> I am thinking some where near lines 60 - 64.  But I think there are definitions that are definied in some of the included libraries.  Lines 66 - 75 are also very important.  Wondering about 255 for the bthresh and what that has to do with it.
[08:19] <WannaBeGeekster> The sscanf looks like it could be important to find out what it does honestly.
[08:25] <WannaBeGeekster> Ahh, so I believe it is using the libyuv filter component.  Which takes U as a color value.  Not sure what YUV stand for (I am still new at video honestly).
[08:26] <WannaBeGeekster> No I think I am wrong.  Trying to figure out what the %u is in this command here:  sscanf(args, "%u:%u", &blackframe->bamount, &blackframe->bthresh);
[08:27] <WannaBeGeekster> sscanf is the function to scan a frame.  So I think that is getting close to the right place here.
[08:29] <WannaBeGeekster> Ahh, so what it is doing is formatting the values for bamount and bthresh and providing those as arguments to the sscanf function.
[08:34] <lake> WannaBeGeekster: interesting
[08:42] <WannaBeGeekster> lake: I honestly think that the sscanf is where you need to spend your resources.  Figuring out what it is really passing back to the sscanf function.  I think that it is formatting the two numbers to fit the criteria in this here:  http://libav.org/libavfilter.html#lut_002c-lutrgb_002c-lutyuv
[08:42] <WannaBeGeekster> The format of the data will depend on the input video though, if it is YUV or RGB.
[08:49] <lake> WannaBeGeekster: okay, thanks for your input. :)
[08:49] <lake> i have to go to sleep now.
[08:49] <lake> hope to talk with you again soon
[08:51] <WannaBeGeekster> good luck!
[10:21] <brocatz>  does anyone know how i tell my usb web cam to stream to me in h264? iut has a chip on it for doing it, but i'm not actually sure how to signal it to do it and how to get the data, i assume if i use dshow it'll convert it to some raster format instead
[10:25] <WannaBeGeekster> depends on the camera obviously.  What model number it is?
[10:26] <darkyen> Is it possible that ffmpeg can return NaN as sample ?
[10:28] <AlRazi> if i'm getting the output ( m4a ) to not be playable on iTunes, do i use l-SMASH muxer, or demuxer
[10:28] <AlRazi> i tried qt-faststart to no avail
[10:29] <iTux> Hi
[10:32] <iTux> I need some help, I would like to sub and convert some files into webm format. The problem is about my subtitles, in ass format; the sub operation is correct but the final file is very pixelized :/
[10:32] <darkyen> ass format . sounds intresting
[10:32] <AlRazi> i'll never grow up
[10:32] <AlRazi> if i'm getting the output ( m4a ) to not be playable on iTunes, do i use l-SMASH muxer, or remuxer
[10:33] <iTux> I use this script : http://pastebin.com/rsQLcuPc
[10:34] <iTux> There is no error while encoding
[10:35] <brocatz> WannaBeGeekster: Logitech C920
[10:36] <burek> iTux, most probably your pixelization comes from combining drawings (fonts) on video, which libvpx handles poorly, imho
[10:36] <burek> try tweaking encoding options a little bit more
[10:36] <burek> also, try -threads 0
[10:36] <burek> let ffmpeg autodetect num of cpus
[10:37] <burek> brocatz, throw it into the garbage
[10:37] <burek> I have that same model
[10:37] <burek> and there is no decent support for it
[10:37] <burek> it took a month just for their support to reply to my email
[10:38] <brocatz> i'd really like to be able to capture decent video
[10:38] <burek> even on windows, their driver doesn't work as advertised
[10:38] <iTux> burek : i'm a beginner whith ffmepg, I need to change "-q" parameter ?
[10:38] <burek> iTux, google for libvpx encoding options
[10:38] <burek> or read here
[10:38] <iTux> Thank you :)
[10:39] <burek> :beer: :)
[10:41] <burek> brocatz, try ffmpeg -f v4l2 -input_format h264 -i /dev/video0 -c copy out.mp4
[10:45] <brocatz> v4l2? video for linux 2?
[10:45] <brocatz> i'm using windows 7 fwiw
[10:46] <burek> good luck then :D
[10:48] <brocatz> yeah input_format is unrecognised
[11:02] <iTux> Hum, do the -map 0:x can be found attachment files in a mkv container ?
[11:02] <iTux> -be
[12:14] <nander> Hi
[12:14] <nander> How do I convert all AVI files in a folder to a different codec?
[12:14] <nander> I have them in 1080p but for some reason they are 6.4 GB for 2 minutes of film
[12:15] <nander> Which craches my video editing software
[13:02] <NuxRo> hi. i want to extend an mp3 with a couple of seconds of "silence"; any ideas how to do it?
[13:58] <AlRazi> if i'm getting the output ( m4a ) to not be playable on iTunes, do i use l-SMASH binaries to fix it ?
[14:02] <AlRazi> i'm almost there people please help, is there an arrangement that involves remuxer and qt-faststart to fix a m4a unplayable file ?
[14:05] <thaman> why use aac? mp3 is much better
[14:05] <iTux> Can ffmpeg take an ass track in a mkv (soft sub) and hardsub it into a WebM ?
[14:05] <iTux> Whit -map argument
[14:07] <sacarasc> thaman: Except it's not...
[14:09] <thaman> i coulda sworn.
[14:10] <thaman> i always associate m4a with apple and well, itunes sucks. ergo m4a suck
[14:29] <iTux> There is a thing that I don't understand, I use "-qmin 10" parameter, but, while encoding, the terminal print "q=0.0". Is that normal ?
[15:40] <sgfgdf> hello, guys! how can i found the container of a video if the file doesn't have a extension?
[15:40] <WannaBeGeekster> Well, just run ffmpeg -i filenamewithoutextension
[15:40] <Mavrik> ffprobe rather :)
[15:40] <WannaBeGeekster> or you can try file filewithoutextension
[15:40] <AnywhereIs> file rather
[15:42] <sgfgdf> file myfile shows this -- "ISO Media, MPEG v4 system, 3GPP". so should i name it .mp4 or .3gp?
[15:42] <sgfgdf> aren't they different containers?
[15:53] <JEEBsv> sgfgdf: they all base onto the ISO media container specification
[15:53] <JEEBsv> they're quite similar
[15:53] <JEEBsv> could check the brands of the file, and decide the file name with that
[15:56] <sgfgdf> JEEBsv, what do you mean by "check the brands of the file"? how can i do it?
[15:56] <JEEBsv> I think ffmpeg showed those?
[15:57] <sgfgdf> JEEBsv,  ah, if you mean this -- "major_brand     : 3gp4", "minor_version   : 768". i still can't figure out what to choose. or probably as you told there should not be any difference.
[15:58] <WannaBeGeekster> I want to convert some videos to 480p.  The current sizes will vary.  Is there a right way to do this without causing weird issues with resizing the video?
[15:58] <JEEBsv> sgfgdf: I guess you'll want to name the file dot-3gp then I guess
[15:59] <JEEBsv> on the big picture it doesn't really matter, as the file formats are similar (based on ISO media container), but in this case I guess 3gp would be "closer"
[16:15] <sgfgdf> JEEBsv, thank you for the help!
[17:19] <burek> WannaBeGeekster, did you try scale video filter
[17:25] <burek> you can use: ffmpeg -i input -vf scale=-1:480 output
[17:33] <WannaBeGeekster> burek:  Nope sure didn't!  Thanks for that.  That is not anything close to what I have found so far.
[17:33] <WannaBeGeekster> It is all in knowing the right words to search for.. lol
[17:33] <burek> :)
[17:39] <zambo> has anyone been able to get filters/concat to work when compiling for Android?
[17:49] <burek> zambo, what exactly does not work for you
[18:12] <tomx> hi
[18:13] <tomx> i have a big problem encoding a video with TWO video streams in it
[18:13] <tomx> Stream #0:0(eng): Video: wmv1
[18:13] <tomx> Stream #0:1(eng):
[18:13] <tomx> Stream #0:2(eng): Video: wmv1
[18:13] <tomx> can somebody help me? :)
[18:15] <relaxed> state the problem
[18:16] <tomx> it tries to map the stream 0:2 to output 0:0 and then throws a Error while opening encoder for output stream #0:0 -
[18:18] <tomx> i can ofc use the -map command but im working on a generalized solution for all kinds of input movies, so -map might not be applicable here
[18:20] <relaxed> You want all the stream present in the output?
[18:21] <relaxed> streams*
[18:23] <tomx> yes
[18:23] <relaxed> Did you try "-map 0"?
[18:24] <tomx> sec!
[18:24] <tomx> Stream mapping:
[18:24] <tomx>   Stream #0:0 -> #0:0 (wmv1 -> libx264)
[18:24] <tomx>   Stream #0:2 -> #0:1 (wmv1 -> libx264)
[18:24] <tomx> looks good!
[18:25] <tomx> but, i dont get any progress feedback
[18:26] <tomx> give me another min, ill investigate :)
[18:26] <relaxed> Maybe because it's encoding both streams in parallel?
[18:28] <tomx> the frame= counter goes up, but also does the dup(licate)= counter
[18:29] <tomx> lets see how this turns out - waiting for encode to finish :)
[18:35] <tomx> yeah! works!
[20:06] <heimlich> anyone know if ffmpeg could be used to create anything like this: http://anotherplotdevice.com/2011/05/06/moviebarcode/
[20:07] <djdingo14> ffmpeging from MOV to AVI causes a hard-to-describe issue when importing the resulting file to Adobe Premiere Pro CS6.
[20:07] <djdingo14> Basically they all become in the wrong (way too transparent) opacity.
[20:21] <lake> heimlich: interesting idea
[20:25] <relaxed> djdingo14: does premiere support any lossless codecs?
[20:26] <djdingo14> relaxed: I am not sure. I always run my ffmpeg command on the files before importing.
[20:26] <djdingo14> Except when it comes to MOV.
[20:27] <lake> heimlich: i'm having trouble conceptualizing how they are acheiving that
[20:30] <relaxed> djdingo14: what's your command?
[20:30] <heimlich> well.. i think it could be done, by taking a movie, splitting it into frames, like lets say every 10 frames, scaling the image to a 1 pixel width, and compositing it onto a HD frame.. then, take the next 10th frame, do the same, offset it by one, etc...i guess, the more I think about it, it seems actually pretty easy to do
[20:31] <lake> i'm a massive noobie w/ ffmpeg but seems doable.
[20:33] <heimlich> seems like a multipstep process would be easy... 1) extract n number of frames from a movie  2) take each of those frames, and composite them onto a black frame by resizing them to a 1 pixel or 2 pixel line, and offsetting them on the large frame until all are composited..
[20:34] <lake> heimlich: i wonder if imagemagick would be suitable to use
[20:35] <heimlich> yeah.. for part 2
[20:37] <lake> heimlich: are you having a go at it?
[20:38] <lake> i would but i don't have free time for it now.
[20:38] <heimlich> yeah. should be quick
[20:38] <lake> nice! will you pastie it when you finish? i'd love to try
[20:38] <lake> and learn
[20:41] <heimlich> sure.. no prob
[20:48] <relaxed> ffmpeg should be able to do it all
[20:50] <heimlich> hmm.. really ?
[20:51] <relaxed> yes, using the pad and overlay filters
[20:51] <heimlich> I'll take a look at that.. thanks for the suggestion... its more out of curiosity than anything else..
[20:54] <relaxed> pad to create the canvas; overlay/movie filter to paint each row
[20:55] <bneff> Can I use ffmpeg to report errors with an RTSP/H.264 source?  The video displays macroblocks with movement, but I'm not sure what the problem is...
[20:58] <shahriman0> guys I just can't seem to get concat working
[20:59] <shahriman0> for example, ffmpeg -i concat:/home/shakkhar/black_24frames.y4m\|/home/shakkhar/input2.mov ~/foo%03d.png
[20:59] <shahriman0> outputs only the first file, not the second
[21:20] <djdingo14> relaxed: Just ffmpeg -i %1 -sameq %1.avi
[21:41] <axisys> how to I convert a .org file to a format that windows can play.. i used recordmydesktop to record some process .. but the people who will watch it unfortunately uses windows
[21:48] <djdingo14> .org?
[21:48] <djdingo14> Maybe you mean .ogg...
[21:48] <axisys> .ogv .. sorry
[22:00] <gapan> Hi. I'm trying to compile ffmpeg 0.11.1 and I get an error "libxvid not found"
[22:00] <gapan> but xvidcore is installed
[22:01] <lake> heimlich: any progress? stuck?
[22:01] <gapan>  /usr/include/xvid.h and /usr/lib64/libxvidcore.so.4.3 are there
[22:01] <gapan> that error comes right after running configure, no progress at all
[22:01] <lake> since there may be more eyes now, does anyone have experience detecting a blueframe, similiar to a blackframe?
[22:03] <lake> i want to timestamp where a blueframe starts and ends during a video containting audio. i have had some luck by using a silence filter, because the blue screen also has silence. however, that has some false positives.
[22:03] <gapan> configure line is here: http://pastebin.com/UGYjVsgm
[22:04] <lake> axisys: can you convert it to flash?
[22:08] <lake> would it be possible to detect the average frame color?
[22:13] <axisys> lake: i am not sure how
[22:16] <lake> axisys: ffmpeg -i foo.ogv bar.flv
[22:17] <axisys> lake: ok
[22:20] <axisys> lake: flash becomes little bigger than avi
[22:22] <axisys> foo.ogv is 23M, foo.flv is 11M .. i guess that is fine
[22:22] <axisys> video is not that clear tho
[22:23] <lake> axisys: what are the dimensions of your ogv
[22:24] <lake> you could also try "bar.wmv"
[22:24] <lake> wmv should be playable on windoze
[22:25] <axisys> [ogg @ 0x7fbe5dfe7940]max_analyze_duration reached
[22:25] <axisys> [lavf] stream 1: video (theora), -vid 0
[22:25] <axisys> VIDEO:  [theo]  1216x768  0bpp  15.000 fps    0.0 kbps ( 0.0 kbyte/s)
[22:25] <axisys> Audio: no sound
[22:25] <axisys> Starting playback...
[22:25] <axisys> Movie-Aspect is 1.58:1 - prescaling to correct movie aspect.
[22:25] <axisys> VO: [xv] 1216x768 => 1216x768 Planar YV12
[22:30] <gapan> fixed, it's an issue with xvidcore, it doesn't create the appropriate symlinks during make install
[22:34] <lake> axisys: ffmpeg -i foo.ogv -s 1216:768 bar.flv
[22:54] <lake> axisys: any luck?
[00:00] --- Tue Jul 17 2012


More information about the Ffmpeg-devel-irc mailing list