[Ffmpeg-devel-irc] ffmpeg.log.20131015

burek burek021 at gmail.com
Wed Oct 16 02:05:01 CEST 2013


[00:26] <AndrzejL> so I am looking at this:
[00:26] <AndrzejL> ffmpeg -i "mms://193.164.132.195:2124/aguizzardi" -vcodec libx264 -vpre default -vpre baseline -g 60 -vb 150000 -strict experimental -acodec aac -ab 96000 -ar 48000 -ac 2 -vbsf h264_mp4toannexb -f mpegts udp://127.0.0.1:10000?pkt_size=1316
[00:26] <AndrzejL> this looks like exactly what I need ;) some parameters would have to be changed and I would prefer the xvid codec rather then x264 - all doable
[00:26] <MarcWeber> hls output, then uploading to a server causes some stalls. Is there a way to tell the client to start about 5 segments before the last one?
[00:37] <buhman> AndrzejL: magic
[00:37] <buhman> AndrzejL: told you so
[00:43] <AndrzejL> :)
[00:51] <AndrzejL> how do I access such stream as lets say udp://127.0.0.1:5999/Stream1.ffm
[00:51] <AndrzejL> from lets say mplayer or vlc?
[00:51] <AndrzejL> I want to test it on local network before I will open ports for wan
[00:56] <wald0> what means -crf 0 ? how i can set the quality/size for libx264 vcodec ?
[00:59] <llogan> wald0: https://trac.ffmpeg.org/wiki/x264EncodingGuide
[01:08] <wald0> thx
[01:11] <wald0> when i record my desktop, the colors of my terminal looks "poor", seems like by recording chars of one-pixel width is not an easy thing, i have read about "subpixel sampling" or a similar name but i dont understand how this works, anybody knows what i need to do for record perfectly those colors?
[01:12] <wald0> so i want to create video tutorials of desktops and i want to be them look perfect before the video-editing step
[01:12] <llogan> thats an effect from RGB to YUV conversion
[01:14] <wald0> mmh, so how i can record in yuv ?
[01:14] <wald0> if this is the answer :)
[01:18] <llogan> i'm not sure what you can do about it
[01:24] <wald0> llogan: you mean that we cannot record videos in yuv format ?
[01:24] <wald0> oh wait
[01:24] <wald0> i meant rgb maybe
[01:27] <llogan> you can (libx264rgb, qtrle, among others, IIRC), but upon playback it will probably be converted to yuv
[01:28] <llogan> huffyuv too
[01:30] <wald0> if is saved correctly is enough for now, so i can edit it correctly in the ML
[01:30] <wald0> NLE
[01:31] <wald0> later will be converted to any other format, but having got resizes and stuff which could make it look better
[01:31] <MorehouseJ09> 	st = avformat_new_stream(context, codec);	
[01:31] <MorehouseJ09> 	c = st->codec;
[01:31] <MorehouseJ09> 	avcodec_get_context_defaults3(c, codec);
[01:31] <MorehouseJ09> 	printf("%p" "\n", c->codec);
[01:31] <MorehouseJ09> 	printf("%p" "\n", codec);
[01:31] <MorehouseJ09> hey, sorry for the weird paste. I'm trying to figure out why when I create a stream, its not linking up the codec properly?
[01:32] <MorehouseJ09> for instance, c->codec returns 0x0
[01:34] <llogan> wald0: i forgot that by default with libx264 the pixel_fmt will be yuv444p, thus avoiding the loss (I think), but it looks shitty because it's being converted to 420 in your player
[01:34] <llogan> (...by default when your input is RGB...)
[01:35] <llogan> i can't remember this stuff.
[08:46] <t4nk426> hi
[08:46] <t4nk426> ffmpeg does not spilt video data to multiple file every segment_time
[08:46] <t4nk426> the command :
[08:46] <t4nk426> ffmpeg -loglevel debug -sn -f video4linux2 -r 30 -s 1280x720 -input_format h264 -i /dev/video1 -vcodec copy http://localhost:23791/feed2.ffm -vcodec copy -map 0 -flags -global_header -f ssegment -segment_time 4 -segment_format mpegts /var/stream%05d.ts
[08:47] <t4nk426> the option "-segment_time 4" does not function
[08:47] <t4nk426> because when I test for about 1 minute, there is file only
[08:47] <t4nk426> does anyone have idea?
[08:47] <t4nk426> because when I test for about 1 minute, there is 1 file only
[08:48] <t4nk426> ffmpeg does not split video data to multiple file by each 4 seconds
[10:12] <Apic> A wonderful spendid Bureaucracy the 69th in the YOLD 3179!
[10:17] <dol> hi. here is my scaling call: sws_scale(convertCtxDec, decodedpic->data, decodedpic->linesize, 0, aHeight, avFrameRGB->data, avFrameRGB->linesize);
[10:17] <Mavrik> mornin everyone
[10:17] <dol> I see a problem in some resolutions that the image is a bit cropped from the right side
[10:17] <dol> I got a comment that I should use viewable width instead of linesize
[10:17] <dol> but I coulndn't succeed
[10:18] <dol> could anyone tell me how I should do it?
[10:18] <dol> btw: this call converts from yuv 420 to rgb32
[10:18] <Mavrik> hmm
[10:18] <Mavrik> you should certanly use linesize
[10:19] <dol> then where is the problem?
[10:19] <Mavrik> now that's a good question, that code does look correct
[10:20] <dol> this code works good for known resolutions
[10:20] <Mavrik> the problem might be in initialization of your sws context or in the player
[10:20] <dol> you mean convertCtxDec?
[10:21] <Mavrik> mhm
[10:22] <dol> I use the follwoing to set it: sws_getContext(aWidth, aHeight, PIX_FMT_YUV420P, aWidth, aHeight, PIX_FMT_RGB32,SWS_SPLINE, NULL, NULL, NULL);
[10:22] <dol> perhaps this is not correct?
[10:23] <Mavrik> well
[10:23] <Mavrik> only if your output image is the same size as the input
[10:24] <dol> by input, you mean the size of the decoded frame ?
[10:24] <Mavrik> the input into sws_scale
[10:24] <Mavrik> that's all it matters :)
[10:25] <dol> i think there is the mistake as you have pointed
[10:25] <dol> then I should set the source width after decoding every frame
[10:26] <Mavrik> basically, your context needs to have correctly set frame sizes for input and output of sws_scale call
[10:27] <Mavrik> if your input size changes, you need to call sws_getCachedContext() before each sws_scale
[10:27] <Mavrik> note that height/width and linesize might not match always :)
[10:27] <cauffe> Hey all, I seem to be having some formatting issues. ffmpeg -f mp4 -i sj.mp4 -ss 871.13 -t 1 sj2.mp4 . Seems to be outputting a invalid video file?
[10:27] <cauffe> is -f mp4 a valid param?
[10:28] <Mavrik> mhm
[10:28] <Mavrik> cauffe, why are you reencoding whole video to just cut it? :)
[10:28] <cauffe> trying to cut a segment :)
[10:28] <Mavrik> yeah
[10:29] <Mavrik> why are you reencoding the video? without encoding parameters? that probably destroys your quality :)
[10:29] <cauffe> When I don't put the -f pram in I get [aac @ 0x7fd40b81cc00] The encoder 'aac' is experimental but experimental codecs are not enabled, add '-strict -2' if you want to use it.
[10:30] <cauffe> then i get an invalid video file again
[10:30] <Mavrik> ywah, because you're reencoding everything
[10:30] <Mavrik> instead of just copying streams
[10:30] <Mavrik> use -codec copy
[10:31] <cauffe> so ffmpeg -codec copy -i ?
[10:31] <Mavrik> em no
[10:31] <Mavrik> after -i
[10:31] <dol> Mavrik, I moved sws_getContext call before sws_scale and changed it to the following: sws_getContext(decodedpic->width, decodedpic->height, PIX_FMT_YUV420P, aWidth, aHeight, PIX_FMT_RGB32, SWS_FAST_BILINEAR, NULL, NULL, NULL);
[10:32] <dol> but everytime decodedpic->width = aWidth and the same for height
[10:32] <dol> so it didn't solve my problem
[10:32] <cauffe> ffmpeg -i -codec copy gave me "-codec: No such file or directory "
[10:33] <Mavrik> cauffe, just how ancient is your ffmpeg?
[10:33] <cauffe> just downloaded it
[10:33] <Mavrik> from where?
[10:34] <cauffe> ~_~ http://ffmpegmac.net/
[10:34] <Mavrik> uh
[10:34] <Mavrik> your error message makes no sense then
[10:34] <Mavrik> what did you type in EXACTLY?
[10:35] <cauffe> ffmpeg -i -codec copy sj.mp4 -ss 871.13 -t 1 sj2.mp4
[10:35] <Mavrik> ah
[10:35] <Mavrik> so you just basically told ffmpeg that input filename is "-codec" ;)
[10:35] <cauffe> >_<
[10:36] <cauffe> its supposed to be after the I though, right?
[10:37] <cauffe> if I put it before the -i then I get "Unknown decoder 'copy'
[10:39] <cauffe> sorry for being such a n00b
[10:39] <cauffe> just super frustrating
[10:39] <Mavrik> yes, because you haven't really read what you're doing and what the options you're using are doing
[10:40] <Mavrik> just blindly stumbling about
[10:40] <Mavrik> you set OUTPUT options AFTER you tell to ffmpeg what the input is
[10:40] <Mavrik> so "-codec copy" paramter goes after the "-i <input file>" parameter
[10:40] <Mavrik> ffmpeg -i <something> -codec copy -ss <bla> -t <bla> output.mp4
[10:43] <cauffe> lol, thanks, got it to output now but with no video
[10:44] <Mavrik> then your input file is broken.
[10:44] <cauffe> yeahhh
[10:44] <Mavrik> can you paste full output of your command on a pastebin?
[10:46] <bouba> hi all, any ffmpeg guru who know a method to extract lossless thumbs from a video ? it seems like jpeg output is lossy... ffmpeg -y -i ${input} -an -vf fps=fps=1/20 -f image2 ${input}_%05d.jpeg
[10:47] <Mavrik> bouba, any thumbnails will be lossy
[10:47] <Mavrik> bouba, you're missing the quality specifier for jpegs that's why they look ugly
[10:48] <Mavrik> bouba, use -qscale:v 3 or something
[10:48] <Mavrik> 1 being the best, 31 the worst quality
[10:48] <bouba> Mavrik, thx, i'll try this method and google it
[11:16] <vl4kn0> Hi, I have continuous growth of memory usage in my program using ffmpeg and valgrind suspects avcodec_video_decode2 (it's from the internet stream). Are there any particular memory allocations I should be aware of?
[11:20] <Mavrik> vl4kn0, the output frame is allocated by the decoder
[11:43] <vl4kn0> Mavrik: I call av_frame_unref(frame) when it's not longer needed
[11:45] <dol> Mavrik, I got disconnected form the chat
[11:45] <dol> I have placed my question in detail here :http://forum.doom9.org/showthread.php?t=169036
[11:45] <dol> could you please have a look at it?
[11:45] <dol> or anybody else
[11:52] <dol> nobody?
[14:40] <dol> guys can anyone please look at the problem that I described in detail here: http://forum.doom9.org/showthread.php?t=169036
[14:41] <JEEBsv> dol: you might get better replies over at the -devel channel
[14:41] <dol> oh really. thanks. I will go there
[16:58] <fling> Who are encoding videos with black bars? I hate them!
[16:58] <zap0> dat wasist!    it's african american bars :p
[17:01] <fling> zap0: :D
[17:01] <foonix> fling: i guess ppl who hate aspect ratio error
[17:02] <zap0> or those that like to be able to read sub titles
[18:05] <edinny> I am trying to convert an MP3 file to a format that Facebook will let me upload.
[18:05] <edinny> So far, no luck
[18:16] <edinny> I found this online, but it does not seem to make a format FB likes:
[18:16] <edinny> ffmpeg -y -i /path_to_your/song.mp3 -f flv -ab 131072 -ac 1 /path_to_your/video.flv
[18:26] <edinny> I am trying to convert an MP3 file to a format that Facebook will let me upload.
[18:26] <edinny> So far, no luck
[18:28] <_8680_> Using ffprobe from MacPorts, it spews information about how it was built and such: <https://dpaste.de/oXHu>. How can I make it be quiet about that and only print the information about the given file?
[20:18] <MorehouseJ09> I'm trying to demux a clip and am getting an "inconsistent channel configuration" error
[20:18] <MorehouseJ09> Shouldn't the ContextFormat take care of this when you open the file?
[20:35] <edinny> I found this online, but it does not seem to make a format FB likes:
[20:36] <edinny> ffmpeg -y -i /path_to_your/song.mp3 -f flv -ab 131072 -ac 1 /path_to_your/video.flv
[20:36] <edinny> any ideas?
[20:48] <khali> edinny: you may have to create a fake video stream
[20:48] <khali> using .flv for audio only seems weird
[20:49] <edinny> got a suggestion on how to make an mpg or avi or mp4 with a static picture?
[20:49] <edinny> khali?
[20:53] <relaxed> edinny: ffmpeg -loop 1 -i input.jpg -t 00:01:00 output
[20:53] <relaxed> that would give you a minute long video of a static image
[20:53] <edinny> I wanted to use an mp3 as the audio in that
[20:55] <relaxed> edinny: ffmpeg -i input.mp3 -loop 1 -i input.jpg -c:a copy -q:v 3 -shortest output
[20:56] <edinny> Unrecognized option 'c:a'
[20:56] <edinny> Failed to set value 'copy' for option 'c:a'
[20:57] <JEEB> by chance, do you have an old or libav-based ffmpeg?
[20:57] <edinny> I have the default for ubuntu 12.04
[20:57] <JEEB> then you have libav
[20:57] <JEEB> use avconv
[20:58] <JEEB> because the ffmpeg binary is not updated in libav
[20:58] <edinny> can you post the entire line?
[20:58] <JEEB> just switch ffmpeg to avconv
[20:58] <edinny> oh!
[20:58] <JEEB> because that's the updated binary with the project that ubuntu/debian uses
[20:59] <JEEB> (which is not the ffmpeg (project))
[20:59] <edinny> installing...
[21:00] <JEEB> huh
[21:00] <edinny> huh? Says already there, but is says it is not...
[21:00] <edinny> got it I think...
[21:01] <JEEB> anyways, if you are going to use libav-based stuff then use avconv, not ffmpeg. If you ever use ffmpeg-based things, then you use ffmpeg
[21:01] <JEEB> :)
[21:01] <edinny> what do I put for "-shortest output"?
[21:01] <JEEB> -shortest is the option
[21:02] <JEEB> output just means the output file name
[21:02] <edinny> so output.avi?
[21:02] <relaxed> no one needs avi in this day and age
[21:02] <JEEB> basically it's an option without a value
[21:02] <relaxed> use output.mp4
[21:03] <llogan> edinny: if you're using libav stuff, and it looks like you are, then please continue this discussion at #libav
[21:03] <edinny> thanks. Did not know
[21:04] <relaxed> edinny: we can discuss llogan there
[21:05] <relaxed> llogan: did you read my thoughts on the compilation guide?
[21:06] <llogan> i don't think so. when? where?
[21:07] <relaxed> Maybe last week here in this channel. I mentioned it would be helpful if each external lib had ffmpeg ./configure requirements for those that use it as a reference.
[21:07] <edinny> relaxed: Facebook does not like that format for video uploads
[21:08] <llogan> relaxed: can you show me an example? i'm not feeling very smrt today
[21:08] <relaxed> for example, libx264 requires --extra-ldflags="-ldl" and --enable-gpl
[21:09] <relaxed> some require --enable-nonfree, etc
[21:09] <llogan> oh, i see. sure. sounds good to me. would you be willing to add the info?
[21:09] <relaxed> Can anyone edit the wiki?
[21:10] <llogan> yes, once you register (i don't think registration requires a humon)
[21:11] <llogan> ...to approve
[23:33] <Maverick|MSG> given the instructions on this page:
[23:33] <Maverick|MSG> https://trac.ffmpeg.org/wiki/Seeking%20with%20FFmpeg
[23:33] <Maverick|MSG> it says you can cut sections out of a movie by specifying the -to option
[23:34] <Maverick|MSG> however if you specify the -ss option before or after the -i you'll get different results
[23:34] <Maverick|MSG> what happens if you specify the -ss twice, as the top section "Fast and accurate seeking" recommends?
[23:46] <cbsrobot> Maverick|MSG: try it ...
[23:46] <cbsrobot> and in what sense different ?
[23:46] <cbsrobot> maybe pastebin some info ?
[23:47] <Maverick|MSG> I don't have my system for ffmpeg setup here so I can't really try much at the moment
[23:47] <Maverick|MSG> was more trying to just pesudocode things
[23:47] <Maverick|MSG> basically in the "Cutting small sections" it shows two simular command lines
[23:47] <Maverick|MSG> the first, the -ss specified first, will produce a 2 minute video
[23:48] <Maverick|MSG> whereas the second, the -ss after the -i will produce a 1 minute video
[23:49] <cbsrobot> what's the value for -to ?
[23:50] <Maverick|MSG> on the page it is 00:02:00
[23:54] <Maverick|MSG> I guess if I specify both it'll just cut it twice
[00:00] --- Wed Oct 16 2013


More information about the Ffmpeg-devel-irc mailing list