[Ffmpeg-devel-irc] ffmpeg.log.20150728

burek burek021 at gmail.com
Wed Jul 29 02:05:01 CEST 2015

[00:00:44 CEST] <Nolski> explodes: by using that project's code, you are violating US copyright law
[00:01:11 CEST] <explodes> Nolski Firstly, I'm not using it. Secondly, dang.
[00:02:06 CEST] <Nolski> explodes: right, not gonna sue you or anything ;) It's normally a misconception that a lot of people on github make. Normally I create an issue letting them know.
[00:02:23 CEST] <explodes> Nolski Is there any way to use ffmpeg in android legally?
[00:02:45 CEST] <Nolski> android is open source right?
[00:03:12 CEST] <explodes> yea, I don't know enough about the relevant licenses to understand what that implies, though
[00:06:10 CEST] <explodes> One person in ##android-dev mentioned if I was going to use FFMPEG was to leave out the "parts" covered by the license, can I do video playback without having to adhere to the GPL license?
[00:06:35 CEST] <explodes> Really my other options are to go through some other party and pay them to use their video player, which is no fun for me
[00:07:42 CEST] <Nolski> explodes: so the GPL essentially says any code that you use it in, you must keep that code open source.
[00:07:50 CEST] <Nolski> So if your code is open source, then you can use the GPL :)
[00:08:05 CEST] <Nolski> even if your code is MIT, BSD, Apache or whatever.
[00:08:16 CEST] <Nolski> If you modify ffmpeg itself, then you must submit those changes upstream
[00:09:48 CEST] <c_14> No you don't.
[00:10:02 CEST] <explodes> Is FFMpeg entirely GPL?
[00:10:07 CEST] <c_14> You just have to release the modified sources in the same channel as the binaries.
[00:10:16 CEST] <Nolski> ^ right, my bad
[00:10:26 CEST] <c_14> explodes: it is if you use --enable-gpl
[00:10:30 CEST] <c_14> Otherwise it's LGPL
[00:11:10 CEST] <explodes> LGPL still requires the project to be open source, correct?
[00:11:58 CEST] <c_14> If you modify the source you still need to publish the changes to the ffmpeg source under the LGPL, but you can link proprietary code against LGPL code
[00:12:12 CEST] <explodes> ok
[00:12:54 CEST] <explodes> So I can include ffmpeg builds that aren't built with --enable-gpl as long as I do not modify the source of the ffmpeg library
[00:13:50 CEST] <c_14> If you distribute binary versions of ffmpeg you also need to distribute source versions of ffmpeg, but you do not need to release source versions of your program that links against the LGPL ffmpeg
[00:16:11 CEST] <explodes> What is a good way to distribute source versions of ffmpeg?
[00:16:31 CEST] <explodes> I have a github- can I just provide a fork?
[00:17:03 CEST] <c_14> You can usually just put a link to the ffmpeg sources on your download page/ your github readme
[00:17:57 CEST] <c_14> ianal though
[00:25:33 CEST] <Nolski> c_14: think you could take a quick look at this. adding amix ends up breaking the command https://gist.github.com/Nolski/5877cee63b2f37be3fd6
[00:26:09 CEST] <pzich> looks like you have a trailing colon in your filters
[00:26:17 CEST] <pzich> what was the actual error?
[00:26:45 CEST] <Nolski> pzich: refresh, I put the output in a comment
[00:27:00 CEST] <c_14> You probably need to actially label the pads
[00:27:09 CEST] <pzich> like [0:a][1:a]amix...
[00:27:13 CEST] <c_14> ye
[00:28:05 CEST] <pzich> also, it looks like your first video input doesn't have an audio track?
[00:29:11 CEST] <Nolski> pzich: possibly. this is a script so it's overlaying two videos and I want it to overlay the audio as well if there is any (on either video)
[00:29:54 CEST] <pzich> I'm not sure if there's a simple way to write this to only mix audio if both inputs have audio, you may need to pre-check that and branch in your code
[00:30:18 CEST] Action: Nolski sighs
[00:30:28 CEST] <Nolski> I figured that'd be the case.
[00:31:11 CEST] <Nolski> not even sure how to figure out how to check that
[00:31:48 CEST] <pzich> ffprobe or ffmpeg can tell you the streams in an input
[00:31:56 CEST] <Nolski> with -i?
[00:31:57 CEST] <pzich> so you need to parse that and then decide how to filter
[00:32:18 CEST] <pzich> I don't think you use -i for ffprobe, but yes for ffmpeg
[00:33:23 CEST] <Nolski> ah I see
[00:33:51 CEST] <pzich> ffprobe also has some different formatting outputs that may make it easier to parse
[00:35:25 CEST] <pzich> e.g. `ffprobe <movie> -show_streams -of json`
[01:08:51 CEST] <Nolski> pzich: wait I'm a bit confused. ffmpeg is showing that there is a stream for audio but then errors when I try to use it
[01:08:56 CEST] Action: Nolski gets paste ready
[01:09:51 CEST] <Nolski> https://gist.github.com/Nolski/da11224bf650ba0059a5
[01:10:05 CEST] <Nolski> pzich: ^
[01:16:50 CEST] <blurider> I've been asking for a while but haven't gotten a response. If ffmpeg's compiled with neon instructions, will it work on non-neon systems?
[01:17:00 CEST] <c_14> probably not
[01:17:08 CEST] <blurider> c_14: Okay, thanks a lot!
[01:27:37 CEST] <edoceo> I've got a source video with an invalid duration: 10757 hours.  I've transcoded with ffmpeg but it still shows this invalid duration
[01:28:13 CEST] <edoceo> Here is the command and output: http://edoceo.io/paste?p=W6OdWN
[01:45:31 CEST] <pzich> Nolski: look at the inputs of your second file, Input #1 has a video and audio stream, Input #0 has just video
[01:47:16 CEST] <Nolski> pzich: hm... I see
[01:47:45 CEST] <Nolski> I was looking at the original files and not the ones i was running ffmpeg on
[01:49:59 CEST] <Nolski> ooooh wait. When you scale a video, it doesn't keep the audio track in the output does it?
[01:59:10 CEST] <c_14> It should. Depends on if you have a map or not.
[04:29:49 CEST] <Mista-D> Has anyone been succesfull in building FFmpeg with QuickSync support on Linux ?? having some build difficulties
[08:31:50 CEST] <lomer> hey people ffmpeg -i input.mkv -vcodec copy -acodec copy -ss 00:04:00 -t 00:04:00 output.mkv (Delay relative to video: -651ms)
[08:33:09 CEST] <lomer> Delay relative to video a mistake?
[09:35:08 CEST] <ruukasu> I have a rip of a blu-ray including "BDMV" and "CERTIFICATE" dirs. at BDMV/STREAM there are .m2ts files, the first two of which each contain two episodes of a show, and the markers for chapters as well as splitting it into two episodes are somewhere else in BDMV
[09:35:23 CEST] <ruukasu> is there any way I can convert this strange format into MKVs without transcoding anything?
[09:35:41 CEST] <ruukasu> I believe the video in m2ts is h.264 so it should be possible
[09:36:45 CEST] <ruukasu> (I'd also like to keep all the audio and subtitle tracks as well as chapter marks, no idea which files those come from)
[09:38:03 CEST] <KarlFranz> If it was a DVD, I would use mencoder to extract each track and then mkvmerge to put it all together in an .mkv
[09:38:12 CEST] <KarlFranz> But I am nto familiar with blue-rays
[09:38:27 CEST] <ruukasu> I think it's a very similar system
[09:38:38 CEST] <ruukasu> the weird part is two episodes per title/track/whatever
[09:38:52 CEST] <KarlFranz> I have seen weirder things.
[09:39:08 CEST] <ruukasu> not sure if there's some kind of chapter-like mark splitting the two episodes
[09:39:12 CEST] <KarlFranz> The problem is gonna be if it is an encrypted blu-ray or something like that.
[09:39:49 CEST] <ruukasu> nah, I'm viewing it perfectly fine in mpv on gnu/linux, so I'm pretty sure the encryption is gone
[09:40:16 CEST] <KarlFranz> Great.
[09:40:52 CEST] <ruukasu> if I give mpv the entire directory it has chapter marks, but if I give it the m2ks file directly it does not, so I think the chapter marks come from elsewhere
[09:40:55 CEST] <ruukasu> this is strange
[11:11:07 CEST] <chungy> Is there a way to set the flag-default property in mkv streams?
[11:11:25 CEST] <chungy> I can edit them after-the-fact with mkvpropedit, but it'd be nicer if I can set it directly from ffmpeg.
[11:11:54 CEST] <termos> I'm trying to compile ffmpeg with --enable-decklink but --disable-sdl, then I get a linker error saying libdl.so.2: error adding symbols: DSO missing from command line. Is this a common problem?
[11:20:09 CEST] <KarlFranz> chungy: I use mkvmerge for that, and it has an option for it.
[11:20:26 CEST] <KarlFranz> heh, it looks like I own stocks on mkvmerge or something of the sort.
[11:21:37 CEST] <chungy> stocks?
[11:23:18 CEST] <KarlFranz> Because I think I have been recommending that software for every mkv related operation since I arrived here.
[11:23:40 CEST] <KarlFranz> YOu can also use mkvpropedit for changing the flags of already existing mkv files.
[11:31:11 CEST] <chungy> yeah I plan on doing that if ffmpeg can't do it directly.
[12:47:11 CEST] <satinder_786> hi I am receiving video udp from network , store it in ring buffer for 10 mins then transmit after 10 mins but my cpu is badly effected . My program use 100% cpu . how I can resolve that issue
[12:47:13 CEST] <satinder_786> ??
[13:44:17 CEST] <satinder>  hi I am receiving video udp from network , store it in ring buffer for 10 mins then transmit after 10 mins but my cpu is badly effected . My program use 100% cpu . how I can resolve that issue??
[15:45:15 CEST] <kubast2> Hey ,I need my audio to be compressed[my moto g doesn't have sdcard slot]. Which audio codec would you recommend to me? libopus ,libfdk_aac ,libvorbis?[flac would be obvious choice for audio quality ,but I need my audio codec to be space+quality efficient].
[15:45:57 CEST] <JEEB> for the least effort (something you don't necessarily need to build yourself) you should pick either vorbis or opus
[15:46:11 CEST] <JEEB> opus is better in really low bit rates, and vorbis is already quite good at 128-192 range
[15:47:02 CEST] <sagax> hiall!
[15:47:13 CEST] <sagax> please, help me
[15:47:38 CEST] <sagax> howto make ffserver streaming with idv3 description
[15:47:41 CEST] <sagax> ?
[15:48:19 CEST] <sagax> i read official docs, but not finded information about this
[15:54:37 CEST] <Fyr> how to convert audio/video only from moment I choose?
[15:54:42 CEST] <Fyr> (to moment I set)
[15:55:57 CEST] <Fyr> the documentation lacks of it.
[15:56:15 CEST] <JEEB> see the documentation of parameters -ss and -t
[15:56:22 CEST] <Fyr> ok
[15:56:26 CEST] <JEEB> https://www.ffmpeg.org/ffmpeg-all.html
[15:56:48 CEST] <Fyr> it's where I'm grepping.
[15:57:23 CEST] <JEEB> then you should be able to find those two :)
[15:58:10 CEST] <Fyr> I tried "time."
[16:14:42 CEST] <sagax> ffserver with idv3 it's impossibe?
[16:14:46 CEST] <sagax> impossible*
[16:17:30 CEST] <Fyr> is there a raw audio format?
[16:17:43 CEST] <BtbN> wav?
[16:17:50 CEST] <Fyr> wav is a container.
[16:17:57 CEST] <BtbN> for raw pcm.
[16:18:10 CEST] <Fyr> with tags
[16:18:19 CEST] <j3imi> pcm_s16le?
[16:18:23 CEST] <Fyr> is there one without them?
[16:18:40 CEST] <BtbN> What are you trying to achive?
[16:18:43 CEST] <c_14> -f s16le etc
[16:18:49 CEST] <Fyr> ok, thanks
[16:19:16 CEST] <j3imi> https://trac.ffmpeg.org/wiki/audio%20types
[16:19:24 CEST] <Fyr> thanks
[16:19:45 CEST] <Fyr> I'm trying to understand what format Audacity uses. it uses ffmpeg to import WAV. =)
[16:20:17 CEST] <BtbN> propably whatever format is in the wav file, unless it supports only one
[16:52:05 CEST] <Fyr> is it possible to use only opencl without involving CPU?
[16:52:22 CEST] <DHE> no. only certain features are available as opencl
[16:52:23 CEST] <Fyr> or at least workload ~1% of CPU?
[16:52:28 CEST] <Fyr> =(
[16:52:42 CEST] <DHE> ffmepg only supports a few filters. libx264 supports some lookahead features
[16:52:54 CEST] <DHE> but for x264 I find opencl only provides ~20% performance improvement by offloading
[16:53:10 CEST] <JEEB> wow
[16:53:19 CEST] <JEEB> you actually get some performance out of that PoC feature?
[16:53:36 CEST] <JEEB> you must have really low latency between your CPU and the thing you're running the OpenCL stuff on
[16:53:37 CEST] <Fyr> it's not about performance.
[16:53:43 CEST] <Fyr> it's about temperature. =))
[16:53:44 CEST] <DHE> not a lot
[16:53:57 CEST] <DHE> at the end of the day, the CPU remains hot and the GPU has seen tougher workloads
[16:54:21 CEST] <JEEB> it's a proof of concept feature and it requires fast bandwidth/low latency to be effective
[16:54:28 CEST] <JEEB> so I'm kind of surprised :P
[16:54:47 CEST] <Fyr> when I'm using openCL-optimized programs, I observe <1% of CPU workload and ~0% of GPU workload.
[16:54:54 CEST] <Fyr> I compared ffmpeg and flacCL. the latter is faster than the formar at ~50 times.
[16:54:55 CEST] <sagax> howto make ffserver+ffpeg streaming with IDv3 description?
[16:55:51 CEST] <Fyr> flacCL can make use of only GPU without burdening CPU.
[16:57:55 CEST] <DHE> JEEB: it depends on the input video and settings as well. I just re-ran it with some quick tests. -preset slow gave an improvement, default gave worse performance...
[16:58:18 CEST] <DHE> and my last test was over a year ago so I don't remember the specifics. either way...
[17:03:58 CEST] <sagax> when ffmpeg started, howto i change metadata?
[17:05:19 CEST] <sagax> dynamicle change metadata for ffmpeg, this need for some player might showing title track and other parameters
[17:07:07 CEST] <sagax> i find key -metadata for set metadata options, but it's method is static
[17:08:06 CEST] <sagax> when i have stream, title must be changing dynamicle, without stoping ffmpeg process
[17:08:40 CEST] <sagax> what i can do with this problem?
[17:16:21 CEST] <loop3r_> hello, I am trying to remove fisheye effect using ffmpeg, I was googling some time, I found that there are lenscorrection in ffmpeg and defish0r plugin, but I can't get it working, do you know where can I find working examples ?:P
[18:24:20 CEST] <Nolski> loop3r_: I'd try looking and seeing if you can find one of the creators or an irc channel for that plugin
[18:28:38 CEST] <ocrete> in _getbuffer2(), the height in the AVFrame is not the size of the final image, but seems to include the padding (it give 1088 for a 1080p video), is there some way to get the final image's size?
[18:37:40 CEST] <Nolski> why would amix screw up how a video is rendered?
[18:42:49 CEST] <loop3r_> Nolski: ok I will try :P
[18:45:07 CEST] <durandal_170> Nolski: use pan
[18:47:00 CEST] <Nolski> durandal_170: sorry I've never used pan before. The examples seem a bit complex. I'm just looking to have the output of an overlay filter also contain both audio streams in the end product as well
[18:49:54 CEST] <Nolski> durandal_170: I think the filter would just be `[0:1][1:1]pan=stereo` if I wanted to mix both channels into one?
[19:08:21 CEST] <Eduardo_1> Hello, someone knows how to do 2 pass encoding using nvenc?
[19:08:49 CEST] <Nolski> ah wait, if you don't map the audio, it just kinda writes it all over the video...
[20:12:17 CEST] <aeikum> hi folks. i'm trying to call libavcodec from wine in order to decode wma v1 & v2 audio into pcm. it's mostly going well, but the decoded audio is corrupt. i think this is because i'm missing the extradata that the decoder expects
[20:12:38 CEST] <aeikum> the WAVEFORMATEX that the applications pass in has a cbSize of 0, so there is no extradata to pass in from the format
[20:13:01 CEST] <aeikum> however, applications do provide some data with each buffer they submit: https://msdn.microsoft.com/en-us/library/windows/desktop/microsoft.directx_sdk.xaudio2.xaudio2_buffer_wma%28v=vs.85%29.aspx
[20:13:31 CEST] <aeikum> i was wondering if that looks like the correct data, and if so, how to pass that in as extradata after the context has been initialized
[20:14:01 CEST] <aeikum> happy to provide data dumps, etc if that would help
[20:16:40 CEST] <aeikum> here's an example of an input format and a buffer: http://pastebin.com/MT3ruPch
[20:21:05 CEST] <aeikum> also worth mentioning, i get many of these errors: [wmav2 @ 0x65ff15a0] overflow (1866 > 1864) in spectral RLE, ignoring
[20:28:30 CEST] <aeikum> oh, just found the comments in >libavformat/xwma.c> about creating some fake extradata. i'll give that a shot
[20:47:38 CEST] <aeikum> hey, it works! :D
[20:55:47 CEST] <JEEB> 'grats
[20:59:39 CEST] <loop3r_> does Dennedy visit irc sometime ?
[21:01:14 CEST] <llogan> loop3r_: if that's his nick, then i haven't ever seen him here.
[21:04:04 CEST] <loop3r_> llogan: unfortunately I am not sure his nickname on IRC, I wanted to ask about frei0r plugin, https://github.com/ddennedy he is owner of the repo
[21:06:35 CEST] <llogan> is that now the official repo?
[21:08:03 CEST] <llogan> i think that's a mirror
[21:08:20 CEST] <loop3r_> ouu
[21:08:24 CEST] <llogan> git://code.dyne.org/frei0r.git
[21:09:11 CEST] <llogan> http://frei0r.dyne.org/
[21:09:47 CEST] <loop3r_> Ive visited dyne.org
[21:10:03 CEST] <loop3r_> contributors are the same
[21:10:32 CEST] <loop3r_> so it doesnt matter at all
[21:10:57 CEST] <llogan> if you have questions for them, then use their mailing list
[21:11:04 CEST] <loop3r_> I wanted to ask about fisheye effect and lenscorrection, becouse I can't find working example
[21:11:50 CEST] <llogan> or #dyne irc.dyne.org:9999
[21:12:10 CEST] <loop3r_> oo this is a really good idea!
[21:12:31 CEST] <loop3r_> thanks i will try to join #dyne there
[21:13:56 CEST] <llogan> loop3r_: ffmpeg has a lenscorrection filter
[21:16:17 CEST] <loop3r_> llogan: I tried http://www.ffmpeg.org/ffmpeg-all.html#lenscorrection but I am not native english speaker, I think I miss something I am doing something wrong there, I thought defish0r might be easier
[21:17:37 CEST] <llogan> an example in the docs would be nice; especially showing usage of the mentioned opencv calibration sample
[21:27:54 CEST] <loop3r_> llogan: could you summarize to me, how to pass my fisheyed movie through ffmpeg with lenscorrection, I will find more specific by myself just in big summarazie :P
[21:30:21 CEST] <llogan> i don't know. i guess if you're lazy you may be able to use blender to give you the k1 and k2 values
[21:32:36 CEST] <loop3r_> llogan: I Have to do it from cmd, i will try it all once again
[21:32:43 CEST] <loop3r_> anyway thanks for help!
[23:09:47 CEST] <Nolski> Hey, think any of yall could help take a crack at this problem? I am overlaying two videos and normally this command works but I'm getting this funky error and it's creating this video thats all garbled up and barely recognizable.
[23:09:49 CEST] <Nolski> https://gist.github.com/Nolski/739936423464e7cd1adf
[23:22:48 CEST] <klaxa> well first of all, using -preset with -c:v huffyuv doesn't make sense afaik
[23:23:15 CEST] <klaxa> also, the only error i see is the out of memory error
[23:23:49 CEST] <klaxa> Nolski: ^
[23:23:51 CEST] <Nolski> Right, i was thinking maybe that's related
[23:24:39 CEST] <klaxa> other than that the command looks fine to me
[23:26:25 CEST] <klaxa> did you try a different codec?
[23:26:42 CEST] <klaxa> also, how is ram usage while ffmpeg is running?
[23:26:52 CEST] <Nolski> Not bad at all
[23:27:23 CEST] <Nolski> klaxa: cpu spikes a bit but memory is fine. I have over 8GB left free
[23:27:36 CEST] <klaxa> huh, that's weird
[23:28:02 CEST] <Nolski> Yeah, it's a bit strange because the videos start off as webm
[23:28:32 CEST] <Nolski> and I encode them with huffyuv because otherwise the end quality looks like crap
[23:32:08 CEST] <coops> I've a ffserver time shifting setup that's happily working, but I would like to extract information about the ffm files that it's buffering to. (e.g. their duration, start time, any gaps, etc). Is there a tool to do this, and if not, where can I read about the specs for FFM files?
[23:53:14 CEST] <Nolski> klaxa: it turns out the issue is coming from when I concat videos (as part of the overlay process)
[23:53:23 CEST] <Nolski> does anything about this command look fishy to you? https://gist.github.com/Nolski/aeb50d0dc9b9558841a6
[23:53:36 CEST] <Nolski> Normally I use filters but I figured it wouldn't be all that different like this
[23:53:43 CEST] <klaxa> depends on the content of loop.txt
[23:54:05 CEST] <klaxa> and the files referenced within
[23:54:06 CEST] <Nolski> All the videos should have the same codec They're all part of the same orginal video just cut up into 3 pieces with the middle one having the overlay
[23:54:10 CEST] <klaxa> if the files have different codecs it will probably not work
[23:54:11 CEST] <klaxa> ah
[23:54:17 CEST] <klaxa> sounds good
[23:54:48 CEST] <Nolski> but the middle video gets all screwed up (upside down, major discoloration, etc) after the concat
[23:55:05 CEST] <Nolski> so maybe it is a codec thing and I need to re-encode using a filter?
[23:56:30 CEST] <klaxa> i don't know, that has never happened to me before
[23:56:40 CEST] <klaxa> can you play the individual files without problems?
[23:56:58 CEST] <Nolski> klaxa: yup. Prior to the concat, everything plays fine
[23:57:12 CEST] <Nolski> it's just that middle video after the concat that gets all screwed up
[23:57:47 CEST] <klaxa> can you pastebin ffprobe for all three files?
[23:59:35 CEST] <Nolski> klaxa: sure thing: https://gist.github.com/Nolski/6c22e7766689fa1d7b5d
[23:59:59 CEST] <Nolski> that middle one says bgra
[00:00:00 CEST] --- Wed Jul 29 2015

More information about the Ffmpeg-devel-irc mailing list