[Ffmpeg-devel-irc] ffmpeg.log.20160201

Tue Feb 2 02:05:01 CET 2016

[00:15:47 CET] <xochilpili> hi there!
[00:16:21 CET] <xochilpili> im having some issues with mkvmerge and subtitles Error: The file 'subtitle.srt' has unknown type
[00:16:26 CET] <xochilpili> any hand  with this?
[00:17:11 CET] <sbraz> xochilpili: that's a mkvtoolnix question, care to upload the file somewhere though?
[00:22:08 CET] <solrize> hmm thanks i'm currently using vp8 encoder which only uses 1 core.
[01:08:25 CET] <p1_> where should I report a company that is using ffmpeg in a product without attribution nor source?
[01:08:46 CET] <c_14> open an issue on trac
[01:13:07 CET] <p1_> yeah, I'd prefer to do it anomomously
[01:14:28 CET] <velusunivers-sys> hello ca\n ffmpeg be used as a streaming server? im hoping to get a online tv station for virtual worlds, im needing to have something that will be able to play videos from a list i.e xml
[01:20:18 CET] <J_Darnley> p1_: email someone directly then
[01:20:48 CET] <J_Darnley> Michael Niedermeier perhaps (michaelni)
[01:21:48 CET] <J_Darnley> There's also a private security mailing list you could email but I'm not sure about the nettiquette using it for non security matters.
[01:23:57 CET] <retard> velusunivers-sys: ffserver
[01:24:21 CET] <velusunivers-sys> how easy is it to set up, and what format does it stream in?
[01:24:37 CET] <c_14> I honestly wouldn't recommend ffserver.
[01:24:45 CET] <c_14> Use HLS/dash or maybe nginx-rtmp
[01:24:45 CET] <velusunivers-sys> why not?
[01:25:05 CET] <velusunivers-sys> i need something that can read from a play list
[01:25:09 CET] <p1_> I think there needs to be a way to report these things anonomously, as people take a risk doing so.
[01:25:23 CET] <velusunivers-sys> i would prefere the playlist be in xml but yeah any format would be good
[01:25:52 CET] <c_14> p1_: create an account on trac with a throwaway email address?
[01:27:20 CET] <J_Darnley> I guess it depends how anon you want to be.  We could set up some sort of tor-only drop box but that sounds like a lot of work.
[01:27:33 CET] <p1_> can I just report it here?
[01:27:57 CET] <J_Darnley> Are you aware that this channel is publicly logged?
[01:28:14 CET] <p1_> do they log ip addresses?
[01:28:39 CET] <velusunivers-sys> how easy is ffserver to set up and what format does it stream in?
[01:28:41 CET] <J_Darnley> Not sure.  I would have to check it to see if joins are there.
[01:28:44 CET] <c_14> I'm pretty sure join/parts are logged. Yes
[01:29:08 CET] <p1_> ok, I'll go to a internet cafe and do it :-)
[01:30:14 CET] <c_14> Hmm, it doesn't look like they're logged on the public logs, but there could always be other ones.
[01:30:52 CET] <p1_> cafe sounds the safest anyway
[01:31:27 CET] <J_Darnley> I wonder if its the NSA and he's in danger of being black bagged
[01:31:41 CET] <J_Darnley> No, the CIA do that.
[01:31:47 CET] <J_Darnley> Is it them?
[01:34:24 CET] <velusunivers-sys> CIA NSA and the  MI6 MI7, and SPC
[01:34:36 CET] <velusunivers-sys> do the black bagging
[01:35:00 CET] <J_Darnley> Don't forget our benevolent overlords from the FSB
[01:38:22 CET] <p1_> It is a company Called echo360, and the product is echo360. They bundle up ffmpeg in a java wrapper to obscure it, but you can see it running in the process table when the wrapper is called.
[01:38:53 CET] <Bray90820> is this a good channel to talk about video editing?
[01:40:43 CET] <J_Darnley> Bray90820: a little off topic but we do know things
[01:42:11 CET] <Bray90820> j am basically looking for a way maybe in adobe premiere to keep the same dolby digital 5.1 audio track as was on the video before I edited it
[01:42:15 CET] <Bray90820> If that makes any sense
[01:43:49 CET] <J_Darnley> If you made edits in time then I guess it won't let you do that.
[01:44:07 CET] <J_Darnley> Maybe you can import the original again as new new track or something.
[01:44:23 CET] <Bray90820> I meant the same encoding
[01:44:45 CET] <Bray90820> Or maybe format it's called
[01:44:48 CET] <J_Darnley> You mean you want it to encode to 5.1 DTS?
[01:45:39 CET] <Bray90820> I am actually not sure if I want 5.1 DTS what I want is whatever the video had originally
[01:46:24 CET] <J_Darnley> Well you probably need to tell it what format you want the audio.
[01:46:58 CET] <J_Darnley> I can't say whether it has a "look at the input and make it like that" setting.
[01:47:14 CET] <Bray90820> Is there a way I can tell it to keep the same format?
[01:48:00 CET] <J_Darnley> I have no idea. I've never used it.
[01:48:33 CET] <Bray90820> Alright
[01:48:41 CET] <Bray90820> Thanks for the help you could give
[01:54:59 CET] <andrey_utkin> JEEB: "doc/examples/transcoding cannot into libx264" with all details emailed to libav-user and ffmpeg-devel @ ffmpeg.org. Awaits moderation approval (not subscribed on my new mail)
[12:31:03 CET] <dorp> I'm trying to use ffmpeg for capturing an audio device under Windows: "CABLE Output (VB-Audio Virtual Cable)" ... but it seems that I can't specify 6 channels, ffmpeg says: "Guessed Channel Layout" ... "Stereo" ... anybody has any idea if/how to enable 6 channels?
[12:40:27 CET] <J_Darnley> What input device are you using (dshow, other)?
[12:41:21 CET] <J_Darnley> That has the channels options you can try
[12:41:53 CET] <dorp> J_Darnley: Input #0, dshow, from 'audio=CABLE Output (VB-Audio Virtual Cable)'
[12:42:28 CET] <J_Darnley> try prepending "channels=6:" to the input argument.
[12:44:32 CET] <dorp> As in: -f dshow -i channels=6:audio="..." ... ? "Malformed dshow input string"
[12:44:48 CET] <J_Darnley> Yes
[12:45:06 CET] <J_Darnley> Oh
[12:45:23 CET] <J_Darnley> Maybe I mean: add -channels 6 before -i
[12:47:05 CET] <dorp> J_Darnley: Not sure what that's about: "Could not set audio only options" ... "Searching for audio device within video devices..."
[12:47:16 CET] <dorp> Do I have to specify a video input device, too?
[12:47:53 CET] <J_Darnley> I have no idea.  I've only used dshow acouple of times to capture a webcam.
[12:48:26 CET] <J_Darnley> You can use -list_option to show the options available for your audio device.
[12:48:41 CET] <J_Darnley> -list_options
[12:49:27 CET] <relaxed> look at "ffmpeg -h demuxer=dshow"
[12:49:59 CET] <dorp> -list_options reveals: Pin "Capture" (alternative pin name "Capture") ... min ch=1 bits=8 rate= 11025 max ch=2 bits=16 rate= 44100
[12:51:27 CET] <dorp> relaxed: Thanks for the hint
[12:51:37 CET] <J_Darnley> There's also http://ffmpeg.org/ffmpeg-devices.html#dshow
[12:52:03 CET] <dorp> It seems that "-channels 2" works, where "-channels 6" produces: "Could not set audio only options"
[12:52:24 CET] <J_Darnley> But I think it looks like 6 channel input for your device isn't available through dshow (or ffmpeg doesn't know how).
[12:52:40 CET] <dorp> I guess I better look for another audio device and see if it works there
[12:53:19 CET] <dorp> (if anybody has a suggestion for such an audio device, to capture 6 channels under Windows, that happens to work with ffmpeg, I'm open to suggestions)
[12:54:10 CET] <tp_> anyway to use -ss while using -re too?
[13:04:30 CET] <mcmillen_> msg
[13:05:50 CET] <mcmillen_> msg how to use videotoolbox in FFmpeg?
[13:07:16 CET] <mcmillen_> msg FFmepg's official wiki and website has no any document or examples about the videotoobox
[13:14:34 CET] <BtbN> What even is "the videotoolbox"?
[13:16:18 CET] <J_Darnley> tp_: Does it not work?
[13:16:20 CET] <mcmillen_> videotoolbox is a framework both in iOS and OS X, for h264 hardware encoding/decoding
[13:17:00 CET] <mcmillen_> i can't find any example code for videotoolbox hwaccel
[13:18:08 CET] <mcmillen_> ffmpeg release 2.8 source include the videotoolbox support in libavcodec/videotoolbox.h, but no actually work
[13:19:26 CET] <J_Darnley> If you can't compile ffmpeg with that support then please tell us more.
[13:19:51 CET] <BtbN> It's propably just like any other hwaccel. ffmpeg just fills the buffers for you
[13:19:57 CET] <J_Darnley> Is there some other problem you're experiencing?
[13:20:13 CET] <J_Darnley> "doesn't work" is not an error message.
[13:22:36 CET] <mcmillen_> is there any document or example for videotoolbox ? like qsv-accelerated decoding? https://www.ffmpeg.org/doxygen/2.8/qsvdec_8c-example.html#a72
[13:24:33 CET] <J_Darnley> It doesn't look like it (judging from grep)
[13:24:57 CET] <J_Darnley> All I can say is read how ffmpeg does it.
[13:29:59 CET] <BtbN> QSV is a "complete" encoder. VT is just a hwaccel module.
[13:30:10 CET] <BtbN> You still have to do the majority of work yourself.
[13:30:35 CET] <mcmillen_> even the vda-acclerated decoding just a wiki with few introduction. https://github.com/dilaroga/ffmpeg-vda/wiki/FFmpeg-vda-usage
[13:31:11 CET] <BtbN> That page is not from ffmpeg.
[13:33:02 CET] <tp_> J_Darnley: nope, it will just delay showing the video until the seconds it gets from -ss
[13:33:29 CET] <J_Darnley> Can ffmpeg seek in the input or is it some sort of stream?
[13:33:40 CET] <tp_> it can see
[13:33:55 CET] <tp_> seek*
[13:34:01 CET] <J_Darnley> okay.  I guess -re does emulate well then.
[13:34:23 CET] <mcmillen_> actually, the developer using the FFmpeg, almost can't using the hwaccel feature. isn't it ? i search "ffmpeg videotoolbox" from google these days, can't find example code any.
[13:34:28 CET] <tp_> i can seek the video with -ss, when im not using -re, but when using -re, i can't
[13:35:03 CET] <J_Darnley> And -re is supposed to make ffmpeg behave as though it is a stream (or something).
[13:35:31 CET] <J_Darnley> Oh I should ask: are you putting -ss before the -i?
[13:35:42 CET] <tp_> yes, before, also tried after
[13:36:06 CET] <J_Darnley> Then if you feel it is a bug then submit a report to trac.
[13:39:40 CET] <tp_> if theres some other way i can encode at normal speed, i dont think its a problem
[13:40:03 CET] <J_Darnley> Why do you want to do that?
[13:42:19 CET] <tp_> because im transcoding to a stream with seeking ability, and if i dont do it, ffmpeg will use 100% cpu, when im using -re it doesnt use 100% cpu, but then i cant seek
[13:57:00 CET] <TekniQue> I have confirmed that ffmpeg is in fact horrible at multithreading
[13:57:57 CET] <TekniQue> and as ou increase the number of outputs, performance goes down but CPU usage hits a ceiling
[13:58:37 CET] <TekniQue> and doing the same work in multiple ffmpeg processes accomplishes a lot more than having one process produce multiple outputs
[13:59:13 CET] <BtbN> Fully using every core seems more like it's good at multi threading.
[13:59:31 CET] <TekniQue> it's not fully using every core, that's the problem
[13:59:47 CET] <TekniQue> encoding drops below real time but cpu usage is at 50% roughly
[14:00:06 CET] <TekniQue> but by running multiple processes I'm able to use more cpu
[14:00:07 CET] <J_Darnley> More encoding -> more cpu usage -> cpu has limits -> encoding speed drops.  Is that surprising?
[14:00:08 CET] <BtbN> ffmpeg itself isn't multi threaded at all, btw. It's strictly single-threaded. Only certain encoders/filters are
[14:00:10 CET] <TekniQue> and encode in real time
[14:00:40 CET] <TekniQue> BtbN: that explains a lot
[14:01:12 CET] <BtbN> No idea how multiple outputs are implemented, but I wouldn't be surprised if it works through them sequentualy for each packet
[14:03:07 CET] <TekniQue> yeah that sounds likely
[14:03:36 CET] <TekniQue> the scenario is that encoding 4 outputs doesn't use any more cpu than encoding 3 outputs
[14:04:34 CET] <TekniQue> and that setting x264 to a faster preset reduces cpu time but doesn't improve encoding speed
[14:04:52 CET] <BtbN> well, the encoder should still only run once, no matter how many outputs you have. So that doesn't sound unusual.
[14:05:09 CET] <TekniQue> running 4 encoders actually
[14:05:18 CET] <TekniQue> encoding at different resolutions and bitrates
[14:05:40 CET] <BtbN> I wouldn't bother using a single ffmpeg process for it then.
[14:05:42 CET] <BtbN> Just run 4 of them
[14:05:50 CET] <BtbN> Way easier to manage, too
[14:06:16 CET] <pkeuter> hey guys, i'm trying to run the following command: http://pastebin.com/LVQKGAKx, and it gives me a "Error when filtering: cannot allocate memory"-error. Any idea what could be wrong? The machine still has 12GB of RAM left, so that shouldn't be the issue...
[14:06:47 CET] <TekniQue> BtbN: yup
[14:09:12 CET] <durandal_1707> Probably amix fault
[14:09:27 CET] <pkeuter> alright BtbN : http://pastebin.com/Tx9RC8fN, is that better?
[14:10:53 CET] <BtbN> Is it a 32bit build of ffmpeg? That's running out of virtual memory for some reason?
[14:11:12 CET] <pkeuter> durandal_1707, well that definately fixes it, removing the audio filters...
[14:12:23 CET] <durandal_1707> pkeuter: use amerge and pan instead of amix
[14:12:24 CET] <pkeuter> BtbN, it's built from source on a 64 bit machine, so no, i guess.
[14:13:05 CET] <BtbN> Strange, what the hell is amix doing to exhaust the memory resources?
[14:15:20 CET] <pkeuter> durandal_1707, what is the difference between amerge and amix?
[14:15:35 CET] <pkeuter> And BtbN; my question exactly
[14:16:54 CET] <durandal_1707> something stupid?
[14:49:11 CET] <pkeuter> well thanks for getting me on the right path!
[14:57:43 CET] <bobby_> I know this question has been asked many times on the mailing list, but I can't find any clear answer: is av_write_frame() functions is thread safe?
[15:15:18 CET] <BtbN> At least not for the same context/packt
[15:15:23 CET] <BtbN> +e
[15:47:16 CET] <roxlu> Hey I'm working on some code which muxes mpeg-ts and I was wondering if someone knows a good approach to restamp the pts/pcr values. I know this is not a easy thing to do, but I was hoping someone here has done something similar.
[16:18:22 CET] <DHE> the mpegts spec calls for the PCR to be very accurate (0.5 microsecond tolerance). any reason not to use ffmpeg since it should already do it properly?
[16:25:45 CET] <Mavrik> ffmpeg's PCR timings are nowhere near spec compliant.
[17:50:16 CET] <HiddenWolf> Hi all. I'm trying to follow the example on the wiki and re-encode a stream to x265, but I'm getting an error 'unrecognised option x265-params'
[17:52:21 CET] <HiddenWolf> I am using ffmpeg 2.8.5,1 on FreeBSD 10.2
[17:53:35 CET] <atomnuker> HiddenWolf: paste your command line
[17:54:25 CET] <dorp> Anybody happens to know if ffmpeg is capable of capturing 6 audio channels via dshow (Windows)? ... I've tried two virtual audio devices, and a physical external device, when I try: "-channels 6" I get: "Could not set audio options"
[17:57:15 CET] <HiddenWolf> atomnuker: http://pastebin.com/Hrjj3Rah
[17:58:09 CET] <HiddenWolf> atomnuker: adapted from the wiki and a stackexhange example. Ideally I'd grab the english subtitle from an idx file and mix that in as default
[17:59:28 CET] <atomnuker> HiddenWolf: you can also set the crf using -crf 28
[17:59:41 CET] <atomnuker> so just remove -x265_params crf=28 and just use -crf 28
[18:00:03 CET] <atomnuker> or maybe just wrap "crf=28" in quotation marks
[18:00:52 CET] <c_14> HiddenWolf: your version of ffmpeg isn't built against libx265
[18:01:04 CET] <HiddenWolf> c_14: ah
[18:01:10 CET] <HiddenWolf> c_14: drat
[18:04:17 CET] <furq> time to update your ports
[18:05:05 CET] <HiddenWolf> furq: I guess. I just grabbed the default pkg though. Why build if they're pre-built for you.
[18:05:19 CET] <furq> because the default options suck
[18:05:26 CET] <furq> although i guess i don't need to tell you that now
[18:05:57 CET] <HiddenWolf> furq: in this case, yeah. but it saves a lot of time and headache, usually.
[18:06:29 CET] <HiddenWolf> At least, for the lazy amateur sysadmin that I am.
[18:45:59 CET] <dorp> Anybody can guess whether ffmpeg should be able to capture 6 audio channels under Windows? (dshow)
[18:46:41 CET] <dorp> (I don't know if I should bother looking for more devices to test with)
[20:58:03 CET] <dorp> So apparently the issue I've stumbled is not specific to ffmpeg, but an issue with multi-channel recording with DirectShow as a whole. I've managed to use something called "WASAPI" with a non-free virtual audio device for capturing 6 audio channels. Not fun
[21:20:42 CET] <markusl> hi
[21:23:32 CET] <markusl> I'm trying to implement DASH audio streaming from alsa using ffmpeg. so far everything works, but now i'm trying to have multiple streams/representations in the manifest
[21:24:18 CET] <markusl> that seems to work, too. but the manifest generated by ffmpeg has the same bitrate (128000) for every representation even if they differ.
[21:24:40 CET] <markusl> I wonder if i'm calling the webm_dash_manifest in the wrong way or if this is a bug.
[21:37:29 CET] <markusl> I've got the command lines (systemd unit) here: http://pastebin.com/22zHuGdE output of encoder + manifest generation: http://pastebin.com/bLaBbS7r
[21:38:19 CET] <markusl> result (w/ all streams reporting the same bandwidth) in shaka-player: https://www.eldoradio.de/experiment/dash/
[21:39:53 CET] <markusl> (currently plays in chrome but not in ff even if shaka says ff is supported after loading the polyfills, but that's a different story)
[21:43:36 CET] <sanket> i am new to contributing to opensource, can anyone help me get started for contributing to ffmpeg
[22:02:36 CET] <J_Darnley> sanket: what can you do?
[22:02:42 CET] <J_Darnley> or what do you want to do?
[22:02:50 CET] <DHE> there's a page on the web site with info and guidelines
[22:03:42 CET] <sanket> i want to contribute by coding
[22:05:43 CET] <J_Darnley> You could start by reading http://ffmpeg.org/developer.html
[22:05:58 CET] Action: J_Darnley notes that "coding" still isn't very specific.
[22:07:03 CET] <genii> sanket: What language were you thinking to code in?
[22:08:27 CET] <sanket> genii: mainly c and c++
[22:08:41 CET] <furq> can i contribute by writing perl 6
[22:10:14 CET] Action: J_Darnley chases furq away with the hose.
[22:10:40 CET] <furq> are you annoyed at my bad joke or at perl 6
[22:49:35 CET] <HiddenWolf> So, I got an ffmpeg built with x265
[22:50:28 CET] <HiddenWolf> Now, Is it possible to grab a subtitle from an .idx file?
[22:50:44 CET] <HiddenWolf> Preferably the right language?
[22:51:06 CET] <c_14> Assuming it's tagged, yes
[22:51:16 CET] <c_14> At least, I think so
[22:51:27 CET] <c_14> Does ffprobe file.idx show several streams?
[22:55:49 CET] <podman> If I'm trying create multiple resolutions representations for DASH. Youtube always vertical height when describing their resolutions but that seems to assume 16x9 videos. Does it make more sense to use the width instead when calculating the size? For instance if I were to upload a video that were 1280x544 to YouTube, it would describe that as 720p
[22:58:00 CET] <HiddenWolf> c_14: It craps out, no such file. And the file certainly is there
[22:59:28 CET] <TD-Linux> podman, well internally youtube steps up/down resolutions to utilize available bandwidth, so it doesn't really matter what exactly you pick
[23:00:17 CET] <podman> TD-Linux: oh, i'm aware. I'm just trying to figure out a reasonable way to generate different resolutions
[23:00:30 CET] <TD-Linux> IIRC youtube's resolution steps actually fit the video into a 16:9 box
[23:00:40 CET] <HiddenWolf> c_14: http://pastebin.com/ahreXqHU
[23:01:11 CET] <TD-Linux> which is a good way to do it if you're dealing with user generated content
[23:01:12 CET] <podman> For instance I could use something like -vf scale=-2:720, but that would wind up with a video that is 1694x720
[23:01:35 CET] <TD-Linux> yeah I'd suggest fitting it into a box, don't know if there is a vf scale option that can do that though (there is for imagemagick)
[23:02:00 CET] <podman> but, if i went with the horizontal, on the other hand, 1280x544 which makes more sense
[23:02:08 CET] <podman> oh, i can do the math
[23:02:30 CET] <TD-Linux> googling gives this horrific string -vf scale="'if(gt(a,4/3),320,-1)':'if(gt(a,4/3),-1,240)'"
[23:02:42 CET] <podman> yeah, i already have something pretty crazy
[23:02:52 CET] <podman> i guess that's assuming a 4:3 aspect ratio?
[23:02:56 CET] <c_14> HiddenWolf: the idx is just an index file, the .sub contains the actual subtitles and it can't find the .sub
[23:03:12 CET] <HiddenWolf> ah, drat, of course
[23:03:27 CET] <podman> TD-Linux: I already have something pretty involved: scale='trunc(min(iw+mod(iw,2),#{width})/2)*2:-2'
[23:05:14 CET] <podman> TD-Linux: do you happen to know what "a" is in that scale filter? Aspect?
[23:05:31 CET] <TD-Linux> it must be.
[23:06:18 CET] <podman> so, if the aspect greater than 4:3, use 320. otherwise -1
[23:06:19 CET] <podman> interesting
[23:08:11 CET] <podman> so, an equivalent might be 'if(gt(a,16/9),1280,-2)':'if(gt(a,16/9),-2,720)'
[23:08:29 CET] <podman> I'll see what that does
[23:10:06 CET] Action: TD-Linux is a little bit leery of rounding errors in that
[23:10:35 CET] <TD-Linux> eh should be fine.
[23:10:40 CET] <podman> Yeah, I might mess around with it a bit more
[23:10:47 CET] <podman> but it did seem to give me 1280x544
[23:12:52 CET] <J_Darnley> podman: yes "a" is the display aspect ratio (within the scale filter)
[23:12:54 CET] <podman> seems broken for 16x9 or less though
[23:12:58 CET] <podman> J_Darnley: Thanks
[00:00:00 CET] --- Tue Feb  2 2016