[Ffmpeg-devel-irc] ffmpeg.log.20170327

[burek]

[02:07:42 CEST] <ZeroWalker> another day to try to figure out hot to compile ffmpeg with libx264
[02:10:18 CEST] <JEEB> with MSVC, you mena
[02:10:31 CEST] <JEEB> I have gotten as far as building libx264 with MSVC
[02:10:51 CEST] <JEEB> I *think* I got it linked against libavcodec but that was some time ago :P
[02:11:29 CEST] <JEEB> I think there were issues wrt to include and lib search paths on the command line, I might have worked around that with the related env variables
[02:17:19 CEST] <ZeroWalker> i kinda gave up on that as i couldn't make libx264 compile with it, i was able to compile it with mingw64, at least i think that's what i compile it with
[02:17:25 CEST] <ZeroWalker> but ffmpeg doesn't detect it;(
[02:17:54 CEST] <JEEB> mingw-w64 should be simple
[02:18:37 CEST] <JEEB> 1) use a prefix (--prefix) 2) install with make install 3) define PKG_CONFIG_PATH=/your/prefix/lib/pkgconfig when configuring FFmpeg
[02:20:17 CEST] <ZeroWalker> ah, i do the first two, wtih ffmpeg i tried using the extra-cflags and whatnot. will try the define. how do you write that though do i write it at it's own line before the configure?
[02:21:34 CEST] <JEEB> any way is fine, I usually define it on the same line: PKG_CONFIG_PATH=... ../configure --params --more-params
[02:22:00 CEST] <ZeroWalker> ah, didn't think that worked, okay, will try it out
[02:22:22 CEST] <ZeroWalker> is it supposed to take ages when configuring ffmpeg btw?
[02:22:37 CEST] <JEEB> yes, although I hear it's a bit faster with dash instead of bash
[02:22:54 CEST] <JEEB> shell's doing a lot of fork()s which are emulated really slowly on windows
[02:23:23 CEST] <JEEB> although if you're running the msys msys2 shell it might be wiser to do PKG_CONFIG_LIBDIR, which leaves out the msys search paths (_PATH appends, while _LIBDIR overrides)
[02:24:44 CEST] <ZeroWalker> don't think i heard of dash, is it the native linux thing or something?
[02:24:59 CEST] <ZeroWalker> hmm didn't seem to work, will try the PKG_CONFIG_LIBDIR instead
[02:25:02 CEST] <JEEB> it's a much simpler shell
[02:25:10 CEST] <JEEB> the effect should be the same, look at config.log
[02:25:33 CEST] <JEEB> also did you actually build libx264?
[02:25:46 CEST] <JEEB> --enable-static or --enable-shared with x264's configure
[02:25:52 CEST] <JEEB> by default it builds neither
[02:28:10 CEST] <ZeroWalker> last build was
[02:28:20 CEST] <ZeroWalker> ../x264/configure --enable-strip --disable-cli --enable-static --prefix=../x264_installed
[02:41:33 CEST] <ZeroWalker> that define didn't work either, guess i am doing something else wrong;(
[02:45:01 CEST] <thebombzen_> out of curiosity, for lossless encoding, deblocking is always disabled, right?
[03:03:05 CEST] <ZeroWalker> JEEB, was my configure wrong?
[03:03:48 CEST] <ZeroWalker> thebombzen_, i would be very surprised if any encoder had deblocking on lossless mode, that would make it semi-lossless, it would just introduce artifacts on lossless content which makes no sense
[10:48:36 CEST] <termos> i'm trying to create filter graph using the examples but I keep getting "Error number 1 occurred" when parsing the filter graph "null"
[12:16:21 CEST] <termos> it seems the time_base of my AVCodecContext is set to 0/2 and not (what I assume is correct) 50/2. The time_base.num is always 0 it seems, this is after avcodec_parameters_to_context
[16:07:15 CEST] <Soni> how can I use ffmpeg to merge/split audio/video and stuff?
[16:07:47 CEST] <c_14> merge concatenate or merge different streams?
[16:08:14 CEST] <Soni> well I first need to chop them up into pieces, then merge pieces from different videos
[16:08:24 CEST] <c_14> so concatenate/trim
[16:08:31 CEST] <c_14> https://trac.ffmpeg.org/wiki/Concatenate
[16:08:48 CEST] <c_14> https://trac.ffmpeg.org/wiki/Seeking
[16:13:31 CEST] <Soni> ok so I have a video that I wanna keep as-is, but I wanna replace part of the audio (and keep the rest)
[16:13:45 CEST] <Soni> how painful is this gonna be?
[16:15:25 CEST] <Soni> do I extract the audio, chop it up, concatenate it, then combine with the video?
[16:15:38 CEST] <c_14> pretty much
[16:15:58 CEST] <c_14> though you can do the chopping up while extracting it
[16:16:44 CEST] <c_14> you can even do it all in one command (if you don't mind reencoding and you're slightly masochistic)
[16:17:40 CEST] <c_14> but yes, the easiest thing is extract audio, cut, concat,  combine
[16:17:57 CEST] <Soni> ffmpeg -i video.mkv audio.flac or something? (not that I expect flac quality, I just don't want the quality to drop while I edit it)
[16:18:13 CEST] <c_14> ffmpeg -i video.mkv -map 0:a -c:a copy audio.mkv
[16:18:18 CEST] <c_14> that will copy the audio track out
[16:20:19 CEST] <Soni> I need it to be lossless for the concatenate part no?
[16:20:58 CEST] <Soni> I mean I don't really know what I'm doing
[16:21:50 CEST] <c_14> hmm?
[16:21:59 CEST] <c_14> you don't _need_ anything to be lossless
[16:22:17 CEST] <c_14> -c:a copy is "lossless" in that there's no reencoding so the quality doesn't change
[16:22:24 CEST] <c_14> you don't need a lossless codec to concatenate audio
[16:24:03 CEST] <Soni> eh how do I just turn it into flac? :/
[16:24:30 CEST] <c_14> you don't have to?
[16:24:50 CEST] <c_14> unless you _really_ _really_ want to
[16:25:14 CEST] <Soni> I do want to
[16:25:36 CEST] <c_14> then just drop the -c:a copy and used a filename ending with .flac
[16:27:59 CEST] <Soni> alright thanks
[16:28:22 CEST] <Soni> I'll probably do the chopping up thing with audacity tbh, since I don't wanna introduce clicks
[16:32:00 CEST] <c_14> whatever works for you
[16:59:53 CEST] <Soni> thanks c_14
[18:21:03 CEST] <TWIST_> I have written simple program which extract video stream. It's working fine but playing very fast, How can I introduce delay? Note: its decoding example given
[18:25:54 CEST] <Darxus> TWIST_: I probably don't know the answer to your question, but I imagine you're going to need to provide a lot more info.
[18:28:41 CEST] <TWIST_> Sorry for that, In a way I am asking how can we manipulate pts and dts of AVPacket or AVFrame?
[18:33:53 CEST] <DHE> the pts and dts are a measure of time relative to the stream's time_base that the packet (and ultimate frame) came from
[18:34:31 CEST] <DHE> the time_base varies. sometimes it's 1/30 in which case 30fps material will just have sequential PTS values. sometimes it's a big number like 1/90,000 or 1/1,000,000 and the PTS moves in large increments
[18:55:27 CEST] <fred1807> how can I render this CSS backroung to a 1920x1080 static video file? https://uigradients.com/#SweetMorning
[19:09:13 CEST] <furq> !filter geq @fred1807
[19:09:13 CEST] <nfobot> fred1807: http://ffmpeg.org/ffmpeg-filters.html#geq
[19:15:22 CEST] <furq> fred1807: -f lavfi -i color=white:s=1920x1080 -vf "geq=r=255:g=(95*(X/W))+(195*((W-X)/W)):b=(109*(X/W))+(113*((W-X)/W))"
[19:15:25 CEST] <furq> something like that
[19:17:28 CEST] <alexpigment> does anyone know the status of HEVC/H.265 interlaced support? I can get it to be interlaced according MediaInfo either by a) using -x265-params interlaced=tff along with the ilme and ildct flags, or b) using -top 1 and the same flags. strangely, this produces two differently-encoded files, but neither of them decode correctly on any players I tested.
[19:18:15 CEST] <fred1807> nice
[19:18:18 CEST] <JEEB> first make sure you can get correct interlaced output with the x265 cli
[19:18:29 CEST] <JEEB> after that start adding libavcodec into the mix
[19:18:41 CEST] <alexpigment> JEEB, good thinking. i'll test that first
[19:18:52 CEST] <furq> fred1807: you can replace 95 with 0x5f etc
[19:19:03 CEST] <JEEB> but yes, -x265-params is probably the way you want to do it in addition to telling ffmpeg.c to output interlaced images
[19:26:11 CEST] <furq> huh
[19:26:33 CEST] <furq> how do i set the sar/dar after using nullsrc/color/etc
[19:26:48 CEST] <furq> setsar is just being ignored
[19:26:54 CEST] <JEEB> o_O
[19:28:24 CEST] <furq> http://vpaste.net/OjAGf
[19:28:35 CEST] <furq> same thing if i move scale before setsar
[19:30:20 CEST] <Tom_B> is it possible to burn dvb_subtitle (or possibly dvb_teletext, I'm not sure which is which to be honest!) into the frame?  I've tried the example from https://trac.ffmpeg.org/wiki/HowToBurnSubtitlesIntoVideo and get either no decoder/filtering impossible or "Impossible to convert between the formats supported by the filter 'Parsed_overlay_0' and the filter 'auto-inserted scaler 1'"
[19:31:07 CEST] <furq> https://trac.ffmpeg.org/wiki/HowToBurnSubtitlesIntoVideo#Picture-basedsubtitles
[19:31:09 CEST] <furq> do you mean that example
[19:31:11 CEST] <JEEB> Tom_B: it should be, dvb subtitles should work with the overlay filter and then the dvb teletext just requires  libzvbi and it gets converted into ASS
[19:31:26 CEST] <JEEB> in which case you can use a text-based subtitle thing to burn it on
[19:31:43 CEST] <furq> oh right yeah dvb_teletext uses the subtitle filter, dvd_subtitle uses overlay
[19:31:50 CEST] <furq> s/dvd/dvb/
[19:32:33 CEST] <Tom_B> yeah, I've been trying to use the picture based subtitles
[19:33:32 CEST] <furq> oh wtf
[19:33:39 CEST] <kepstin> furq: that's behaving exactly as expected, you use setsar, then override with setdar
[19:33:43 CEST] <kepstin> furq: use one or the other
[19:34:00 CEST] <furq> it works with just setdar
[19:34:20 CEST] <Tom_B> I've tried [0:v][0:s:0]overlay[v] and [0:v][0:s:1]overlay[v] neither works, am I missing something obvious?
[19:34:22 CEST] <furq> just setsar is broken
[19:34:57 CEST] <kepstin> furq: setsar takes sar as input and sets both sar and dar -- setdar takes dar as input and sets both sar and dar
[19:35:11 CEST] <kepstin> so whichever one you use last takes effect
[19:35:52 CEST] <kepstin> it doesn't make sense to set sar and dar separately, because sar and dar are related based on the frame saize
[19:36:39 CEST] <furq> i figured just setsar=1 would work, but it does nothing
[19:36:46 CEST] <furq> i have to setdar=16/9
[19:36:56 CEST] <furq> which seems stupid when i'm scaling to a 16:9 resolution
[19:37:28 CEST] <kepstin> furq: nullsrc makes 4:3 frames by default, so setsar sets sar to 1:1 on the 4:3 frame, then the scale filter preserves the dar
[19:37:37 CEST] <kepstin> works as expected
[19:37:48 CEST] <furq> it sets it to 3:4
[19:38:52 CEST] <kepstin> furq: no, the scale filter updates the sar to 3:4 when it's doing the scaling, so that the dar of 4:3 is preserved
[19:39:13 CEST] <kepstin> you have 320x240 [SAR 1:1 DAR 4:3] -> scale -> 1280x720 [SAR 3:4 DAR 4:3]
[19:39:18 CEST] <kepstin> works as expected.
[19:41:13 CEST] <kepstin> technically, you have nullsrc -> 320x240 [SAR 1:1 DAR 4:3] -> setsar=1 -> 320x240 [SAR 1:1 DAR 4:3] -> scale -> 1280x720 [SAR 3:4 DAR 4:3]
[19:42:15 CEST] <Teduardo> Howdy, is there a way to automatically cut the first frame off of a video?
[19:42:56 CEST] <Tom_B> What commands do I need to convert dvb_teletext to ass?  Presumably I need something like -vf "ass=[0:2]"
[19:43:22 CEST] <kepstin> Teduardo: assuming you're re-encoding, you can use "-vf trim=start_frame=1", I think (the first frame is #0)
[19:43:41 CEST] <kepstin> Teduardo: that doesn't update the audio tho, so it might slightly desync audio.
[19:43:50 CEST] <Teduardo> there is no audio
[19:44:15 CEST] <Teduardo> so just specify the input as the current file and the output as the new file and -vf trim=start_frame=1 and that should possibly work?
[19:44:18 CEST] <Teduardo> thanks
[19:44:59 CEST] <kepstin> Teduardo: should work, yes. Keep in mind that it will require re-encoding the video, with the associated generational quality loss.
[19:45:12 CEST] <kepstin> (you can't remove arbitrary frames from video without re-encoding)
[19:45:26 CEST] <Teduardo> ah. it sucks that I can't just fix the flips in premier quickly
[19:45:29 CEST] <Teduardo> err, clips
[19:45:50 CEST] <Teduardo> someone cut a long video into clips but they cut it so that the last frame of the previous clip is in the next clip
[19:45:59 CEST] <Teduardo> because they didnt know what they were doing
[20:20:04 CEST] <alexpigment> Teduardo: you may want to look into VideoRedo. It does "smart rendering", where it only re-renders the GOPs right around the cut points. I use it every day and consider it easily worth the cost
[20:39:52 CEST] <bf_> Good evening everyone
[20:41:10 CEST] <bf_> Is it advised to use ffmpeg to merge 10+ audio streams into a single audio stream with some equalizer fiddling?
[20:41:40 CEST] <bf_> I can't think of a better tool for the job, but I'd like to ask you guys first before finding out the caveats myself
[20:41:51 CEST] <JEEB> there's plenty of audio filters
[20:42:01 CEST] <JEEB> so in that sense you should be able to do what you want
[20:43:01 CEST] <bf_> Is it a good thing to do it with ffmpeg or would you advise a different tool?
[20:43:15 CEST] <durandal_1707> paid pro tool
[20:43:34 CEST] <bf_> I'd like to roll my own and learn :)
[20:44:38 CEST] <JEEB> bf_: I don't see any reason why you wouldn't be able to do it
[20:44:38 CEST] <bf_> which paid tool do you have in mind durandal_1707
[20:44:50 CEST] <JEEB> you just need to know what exactly you'd be doing :P
[20:45:21 CEST] <bf_> so okay please don't laugh :)
[20:45:39 CEST] <bf_> I want to build this thing over the browser where up to 1000 people can speak into their mics simultaneously
[20:45:44 CEST] <durandal_1707> bf: only very expensive one, but if ffmpeg can do it and do it fast, there you go
[20:45:58 CEST] <bf_> order the audio sources in space via equalizer so you it feels like a stadium
[20:46:20 CEST] <durandal_1707> ugh
[20:46:43 CEST] <bf_> I know it won't scale but I want to build a prototype to show it potentially works
[20:47:45 CEST] <durandal_1707> you can use amix ffmpeg filter to mix multiple sources easy way
[20:49:17 CEST] <bf_> thanks for your input, I will try it
[21:38:34 CEST] <bencc1> are aac and libx264 suitable formats for hls?
[21:38:35 CEST] <bencc1> '-acodec', 'aac', '-strict', 'experimental', '-b:a', '96k',
[21:38:36 CEST] <bencc1> '-vcodec', 'libx264', '-pix_fmt', 'yuv420p', '-preset:v', 'veryfast' ,'-crf', '23',
[21:39:07 CEST] <bencc1> I'm trying to capture the screen and output hls
[21:39:46 CEST] <furq> they're the only codecs you can use with hls
[21:40:07 CEST] <bencc1> furq: what I pasted above?
[21:40:54 CEST] <alexpigment> bencc1 - i don't think -strict and experimental are necessary for any recent versions of FFMPEG
[21:41:55 CEST] <bf_> I have another stupid question. I am unable to find out if ffmpeg can listen on a port for an incoming rtmp connection and convert it to ffmpeg
[21:41:59 CEST] <bencc1> alexpigment: I'm using ffmpeg on ubuntu: 2.8.11
[21:42:05 CEST] <bencc1> recent enough?
[21:42:14 CEST] <bf_> On the mailing list I found something like "nc -l 8888 | ffmepg -i pipe:0 ..." but it ain't working
[21:42:19 CEST] <bf_> Is there a recommended way for it?
[21:42:25 CEST] <alexpigment> not sure, although i'd update if you can. the AAC encoder is much better these days than it was a year ago or so
[21:42:41 CEST] <furq> you need 3.0 or above for the new aac encoder
[21:43:17 CEST] <furq> https://www.johnvansickle.com/ffmpeg/ if you don't want to build it yourself
[21:43:31 CEST] <bencc1> alexpigment: better in what sense?
[21:43:42 CEST] <bencc1> alexpigment: what version you'll consider recent enough?
[21:43:54 CEST] <alexpigment> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=d9791a8656b5580756d5b7ecc315057e8cd4255e
[21:43:57 CEST] <bencc1> 3.0.7 is good?
[21:44:13 CEST] <JEEB> just go with the latest unless you have a reason not to
[21:44:32 CEST] <JEEB> (if you need some sort of check just check fate.ffmpeg.org for your OS/arch)
[21:44:48 CEST] <furq> don't install from a ppa if that's what you're thinking of doing
[21:45:15 CEST] <bencc1> I prefer to install from the official ubuntu repo
[21:45:22 CEST] <bencc1> isn't this good enough?
[21:45:22 CEST] <bencc1> http://packages.ubuntu.com/yakkety/ffmpeg
[21:45:40 CEST] <furq> are you on yakkety
[21:46:17 CEST] <furq> i assume you're on xenial if you have 2.8.11, in which case don't install shit from other repos
[21:46:30 CEST] <furq> either do a full upgrade for 3.x or just use those static builds
[21:46:56 CEST] <bencc1> furq: I'm using ffmpeg inside a docker container
[21:46:58 CEST] <bf_> Is there an alternative to using nginx-rtmp for rtmp server? I have a hard time to find something that listens for incoming RTMP stream and calls ffmpeg to convert to HLS
[21:47:11 CEST] <bencc1> so I don't mind upgrading to yakkety if needed
[21:47:21 CEST] <furq> well yeah that's fine then
[21:47:34 CEST] <JEEB> at that point just go to 17.04 as that is pretty much done
[21:47:40 CEST] <furq> bf_: what's wrong with nginx-rtmp for that
[21:47:42 CEST] <JEEB> and has an even newer thing
[21:47:51 CEST] <bencc1> JEEB: good point
[21:47:54 CEST] <furq> that's more or less the main use case for it
[21:48:02 CEST] <BtbN> nginx-rtmp can convert to HLS on its own.
[21:48:14 CEST] <furq> yeah
[21:48:26 CEST] <bf_> furq: there might be something more lightweight? I just want to convert incoming stream to HLS, and don't want NGINX to serve the hls files
[21:48:38 CEST] <bf_> hls files -> cdn
[21:48:40 CEST] <furq> i'm not aware of anything more lightweight
[21:48:54 CEST] <bf_> furq: okay, thank you
[21:48:56 CEST] <furq> you can build ffmpeg without any of the http modules if you really don't want to use it
[21:49:03 CEST] <bf_> ffmpeg?
[21:49:06 CEST] <furq> er
[21:49:06 CEST] <furq> nginx
[21:49:06 CEST] <bf_> you mean nginx?
[21:49:09 CEST] <bf_> ok
[21:49:46 CEST] <bf_> https://github.com/arut/nginx-rtmp-module/issues -> 500 open issues o.o
[21:50:01 CEST] <JEEB> there's a fork that's more active
[21:50:01 CEST] <furq> https://github.com/arut/nginx-rtmp-module/issues/985
[21:50:03 CEST] <furq> 499 open issues
[21:50:04 CEST] <bf_> I heard twitch is using a go server for incoming streams
[21:50:17 CEST] <furq> a lot of these appear to be feature requests
[21:50:24 CEST] <bf_> anybody know if some parts of that are open sourced?
[21:50:31 CEST] <furq> although yeah it's pretty inactive
[21:50:37 CEST] <furq> i expect most of those issues would have been closed otherwise
[21:50:58 CEST] <bf_> yes, even though arut seems to be employed at nginx
[21:51:48 CEST] <bf_> Ok so this proprietary go rtmp listener is twitch's secret sauce
[21:55:07 CEST] <utack> is it possible that benchmark+seek uses the real time starting when ffmpeg launches, instead of actually going to the point it should seek to and then start the benchmark?
[22:56:05 CEST] <ogopogo> What is the command to remove black bars from a .mp4 file?
[23:02:34 CEST] <Darxus> ogopogo: There's a command for that?
[23:09:44 CEST] <pbos> Possibly-wrong channel, but does anyone happen to know if openh264/h264enc is usable without welsenc.cfg config files?
[23:09:53 CEST] <pbos> Getting: [OpenH264] this = 0x0x203c010, Error:Invalid settings in bitrate. the sum of each layer bitrate(2236000) is larger than total bitrate setting(100000)
[23:10:14 CEST] <pbos> which is because I haven't specified multiple layers, hoping it would just ignore that.
[23:11:03 CEST] <pbos> don't see any example use of this app online, so it might just be "guh no one uses it"
[23:15:50 CEST] <pbos> Oh, it just hardcodes defaults implicitly even though I haven't provided anything. That's scary.
[23:17:38 CEST] <thebombzen> I'm getting a slight audio desync in mpv
[23:17:40 CEST] <thebombzen> compared to ffplay
[23:20:26 CEST] <thebombzen> oops wc
[23:22:58 CEST] <ader> I have written simple video player, which is working fine when I play audio and video, but when I disable audio play codes video plays too fast. So my queston is, is it the responsibilty of the programmer to set play timer or does library take care with dts and pts?
[23:25:08 CEST] <BtbN> I'd assume it's entirely up to you how fast you show the pictures
[23:29:48 CEST] <ader> Thanks, So you are saying I have to take care of delay with some timer and make thread sleep for while to slow playing speed right
[23:39:58 CEST] <kepstin> yep. normally when you have audio, the sound card limits your playback rate because it plays the audio in realtime. but if you're not playing audio, you have to time the video to some other clock.
[23:46:35 CEST] <bf_> furq: JEEB: durandal_1707 I have tried out nginx-rtmp. Unfortunately it is very buggy.
[23:47:46 CEST] <BtbN> Works for me.
[23:48:44 CEST] <furq> same
[23:59:51 CEST] <vader-> hmmm strange issue... so I am capturing my video on a Windows 10 machine, with a Blackmagic H.264 Recorder. The video looks correct on the computer, during capture, during playback in both Windows Media Player and Quicktime. When I transfer that video to another computer is comes up really dark. I have tried playing in Windows media player, Quicktime, VLC,
[23:59:51 CEST] <vader-> etc. Always comes up dark... Any thoughts?
[00:00:00 CEST] --- Tue Mar 28 2017