[Ffmpeg-devel-irc] ffmpeg.log.20170406

burek burek021 at gmail.com
Fri Apr 7 03:05:01 EEST 2017


[00:00:20 CEST] <Tatsh> smart flags maybe in the future: so if you do -preset veryslow with libx264, and you leave that as is with -c:v h264_nvenc, it automatically converts :)
[00:00:47 CEST] <Tatsh> i'm so used to doing libx264 with SD media so i don't mind the slow speed there, but these are 1080p and i didn't like going at 1/5 the speed
[00:31:54 CEST] <stryder> hi I have a question about encoding audio into AAC 4.0 on Android
[00:32:21 CEST] <stryder> I set my channel count to 4
[00:32:31 CEST] <stryder> and my channel layout to AV_CH_LAYOUT_4POINT0
[00:33:07 CEST] <stryder> but when I try to configure my filter graph (with that as an output) I get
[00:33:08 CEST] <stryder> FFMPEG Error Level:16 Log: Output channel layout '5 channels (FL+FR+BR+FLC+SL)' is not supported
[00:48:56 CEST] <bre> Hello. Is it possible to have conditions in filter value expressions? For example, setpts=PREV_OUTPTS + 1 if PTS - PREV_INPTS > 1 else PTS - PREV_INPTS + PREV_OUTPTS.
[00:53:15 CEST] <bre> Normally filter value expressions evaluate to a single number. Is it possible to have an expression that includes checking a condition and the value is calculated differently depending on the condition?
[01:16:27 CEST] <DHE> bre: see https://www.ffmpeg.org/ffmpeg-all.html#Expression-Evaluation
[01:16:35 CEST] <DHE> yes, there are simple if/else formula options available
[01:22:08 CEST] <bre> DHE: That is fantastic. Thanks very much
[02:06:11 CEST] <bre> Filtering and streamcopy cannot be used together
[02:08:24 CEST] <bre> Is it possible to change frame time stamps without changing the frame pixels?
[02:09:06 CEST] <bre> I'm working on a VP9 video in case it makes a difference
[02:18:02 CEST] <bray90820> Slightly off topic but is there a way to check if a disc is encripted on OSX
[02:51:20 CEST] <Tatsh> i put -crf 21 on h264_nvenc, terrible
[02:51:26 CEST] <Tatsh> with -preset slow
[02:52:33 CEST] <furq> neither of those options affect nvenc
[02:55:01 CEST] <AndrewMock> Is it possible to display local time using drawtext with milliseconds?
[02:55:21 CEST] <AndrewMock> I can show linear timecode with milliseconds, but not the current time.
[02:57:06 CEST] <Tatsh> furq, preset slow does but not crf
[02:57:29 CEST] <Tatsh> i've been trying to figure out good equivalents to what i like with libx264 but not getting anything
[03:58:51 CEST] <DHE> Tatsh: there's a -cq which I think is the closest thing to crf you'll get
[03:59:07 CEST] <DHE> haven't tried it though. just reading the codec fields
[03:59:52 CEST] <Tatsh> yea i'm going to experiment with it
[04:00:06 CEST] <Tatsh> right now for this content which isn't from a great camera i'm testing various -crf qualities
[04:00:22 CEST] <Tatsh> the camera saves to MJPEG format in AVI container
[04:00:45 CEST] <Tatsh> it's pretty low-bitrate even on 1080
[06:06:51 CEST] <vader-> Uhg... So pissed my archiving project was going so well. I was about 50 tapes into about 500 and my VCR decided to act up... Getting crazy lines and stuff. I think the capacitors need to be replaced.
[09:57:47 CEST] <AndrewMock> Is it possible to display local time with milliseconds using drawtext?
[11:09:21 CEST] <Kadigan> Hey... Can I use ffmpeg to insert an empty subtitle stream? I have a number of video files that have English audio w/ English subs, and I'd like to keep the subs but NOT have them be the default. I realize I could change this behavior in my video player, but my player doesn't allow me to differentiate decisions based on audio streams present (have this configured for LAV but not everything uses it)
[11:09:45 CEST] <Kadigan> That way I could add an empty sub stream and -then- the original sub stream as an additional stream (non-default)
[11:18:31 CEST] <furq> Kadigan: printf "001\n00:00:00,000 --> 00:00:00,000" > empty.srt; ffmpeg -i foo.mp4 -i empty.srt ...
[11:26:03 CEST] <wh> #0  nppscale_filter_frame (link=link at entry=0x34b38c0, in=0x34c0340) at libavfilter/vf_scale_npp.c:496
[11:26:03 CEST] <wh> 496         AVCUDADeviceContext *device_hwctx = frames_ctx->device_ctx->hwctx;
[11:26:03 CEST] <wh> [Current thread is 1 (Thread 0x7fc45705e740 (LWP 14602))]
[11:26:03 CEST] <wh> (gdb) bt
[11:26:03 CEST] <wh> #0  nppscale_filter_frame (link=link at entry=0x34b38c0, in=0x34c0340) at libavfilter/vf_scale_npp.c:496
[11:26:04 CEST] <wh> #1  0x00000000004b658f in ff_filter_frame_framed (frame=0x34c0340, link=0x34b38c0) at libavfilter/avfilter.c:1116
[11:26:07 CEST] <wh> #2  ff_filter_frame_to_filter (link=0x34b38c0) at libavfilter/avfilter.c:1264
[11:26:09 CEST] <wh> #3  ff_filter_activate_default (filter=<optimized out>) at libavfilter/avfilter.c:1315
[11:26:13 CEST] <wh> #4  ff_filter_activate (filter=<optimized out>) at libavfilter/avfilter.c:1476
[11:26:15 CEST] <wh> #5  0x00000000004b9ebc in ff_filter_graph_run_once (graph=graph at entry=0x3af41c0) at libavfilter/avfiltergraph.c:1446
[11:26:18 CEST] <wh> #6  0x00000000004bae2c in push_frame (graph=0x3af41c0) at libavfilter/buffersrc.c:181
[11:26:20 CEST] <wh> #7  av_buffersrc_add_frame_internal (ctx=ctx at entry=0x34b2e20, frame=frame at entry=0x3acbce0, flags=flags at entry=4) at libavfilter/buffersrc.c:262
[11:26:23 CEST] <wh> #8  0x00000000004bb51d in av_buffersrc_add_frame_flags (ctx=0x34b2e20, frame=frame at entry=0x3acbce0, flags=flags at entry=4) at libavfilter/buffersrc.c:164
[11:26:26 CEST] <wh> #9  0x0000000000498faf in ifilter_send_frame (frame=0x3acbce0, ifilter=0x348a4a0) at ffmpeg.c:2206
[11:26:28 CEST] <wh> #10 send_frame_to_filters (ist=ist at entry=0x366a460, decoded_frame=decoded_frame at entry=0x3acbce0) at ffmpeg.c:2284
[11:26:31 CEST] <wh> #11 0x0000000000499776 in decode_video (ist=ist at entry=0x366a460, pkt=pkt at entry=0x7ffe2219e8c0, got_output=<optimized out>, eof=<optimized out>,
[11:26:34 CEST] <wh>     decode_failed=<optimized out>) at ffmpeg.c:2484
[11:26:36 CEST] <wh> #12 0x000000000049af33 in process_input_packet (ist=0x366a460, pkt=0x7ffe2219eb00, no_eof=0) at ffmpeg.c:2633
[11:26:39 CEST] <wh> #13 0x0000000000479dba in process_input (file_index=<optimized out>) at ffmpeg.c:4407
[11:26:43 CEST] <wh> #14 transcode_step () at ffmpeg.c:4518
[11:26:45 CEST] <wh> #15 transcode () at ffmpeg.c:4572
[11:26:47 CEST] <wh> #16 main (argc=<optimized out>, argv=<optimized out>) at ffmpeg.c:4777
[11:26:49 CEST] <wh> anybody seem this before ?
[11:33:22 CEST] <Kadigan> furq: thanks :)
[11:35:10 CEST] <Kadigan> furq: can it be Windows newlines if I'm doing it on Windows? ie. \r\n - or does it have to be \n ?
[11:36:37 CEST] <furq> either's probably fine
[11:37:22 CEST] <Kadigan> So basically, if I have say: video audio[eng] audio[fre] audio[eng] subs[eng], and I want to drop the french audio but keep the subs as non-default, "-i empty.srt -map 0:0 -map 0:1 -map 0:3 -map 1:0 -map 0:4" ?
[11:37:52 CEST] <Kadigan> should net me video audio[eng] audio[eng] subs[empty] subs[eng]
[11:38:23 CEST] <furq> you might need to use -disposition as well
[11:38:56 CEST] <furq> you can also use -map -0:m:language:eng
[11:40:05 CEST] <furq> -disposition:s:0 0 should make the first sub stream non-default fwiw
[11:40:10 CEST] <furq> if that works with your player
[11:40:38 CEST] <furq> otherwise add the empty stream and -disposition:s:0 default -disposition:s:1 0
[11:41:14 CEST] <Kadigan> that relating to output, yes?
[11:41:18 CEST] <furq> yeah
[11:41:46 CEST] <Kadigan> what does -map 0:m mean?
[11:42:26 CEST] <Kadigan> (I'm building the stream list based on parsed output from ffprobe, so I'm not using ffmpeg to make the decisions)
[11:51:07 CEST] <Kadigan> Ouch, I don't have -disposition.
[11:52:28 CEST] <thebombzen> Kadigan: I don't think you should use -map 0:m
[11:53:33 CEST] <Kadigan> I haven't used it because I don't know what it means.
[11:54:00 CEST] <thebombzen> interesting
[11:54:15 CEST] <Kadigan> Still, I can't get the original subtitles to show after I'm done.
[11:54:23 CEST] <thebombzen> hm so normally ffmpeg uses stream specifiers with map
[11:54:27 CEST] <thebombzen> but it defies this convention with m
[11:54:34 CEST] <Kadigan> They still default to stream 1 (out of 3), but there's nothing there...
[11:54:53 CEST] <thebombzen> m is "metadata" so by using -map 0:m:language:eng you're saying "map streams with the metadata key/value pair langauge=eng"
[11:55:13 CEST] <Kadigan> Ah, I see.
[11:55:25 CEST] <Kadigan> So no, I don't think I want that.
[11:55:49 CEST] <thebombzen> You probably want -map 0 -map -0:m:language:fre
[11:56:09 CEST] <thebombzen> sorry not that
[11:56:09 CEST] <thebombzen> I mean
[11:56:29 CEST] <thebombzen> you probably want -map 0 -map -0:a -map 0:a:language:eng
[11:56:41 CEST] <thebombzen> what that does is it maps all streams except non-english-language-audio
[11:56:47 CEST] <Kadigan> Like I said, I'm not using ffmpeg to make the decisions.
[11:56:59 CEST] <Kadigan> I dump the streams using ffprobe and parse the output to select the streams I want.
[11:57:42 CEST] <Kadigan> Namely, if first inserted language is not English, I simply map the subs as-is, but if the first language -is- English, I add the empty.srt and insert it first.
[11:57:58 CEST] <Kadigan> Seems my ffmpeg doesn't know about disposition, so I can't set default streams.
[11:59:00 CEST] <Kadigan> Still, I don't understand why the subs that are originally there don't seem to insert back in.
[11:59:22 CEST] <thebombzen> well what I said was wrong so whatever
[11:59:25 CEST] <thebombzen> upon testing
[12:00:00 CEST] <Kadigan> Thanks for trying though. Always appreciate the willingness to help, regardless of outcome. :)
[12:09:25 CEST] <furq> if you don't have -disposition then probably update ffmpeg
[12:17:15 CEST] <wh> hello , anybody tested Full hardware transcode with CUVID and NVENC ?
[12:18:33 CEST] <furq> what a nice young man
[12:21:52 CEST] <Kadigan> furq: yeah, but I'm on Windows, and I want the non-free (libfdk_aac for example), so I either need to build it myself (on Windows -- not recommended :D) or wait for someone to make and distribute a Windows build w/ them.
[12:22:12 CEST] <JEEB> well enable-nonfree means that the binary cannot be redistributed
[12:22:30 CEST] <JEEB> in which case you go on testing the internal aac encoder as it was improved in late 2015/early 2016
[12:22:37 CEST] <Kadigan> Yes, I know. But some people actually do it. And don't get me wrong - I'm happy that they do.
[12:22:52 CEST] <JEEB> and building on windows really isn't that hard
[12:22:58 CEST] <furq> i should really push all these updates to my mingw build thingy
[12:23:00 CEST] <Kadigan> I've tried 5 times, and I'm 0 for 5.
[12:24:17 CEST] <Kadigan> I'm currently in the slow-progressing process of setting up a multitarget build environment on a Debian VM, because I've a number of things I'd like to build, and not just for Windows. But until that gets done...
[12:24:22 CEST] <furq> it's much easier to build in a vm
[12:24:59 CEST] <JEEB> ok, start with empty msys2 with git,yasm,make,diffutils (and whatever package in msys2 contains autotools such as automake and autoconf) and then extract nev's toolchain (that is tested enough with his FATE and his LAV Filters) and add it to PATH
[12:25:02 CEST] <Kadigan> Well, for starters, I have -no- idea how to build against a specific target, or with specific options. I only know "make". :) And it's outside the scope of this channel anyway.
[12:25:06 CEST] <JEEB> that should make it work
[12:25:37 CEST] <furq> Kadigan: https://github.com/qruf/ffmpeg-mingw
[12:25:56 CEST] <JEEB> the last thing I mentioned is available from https://files.1f0.de/mingw/
[12:26:10 CEST] <furq> some of the package versions are a bit out of date though
[12:26:20 CEST] <JEEB> but yea, your problem most likely is not building for windows but in generally building stuff :P
[12:26:21 CEST] <Kadigan> Okay, fun fact: the thing doesn't transplant the subtitles properly. :/ At all. They work in the original file, they don't work in the processed one.
[12:27:07 CEST] <Kadigan> I'm only digging into this because I need to fix VFR on some 120GB worth of media, incidentally. Thought I'd get rid of the french audio while I was at it...
[12:27:17 CEST] <furq> i got stuck figuring out the correct cmake summoning spell to get it to work with --enable-shared
[12:27:22 CEST] <furq> er
[12:27:24 CEST] <furq> s/it/libgme/
[12:28:04 CEST] <JEEB> Kadigan: also wrt the fdk-aac thing, are you still sure you absolutely positively need it :P
[12:28:06 CEST] <Kadigan> JEEB: that's a hole in one, good sir. Exactly it. I need to learn how to build things on Linux in general.
[12:28:28 CEST] <furq> i should probably take another look at it today. i have a bunch of uncommited changes
[12:28:31 CEST] <JEEB> Kadigan: not limited to *nix, just the general build system ways from build systems originating there
[12:28:31 CEST] <furq> +t
[12:28:32 CEST] <Kadigan> JEEB: well, whenever I encode with -c:a aac -strict -2 I get pops and clicks.
[12:28:39 CEST] <JEEB> well if you need -strict experimental
[12:28:44 CEST] <JEEB> then you have f**cking old thing
[12:28:46 CEST] <furq> if you need -strict -2 then that's the pre-3.0 aac encoder
[12:28:52 CEST] <furq> which sucks
[12:28:59 CEST] <Kadigan> I have two machines I need ffmpeg on.
[12:29:02 CEST] <Kadigan> The first one is Windows,
[12:29:05 CEST] <Kadigan> the other one is a Mac.
[12:29:06 CEST] <Kadigan> :D
[12:29:11 CEST] <furq> try -c:a aac with one of the latest zeranoe builds
[12:29:17 CEST] <Kadigan> For Windows I use pre-built stuff, on Mac I use macports.
[12:29:18 CEST] <JEEB> well with mac you have brew
[12:29:22 CEST] <furq> if that still sucks then yeah, you'll need to build yourself
[12:29:26 CEST] <JEEB> which generally seemed less insane than macports
[12:29:33 CEST] <JEEB> at least the last time I checked
[12:29:40 CEST] <furq> macports is better based on hearsay from osx users i know
[12:29:46 CEST] <JEEB> it's the older thing
[12:29:49 CEST] <Kadigan> JEEB: maybe so, but I don't want to mix brew and ports, lest I end up with an insane headache about what-is-where
[12:29:55 CEST] <JEEB> yea, that's true
[12:29:58 CEST] <JEEB> you'd have to uninstall one
[12:30:04 CEST] <JEEB> to make sure they don't conflict somehow
[12:30:05 CEST] <Kadigan> and I'm using a shitload of stuff from ports already
[12:30:06 CEST] <Kadigan> :)
[12:30:24 CEST] <furq> i take it macports lets you build with fdk
[12:30:42 CEST] <furq> i've never touched osx but i assume it's some kind of freebsd ports clone
[12:31:32 CEST] <ritsuka> on mac you can just use the apple aac encoder
[12:32:12 CEST] <Kadigan> ritsuka: maybe. I never dug that deep into it. All I know is that I needed to encode a bunch of deliveries, incl. TV, and ffmpeg (and then ffmbc) came to the rescue.
[12:32:17 CEST] <furq> oh right yeah you can build with audiotoolbox on osx
[12:32:20 CEST] <Kadigan> Have been using them for 4+ years in that role.
[12:34:17 CEST] <Kadigan> Truth be told, I'm ultimately wanting a one-stop-shop for all my transcoding needs on Windows. I have one format I use (h264/x264 w/ AAC or AC3 in MKV) and I'd like all of my media to one day use that specific format. I have a lot of stuff, incl. anime, incl. dual-audio stuff which I -sometimes- use, but most of the time I need to strip off the english audio, OR the english subs (if the audio is in english)
[12:34:41 CEST] <Kadigan> so I'm trying to build a good decision-making script that'd analyze the files and choose the correct processing paths.
[12:34:58 CEST] <Kadigan> Something I could sic at my media library recursively and check up on a week later. :)
[12:35:37 CEST] <Kadigan> And the thing with anime is that it quite often uses variable frame rate, and I need a smooth rate for SVP. So yeah, specific needs.
[12:36:04 CEST] <Kadigan> (SVP drops the buffer each time framerate changes, and if it happens at each scene change, I get video jerks every few seconds, which is unacceptable)
[12:53:30 CEST] <Kadigan> Well, I guess I'll get to it in earnest when I have an up-to-date version of ffmpeg. :) Thank you all for contributing.
[12:57:31 CEST] <Kadigan> Different question:
[12:58:00 CEST] <Kadigan> can I somehow hide all of the stuff ffmpeg shows (similar to -loglevel error) but still have updates on encoding?
[12:58:59 CEST] <furq> -v error -stats
[13:01:00 CEST] <Kadigan> Thank you kindly.
[13:20:39 CEST] <Threads> <JEEB> and building on windows really isn't that hard
[13:20:49 CEST] <Threads> not hard but it takes a while to compile the exe
[14:35:55 CEST] <Dan39> hi \o
[14:36:25 CEST] <Dan39> i am curious, does ffplay use a specific output video driver? like with mplayer i specify xv..
[14:36:37 CEST] <Dan39> and can i specify which to use?
[14:37:14 CEST] <durandal_1707> nope, it usses only what sdl uses
[14:37:47 CEST] <durandal_1707> see sdl manual about environmental variables
[14:41:44 CEST] <Dan39> had a feeling that was the case, thanks
[14:45:01 CEST] <Shinao> hello, trying to convert list of imgs to video, I get a black screen - ffmpeg -framerate 60 -i shot%03d.png outputimgs.mp4
[14:49:06 CEST] <Shinao> Okey fixed my black screen by updating ffmpeg to latest version, adding option -pix_fmt yuv420p and adding option -vf "scale=trunc(iw/2)*2:trunc(ih/2)*2" because there is another problem. damn that's not really easy for such basics needs
[14:49:57 CEST] <Kadigan> Shinao: ffmpeg was made to be as versatile as possible, and some options may end up being convoluted because of all the added functionality. :]
[14:50:12 CEST] <Kadigan> At least that is my experience and the view of the matter.
[14:52:02 CEST] <Shinao> I suppose all the possibility makes it worth it but just to convert images to a video ? meh.
[14:52:56 CEST] <Kadigan> Actually, I think that's not complex at all. Specifying yuv420 as the color space is necessary for a number of players (notably quicktime) which are broken and can't accept more modern formats.
[14:53:27 CEST] <Kadigan> TVs have that issue, too.
[14:54:53 CEST] <Kadigan> As for the images and video size -- if you want picture-perfect quality, prepare your images to be of size required beforehand - it's what I do, anyway. That way you will have complete control over how they're framed into the video. I've never had issues converting 4k by 3k photos into video, to be clear.
[14:54:54 CEST] <Shinao> I tried VLC, media player and another random one. But even after that there is some problem with its scale. Anyway still great soft
[14:55:48 CEST] <Kadigan> In any case, even if they don't play back properly, they would still be properly processed by NLEs (video editor software), to produce the final, standardized output.
[15:06:39 CEST] <Dan39> why does ffplay move my mouse? :| is that SDL doing it?
[15:14:56 CEST] <Dan39> got it, set SDL_NOMOUSE
[15:18:50 CEST] <Dan39> i take that back, maybe i dont. forgot i wanted to fullscreen it...
[15:20:22 CEST] <Dan39> yea when fullscreening it still moves the mouse on me -_-
[15:24:52 CEST] <Kadigan> You know... There was a joke that a guy invited a gf over to watch some movies, but the screen shut off every 15min while watching, and the guy decided to move the mouse every so often to prevent that. He ended up coding half the night to make it do it automatically (she left mid-coding, of course)... Maybe there's some truth to this joke after all? ;]
[15:26:17 CEST] <Dan39> ha
[15:26:30 CEST] <Dan39> it doesnt "keep moving it"
[15:26:46 CEST] <Dan39> it just moves it to the screen that the video is playing on when it first starts
[15:26:58 CEST] <Kadigan> Well, that's more specific.
[15:27:17 CEST] <Kadigan> I don't know how that works on Linux, but on Windows full-screen apps tend to 'snatch' the mouse.
[15:27:22 CEST] <Kadigan> (when they enter full-screen)
[15:27:31 CEST] <Dan39> which i actually have setup as a seperate screen that the mouse cant move over to
[15:27:37 CEST] <Dan39> so once it grabs the mouse, its stuck over there
[15:27:44 CEST] <Kadigan> Windows?
[15:27:47 CEST] <Dan39> linux
[15:27:50 CEST] <Kadigan> Hm.
[15:28:00 CEST] <Dan39> mplayer -fs doesnt grab the mouse on me
[15:28:06 CEST] <Kadigan> That's something I've come to expect from full-screen focus-exclusive apps, like games.
[15:28:13 CEST] <Dan39> but then i believe mplayer doesnt use the mouse at all unlike ffplay
[15:29:35 CEST] <Kadigan> Yeah, from what I can tell from a cursory Google search is that the solution is not to use ffplay. ;]
[15:29:50 CEST] <Dan39> haha
[15:30:02 CEST] <Kadigan> (ie. use mplayer)
[15:30:05 CEST] <Dan39> yea
[15:30:06 CEST] <furq> use mpv
[15:30:12 CEST] <furq> ffplay is for debugging
[15:30:20 CEST] <Dan39> thats what i was kind of using it for
[15:30:22 CEST] <Dan39> debugging
[15:30:51 CEST] <Kadigan> Seems the coders of ffplay have a definition of "debugging" that makes assumptions different from your own, then.
[15:31:06 CEST] <Dan39> im having issues with audio crashing. mplayer outputs nothing even with highest log level
[15:31:13 CEST] <Dan39> ffplay at least throws something
[15:31:51 CEST] <Dan39> but yea, dont worry about it :P
[15:31:58 CEST] <Dan39> i want to use mpv
[15:32:18 CEST] <Dan39> if i have time to set it up on this install..
[15:32:36 CEST] <Dan39> doesnt seem there are any centos7 packages for mpv
[16:45:13 CEST] <joinx> hi team
[16:45:33 CEST] <joinx> i need help with "joining two mp4/h264 files into single one"
[16:45:48 CEST] <joinx> i'm using the concat feature, but... i don't know why, only the first video is player
[16:46:19 CEST] <joinx> im using the simple concat demuxer
[16:46:44 CEST] <joinx> the output file has the apparently great size, but. the content is only played for the first video
[16:47:16 CEST] <joinx> any idea?
[16:47:39 CEST] <joinx> (i googled a lot ... and i don't want to re-encoding to mpeg temporary file..)
[16:48:09 CEST] <joinx> cbote at b3:~/test$ file video.mp4 texto.mp4  video.mp4: ISO Media, MPEG v4 system, version 1 texto.mp4: ISO Media, MPEG v4 system, version 1
[16:48:29 CEST] <joinx> both files are similar with the same codec, and both are generated with ffmpeg.. so , i don't know what's happening
[16:48:54 CEST] <joinx> thank you
[16:49:28 CEST] <dystopia_> are they the same spec exactly?
[16:49:36 CEST] <dystopia_> same resolution, same codec, same settings?
[16:49:37 CEST] <joinx> generated with the same options
[16:49:40 CEST] <dystopia_> ok
[16:49:55 CEST] <dystopia_> i dump files i want to cat to a text file with a for loop
[16:50:15 CEST] <dystopia_> (for %%i in (*.mp4) do @echo file '%%i') > cat.txt
[16:50:31 CEST] <dystopia_> ffmpeg -f concat -i cat.txt -c copy outfile.mp4
[16:50:42 CEST] <dystopia_> you could give it a try
[16:51:24 CEST] <joinx> let me try
[16:56:15 CEST] <joinx> it works dystopia!
[16:56:31 CEST] <joinx> i don't know why, but it works
[16:56:38 CEST] <joinx> thank you very much
[16:56:55 CEST] <dystopia_> :)
[16:57:04 CEST] <joinx> wow, irc still works!!!
[16:57:16 CEST] <joinx> god save netsplits :D
[16:57:45 CEST] <joinx> please keep pushing, it's great to get younger again ;-)
[16:57:48 CEST] <joinx> see ya!
[17:09:02 CEST] <Kadigan> Sooo... I know ffprobe will show me the -average- framerate of a file. Can I however use ffmpeg to determine all of the VFR changes in a file? As in, truly detect VFR.
[17:09:15 CEST] <Kadigan> Well, ffmpeg or ffprobe.
[17:10:21 CEST] <Kadigan> If I could make a nice graph out of it in the end (so pairs of time+new fps?), it would actually help I suppose.
[17:10:35 CEST] <DHE> it would require scanning the whole file. which ffprobe can do but is not the default
[17:10:51 CEST] <DHE> I think you'd have to run with -show_frames and search for duration or pts inconsistencies.
[17:11:22 CEST] <Kadigan> I don't mind scanning the whole file,
[17:11:43 CEST] <Kadigan> it -must- be faster than decoding the whole file into frames, reencoding frames into video-only track and then remuxing w/ the original audio.
[17:11:55 CEST] <Kadigan> I want to detect which files are -not- VFR so I can skip the whole processing for them.
[17:27:04 CEST] <kepstin> Kadigan: also note that some file formats with fixed timebase (e.g. matroska/webm) are really always vfr - the best you can do is check if the frames are mostly constant rate modulo jitter.
[17:31:07 CEST] <Kadigan> Oh... That I didn't know...
[18:15:57 CEST] <heeen> hi, I am looking for libavutil/hwcontext.h, but the ffmpeg tree I'm looking at does not have it
[18:16:06 CEST] <heeen> seems to be 3.0
[18:17:11 CEST] <JEEB> always when updating check your os and architecture from fate.ffmpeg.org
[18:17:41 CEST] <JEEB> that lets you know if the automated tests pass with the latest heas
[18:17:45 CEST] <JEEB> *head
[19:10:47 CEST] <Kadigan> Okay... I'm really at a loss. So I have a video file which, according to ffdshow, is encoded at 23.976fps, but from time to time it switches to 24.something - for a split second, mind. It goes up to 24, then back to 23, and this happens within the space of a few frames or so.
[19:11:19 CEST] <Kadigan> Unfortunately, it's enough for the playback layer to drop the entire pre-buffer and buffer again, which causes a very noticeable stall for a second or so.
[19:11:37 CEST] <Kadigan> Now, this happens every few seconds, so I'm slowly building up Krogan-style blood rage here.
[19:12:07 CEST] <Kadigan> I thought by forcing the video into separate frames and then re-combining again at a known rate, I would end up forcing a constant rate...
[19:12:27 CEST] <Kadigan> This is normal for MKV, you say? And yet, there are other video files where this doesn't happen.
[19:12:35 CEST] <Kadigan> (also Matroska)
[19:12:44 CEST] <Kadigan> Is there -anything- I can do, at all, to fix this?
[19:17:49 CEST] <Kadigan> Okay, scratch that.
[19:18:16 CEST] <Kadigan> I have just re-tested it w/ the "Movie FPS" OSD entry, and it seems to behave differently between re-starts of the player, even among the same files.
[19:18:25 CEST] <Kadigan> My problem is somwhere at the "feeding into the decoder" stage.
[19:18:43 CEST] <Kadigan> somewhere*
[19:19:10 CEST] <Kadigan> (or even earlier; in any case, it must have to do with some variable controlled by the system, and it does not appear to be a fault with the files themselves)
[19:22:47 CEST] <JEEB> well at the point that you're using FFDShow-tryouts you're already out of luck
[19:23:10 CEST] <JEEB> but for exact timestamps you can use mkvinfo for matroska and L-SMASH's boxdumper for mp4
[19:23:19 CEST] <JEEB> those let you view exact values coded in the container
[19:27:31 CEST] <Kadigan> Tryouts?
[19:28:17 CEST] <Kadigan> Unless it's been in the "tryouts" stage for over a year, I'm using a fairly old decoder. Haven't noticed these issues much before today.
[19:28:53 CEST] <Kadigan> But it doesn't matter. The source of the problem is somewhere with my system, and I'll hunt it down.
[19:29:02 CEST] <JEEB> Kadigan: tryouts is the fork that was make after 2006
[19:29:06 CEST] <JEEB> *made
[19:29:10 CEST] <Kadigan> Ah, I see.
[19:29:12 CEST] <JEEB> actual FFDShow has been dead since
[19:29:25 CEST] <Kadigan> So please explain to me why is it that I'm out of luck.
[19:29:32 CEST] <JEEB> and FFDShow-tryouts pretty much dead since ~2011
[19:29:41 CEST] <JEEB> you get a minimal amount of commits after that
[19:30:00 CEST] <Kadigan> What would you propose I use, then? SVP expects an ffdshow video stream.
[19:30:16 CEST] <JEEB> that's because they've hacked themselves into the Avisynth filtering path
[19:30:29 CEST] <Kadigan> Seems that way, yeah.
[19:30:40 CEST] <JEEB> they do have a vapoursynth thing for mpv IIRC, so you could try that. Otherwise just complain to SVP
[19:30:53 CEST] <JEEB> you're a paying customer, you have the right to say that you want less crappy solution
[19:30:56 CEST] <Kadigan> There's actually an SVP4, but it's pay-for software now.
[19:31:08 CEST] <Kadigan> No, I'm still using the old SVP3, which is free.
[19:31:10 CEST] <JEEB> oh, then just give up
[19:31:14 CEST] <Kadigan> (I don't have the money for SVP4)
[19:31:26 CEST] <JEEB> also you can try the built-in interpolation in mpv
[19:31:33 CEST] <Kadigan> What's mpv?
[19:31:40 CEST] <JEEB> the least retarded mplayer fork
[19:31:49 CEST] <Kadigan> How does it compare to mpc-hc?
[19:32:10 CEST] <JEEB> no graphical interface other than the on-screen controls, heavily customize'able and people can make GUIs with it if they want
[19:32:18 CEST] <Kadigan> Windows port?
[19:32:23 CEST] <JEEB> it's cross-platform
[19:32:30 CEST] <Kadigan> Might give it a try, then.
[19:32:34 CEST] <JEEB> and windows is a first-level citizen
[19:32:40 CEST] <JEEB> or however you say that
[19:32:45 CEST] <Kadigan> first-class
[19:32:46 CEST] <Kadigan> Yeah.
[19:32:46 CEST] <JEEB> https://mpv.srsfckn.biz/
[19:33:00 CEST] <JEEB> that guy is on #mpv and thus the least un-trustworthy of those making windows binaries
[19:33:30 CEST] <llogan> plus it's a .biz which adds +2 to trustworthy
[19:33:41 CEST] <Kadigan> Oh sure.
[19:33:46 CEST] <llogan> unlike .horse
[19:34:59 CEST] <Kadigan> SVP seems to demand their own mpv build.
[19:35:15 CEST] <JEEB> mostly because nobody builds it with vapoursynth :P
[19:35:18 CEST] <Kadigan> (since apparently MPV doesn't provide one w/ Vapoursynth
[19:35:20 CEST] <Kadigan> on Windows)
[19:35:35 CEST] <JEEB> mpv itself doesn't provide any binaries
[19:35:55 CEST] <Kadigan> I've already been down the road of using {x}-provided stuff (for example CCCP provides its own mpc-hc, which is a source of a solid number of nightmares)
[19:36:06 CEST] <Kadigan> (worse, it can't be de-selected)
[19:37:30 CEST] <JEEB> umm, I'm pretty sure it should be
[19:37:38 CEST] <JEEB> <- "maintains" CCCP
[19:38:13 CEST] <JEEB> also if you want vanilla you grab vanilla, we have slightly changed defaults (nowadays much less than before, though)
[19:39:53 CEST] <JEEB> https://github.com/cccp/mpc-hc/commits/blitzkrieg (patches for the latest thing I got out)
[19:41:17 CEST] <JEEB> and yes, when changes were globally applicable they went upstream
[20:17:03 CEST] <benbro1> is there a virtual cam/mic device for windows based on ffmpeg?
[20:17:18 CEST] <benbro1> not sure it's related but I don't know where to look
[20:17:31 CEST] <BtbN> Virtual Cam?
[21:43:08 CEST] <benbro1> BtbN: it looks like a usb webcam to other apps
[22:31:56 CEST] <draean> Hi, I'm trying to figure out how to gracefully handle an IP stream that can start and stop. Is there some way to take a live audio stream, like, '-i "http://blah/;stream.mp3"' map it to the audio on my output, and have ffmpeg use it if it's there, and just do black audio if it's not?
[22:40:00 CEST] <draean> here's a command line "ffmpeg -f decklink -i "Intensity Pro at 16" -i "http://198.20.181.70:8000/;stream.mp3" -map 0:1 -map 1:0 -r 30 -c:v libx264 -preset veryfast -c:a libvo_aacenc -f mpegts -flush_packets 0 "udp://<ip_addr>:40000?pkt_size=1316&buffer_size=65535"" when the mp3 stream goes away, I eventually get decklink input buffer overrun, and then get weird behavior after that. I'm wondering if there's a way to get that audio input to gracefull
[22:40:00 CEST] <draean> y handle dropouts
[23:01:27 CEST] <bray90820> How would I do something like this but also copy the subtitle track "ffmpeg -i file.mkv -c:a copy "file.mp4"
[23:15:52 CEST] <klaxa> bray90820: try ffmpeg -i file.mkv -c:a copy -c:s mov_text file.mp4 ?
[23:16:00 CEST] <klaxa> https://en.wikibooks.org/wiki/FFMPEG_An_Intermediate_Guide/subtitle_options
[23:17:17 CEST] <bray90820> klaxa: The subtitles are already in the MKV it's not a separate file
[23:18:05 CEST] <bray90820> could I use something like -map 0?
[23:19:53 CEST] <kerio> bray90820: have you tried it
[23:20:12 CEST] <alexpigment> bray90820: i think the MP4 container (significantly more limited than MKV) only supports mov_text subtitles
[23:20:45 CEST] <alexpigment> as klaxa said earlier, try -c copy -c:s mov_text [file].mp4
[23:21:21 CEST] <bray90820> alexpigment: Do the subtitles have to be a separate file to be able to use that command?
[23:21:23 CEST] <alexpigment> you mentioned that it's not a separate file, but i don't know if that meant you ran into some sort of problem when you tried it
[23:21:40 CEST] <alexpigment> i don't usually mess with subtitles. did you try it or just assume it needed to use a separate file?
[23:25:52 CEST] <alexpigment> does anyone here know how to prioritize quality/bitrate to iframes when encoding with mpeg2video? i have a feeling i could get rid of static-image pulsing if i could make the bitrate spike a little bit on every I frame
[23:26:21 CEST] <klaxa> you can use a custom keyframe interval with -g
[23:26:35 CEST] <klaxa> like -g 60 with 30 fps video will force an i-frame every 2 seconds
[23:26:41 CEST] <alexpigment> i'm trying to stay within the DVD specification, so i'm at -g 18
[23:26:56 CEST] <bray90820> alexpigment: Currently converting as we speak but it's gonna take a little bit
[23:27:04 CEST] <alexpigment> i know other encoders don't have this problem with the same settings, but bitrate viewer shows little spikes here and there, which i assume are on the I frames
[23:27:27 CEST] <klaxa> bray90820: you can test a short segment by using -t 60 or something similar
[23:27:41 CEST] <alexpigment> bray90820: if you do -c copy, it'll copy both the video and audio (rather than re-encoding). that'll save a bunch of time on your tests. you can choose to re-code the video if it works
[23:27:50 CEST] <alexpigment> also what klaxa said ;)
[23:28:00 CEST] <klaxa> that will encode 60 seconds of video
[23:28:14 CEST] <klaxa> i would recommend to test encodes before wasting hours of time
[23:29:06 CEST] <klaxa> and yeah if your device can play the codec use -c copy to copy audio and video
[23:34:08 CEST] <bray90820> Yeah that didn't work at all https://pastebin.com/raw/GweKLz62
[23:36:02 CEST] <alexpigment> oh so these are dvdsubs?
[23:37:18 CEST] <bray90820> alexpigment: I think so but not 100% sure while they did come from a dvd I am strongly thinking the DVD might be a bootleg
[23:37:30 CEST] <alexpigment> do they look really bad and blocky in the MKV?
[23:37:35 CEST] <alexpigment> if so, they're DVD subs
[23:37:48 CEST] <alexpigment> anyway, i have no idea how to maintain those in an MP4
[23:38:42 CEST] <bray90820> They don't seem to be dvdsubs
[23:38:43 CEST] <bray90820> https://www.dropbox.com/s/hcvm19xnin9dsnw/Screen%20Shot%202017-04-06%20at%204.38.08%20PM.png?dl=0
[23:38:47 CEST] <bray90820> But what the hell do I know
[23:39:22 CEST] <alexpigment> yeah, that looks like DVD subs to me. plus it says so in your pastebin :)
[23:40:23 CEST] <bray90820> Aaahhh well nevermind then haha
[23:40:29 CEST] <bray90820> So anyone else know how to copy dvdsubs
[23:42:16 CEST] <alexpigment> another quick question - are you wanting to have them embedded, or do you want them always shown (i.e. burned into the video stream)?
[23:43:00 CEST] <bray90820> Embedded would be preferred
[23:43:29 CEST] <alexpigment> well, in case you give up on embedded, see this article: https://trac.ffmpeg.org/wiki/HowToBurnSubtitlesIntoVideo
[23:43:42 CEST] <alexpigment> the first example under "Picture-based subtitles
[23:44:29 CEST] <bray90820> I'm gonna still try fot embedded subtitles but thanks
[23:59:33 CEST] <alexpigment> bray90820: http://ffmpeg.gusari.org/viewtopic.php?f=11&t=2951
[23:59:41 CEST] <alexpigment> that pretty much confirms my suspicions
[23:59:52 CEST] <alexpigment> you can't convert one image-based sub into another with ffmpeg
[00:00:00 CEST] --- Fri Apr  7 2017


More information about the Ffmpeg-devel-irc mailing list