[Ffmpeg-devel-irc] ffmpeg.log.20180104

Fri Jan 5 03:05:02 EET 2018

[00:03:23 CET] <kingsley> durandal_1707: I use vlc, melt and ffplay.
[00:31:46 CET] <kingsley> furq: FYI: I reported the bug to kdenlive at
[00:31:54 CET] <kingsley> https://bugs.kde.org/show_bug.cgi?id=388503
[00:32:01 CET] <kingsley> and updated debian's bug report at
[00:32:25 CET] <kingsley> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=884268
[00:33:20 CET] <kingsley> Changing line 6 of my local copy of /usr/share/kdenlive/export/profiles.xml seems to have worked around the bug.
[00:33:25 CET] <kingsley> Thanks again.
[08:20:29 CET] <fletchii> Hi, I'm having a bit of an issue doing hardware acceleration with NVIDIA cards - actually, the full transcoding process is working fine, but it's all starting to fall apart when using the scale_npp filter and when trying to add an overlay to the video, the latter being a topic for another day. When I try to scale the video using scale_npp my output file just becomes green with a really low file size and therefore bitrate.
[08:20:49 CET] <fletchii> I've followed everything on the HWAccelIntro page, here's the command and output: https://pastebin.com/TFHgADEN
[08:24:28 CET] <fletchii> Scaling using NVDEC seems to work just fine, that said I'm planning to transcode the same file into multiple variants in one go. So scaling with the decoder isn't really an option, unfortunately.
[12:53:46 CET] <broman> how to stream a pipe? I always get an error "illegal dc vlc" or "mcbpc damaged at xx xx"
[13:03:24 CET] <DHE> you might need to specify information about the container/codec then. that means the stream is corrupted
[13:03:31 CET] <DHE> unless, of course, the stream actually is corrupted
[13:21:18 CET] <iAmSlow> hi
[13:21:29 CET] <iAmSlow> https://trac.ffmpeg.org/wiki/StreamingGuide
[13:21:38 CET] <iAmSlow> am reading this and have some confusion
[13:21:57 CET] <iAmSlow> i want to stream from my linux pc to rpi
[13:22:18 CET] <iAmSlow> the point that moste confuses me  is
[13:22:19 CET] <iAmSlow> ffmpeg -f dshow  -framerate 20 -i video=screen-capture-recorder -vf scale=1280:720 -vcodec libx264 -pix_fmt yuv420p -tune zerolatency -preset ultrafast -f mpegts udp://236.0.0.1:2000
[13:23:14 CET] <iAmSlow> udp://236.0.0.1:2000  ? shoudent i just specify runing poort on streaming mashine and just do 192.168.1.5:port  and connect to stream?
[13:24:57 CET] <iAmSlow> and whenwer i use -dshow i get Unknown input format: 'dshow'
[13:29:39 CET] <iAmSlow> how do i stram p2p
[16:45:48 CET] <pos> so, i've got these rtsp sources which, when using ffmpeg with -vcodec copy, end up with funky files. let's say i've captured 15 seconds of video, i get a video file sized about 5MB but all players think it's 50 minutes long
[16:45:53 CET] <pos> <pos> mplayer/mpv complain about invalid timestamps, vlc crashes
[16:48:36 CET] Action: kepstin hopes you opened a ticket against vlc about that, players crashing on certain input files is sometimes a security problem
[16:52:55 CET] <pos> oh, vlc 2.x really doesn't play well with stuff made with lavf57
[16:53:27 CET] <pos> in a lot of cases. vlc 3rc appears to have a fix
[17:10:08 CET] <AlexMG> i've been using ffmpeg to test streams for errors using `-v error` switch, then i was looking deeper into it and from what i have read it seems to ffmpeg tries to encode the stream to null and find errors allthough i could not find any concreate proof on that. I have couple cmd examples here: https://pastebin.com/GvfBa2v7
[17:11:53 CET] <AlexMG> if i am looking to find errors only while demuxing/decoding the stream without re-encoding will the first example on https://pastebin.com/GvfBa2v7 suffice? if not what is then the best approach to detect stream errors only while decoding?
[17:22:01 CET] <c_14> yeah
[17:22:12 CET] <c_14> ffmpeg only detects fatal errors though
[17:22:19 CET] <c_14> not sure what your use case is
[17:22:34 CET] <c_14> there is -xerror to make it stricter in some cases
[18:33:09 CET] <esc-reality> FFMPEG doesn't correctly follow the ID3v2.4 specs (http://id3.org/id3v2.4.0-frames). The following is stated there; All text information frames supports multiple strings, stored as a null separated list, where null is reperesented by the termination code for the charater encoding.
[18:33:34 CET] <esc-reality> i.e. ffprobe only shows the first value in the tag
[18:37:06 CET] <esc-reality> Look at the differences between mid3v2 (https://pastebin.com/Q6QEXVsL) and ffprobe (https://pastebin.com/E2027vMF) on the TPE1 and TCON fields.
[18:42:43 CET] <esc-reality> Is there something I'm missing, or should I write a bug report?
[19:17:10 CET] <hexhaxtron> I got 5 m4v videos. How can I join them together without encoding?
[19:23:50 CET] <EvalCopy> Alright, so (I hope i'm in the right channel,) when converting from libx264(high) at 15150 kbps to zmbv pal8 I try to set the average bitrate significantly lower due to the decreased resolution, but no matter what i set with bv or min/maxrate it starts churning out things x4 as large.
[19:26:39 CET] Last message repeated 1 time(s).
[19:26:39 CET] <EvalCopy> whoops
[19:28:06 CET] <kepstin> EvalCopy: zmbv doesn't support targetting a bitrate, it just compresses as much as it can and no more.
[19:28:37 CET] <EvalCopy> ohhh.
[19:28:45 CET] <kepstin> as is typical for lossless codecs
[19:28:54 CET] <EvalCopy> I wasn't aware it was lossless. my bad, guys.
[19:29:01 CET] <EvalCopy> that explains a lot
[19:29:41 CET] <kepstin> I mean, it only supports pal8, so converting from something else to pal8 will incur some loss. but the codec itself is lossless :)
[19:29:52 CET] <EvalCopy> I appreciate it!
[19:30:31 CET] <hexhaxtron> kepstin: can you tell me?
[19:43:02 CET] <AlexMG> i have ts file and streams, and i am looking for a way to detect if the stream or the ts file has any issues during decoding like missing frames, corrupt frame, etc... for which i am trying to figure out if `  ffmpeg -v error -i file.ts -f null -` is good enough or i should try to use `ffmpeg -v error -c:v copy -i file.ts -f null -` ?
[19:43:09 CET] <furq> hexhaxtron: https://trac.ffmpeg.org/wiki/Concatenate#demuxer
[19:43:16 CET] <AlexMG> https://pastebin.com/GvfBa2v7
[19:43:32 CET] <furq> AlexMG: copying is actually worse, it won't decode the stream
[19:43:47 CET] <AlexMG> ah
[19:44:06 CET] <AlexMG> so this is the best: `ffmpeg -v error -i file.ts -f null -`
[19:44:12 CET] <furq> there's also -xerror which will bail out with an exit status on error
[19:44:20 CET] <furq> but sadly not on warnings
[19:44:49 CET] <AlexMG> i am trying to capture all issues that occur, to improve the stream
[19:44:57 CET] <furq> yeah just stick with what you had then
[19:45:51 CET] <AlexMG> than you [furq], one more question, the cli, does it demux/decode only or, does it try to encode too?
[19:46:41 CET] <furq> that command line just decodes
[19:48:01 CET] <AlexMG> ah perfect, i wanted to be sure what it does, as i have read from one of the forums that command encodes to null also, which is something i did not want to do
[19:50:17 CET] <ascheel> can someone tell me why this is only grabbing 4 images from a directory totally over 19,000?  https://gist.github.com/ascheel/90cdf0871fad66fe5fc282132eb65f7a
[19:50:38 CET] <furq> ascheel: are all the images sequentially numbered
[19:50:39 CET] <ascheel> Joining 19,000+ images into a video.
[19:50:47 CET] <furq> if they're not then you'll need to use a glob
[19:50:48 CET] <ascheel> furq: They are.
[19:51:31 CET] <ascheel> furq: https://gist.github.com/ascheel/f415247622f88f82ec30b4388599a440
[19:51:44 CET] <ascheel> That's just the first 10, but it does show the sequential order
[19:51:56 CET] <ascheel> mediainfo shows only 4 frames.
[19:52:13 CET] <ascheel> There are no errors, it just doesn't grab the full gamut of images
[19:52:46 CET] <ChocolateArmpits> ascheel, have you tried a more recent version of ffmpeg ?
[19:52:59 CET] <furq> yeah that looks as though it should work
[19:53:01 CET] <ascheel> ChocolateArmpits: This was just pulled from nux-dextop's repo.
[19:53:08 CET] <furq> 2.6.8 is pretty old
[19:53:23 CET] <ascheel> I'll compile from source.  CentOS doesn't like to have modern repos.
[19:53:45 CET] <furq> but yeah you could try -pattern_type glob -i G00*.JPG
[19:53:59 CET] <ascheel> I'll try that, first.  Thank you.
[19:54:01 CET] <furq> oh hang on a minute
[19:54:04 CET] <furq> -rw-r--r--. 1 arts arts 20998 Jan  4 06:51 G0010088.JPG
[19:54:05 CET] <furq> -rw-r--r--. 1 arts arts 24301 Jan 4 06:51 G0020089.JPG
[19:54:08 CET] <furq> G001, G002
[19:54:35 CET] <ascheel> Huh....  good catch.  WTF happened there?
[19:54:41 CET] <furq> something fun no doubt
[19:54:45 CET] <furq> glob should still work with that anyway
[19:55:33 CET] <ascheel> yep.  Globbing is OK with it.  Need to investigate that.
[19:56:39 CET] <ascheel> Going to trash those first x images, anyway.
[19:56:42 CET] <ascheel> Much appreciated, folks.
[20:33:33 CET] <TheWild> hello
[20:39:56 CET] <TheWild> soooo... I wanted to ask something but it pointed out to be me not reading manual with understanding.
[20:49:42 CET] <ayum> there is a filter has a flags option, and I want to specify multiple flags for this option. but how to do? "flags=draw_bbox,enable_sdl" seems wrong
[20:52:38 CET] <DHE> what filter is this? I don't see anything like that in the docs
[20:52:55 CET] <ayum> I am writing a filter
[20:53:09 CET] <ayum> can I show you some code?
[20:55:55 CET] <ayum> https://paste.ofcode.org/sJWngH8etMyLkxv72xR33M
[20:56:55 CET] <ayum> it's my AVOption, this filter named "compare", but I want to specify some flags when I need debug the filter.
[20:57:16 CET] <ayum> like: "compare=flags=draw_bbox:sync_frames=64"
[20:58:09 CET] <ayum> but is it possible to specify multiple flags? like this one: "comapre=flags=draw_bbox,open_sdl:sync_frames=64".
[20:59:48 CET] <ayum> I think I found how to do it. just add "+" sign.
[21:00:40 CET] <ayum> "comapre=flags=+draw_bbox+open_sdl:sync_frames=64". if you need specify multiple flags, just need add "+" sign
[21:05:50 CET] <kepstin> yeah, or a - sign to remove a flag. I think it uses the same parser as other "flag" type options in ffmpeg, e.g. on codecs and formats.
[21:07:19 CET] <ayum> yes, I didn't found any information how to set multiple flags in avfilter options, but I remember some codec settings supports "-option +flags1,+flags2".
[00:00:00 CET] --- Fri Jan  5 2018