[Ffmpeg-devel-irc] ffmpeg.log.20190116

burek burek021 at gmail.com
Thu Jan 17 03:05:02 EET 2019


[02:17:14 CET] <KombuchaKip> Is there a way to use the ffmpeg API to detect if an audio file is corrupted? I know there is no way to do this if the container doesn't contain a checksum, but for when there is I am assuming there is a way to do this?
[02:24:54 CET] <DHE> I think you just have to decode the audio (assuming it's compressed) and look for errors in the decompressor
[03:15:46 CET] <Hello71> if you want to try harder you can calculate the SNR
[03:16:09 CET] <Hello71> I would assume there there is something for this somewhere in ffmpeg
[04:55:00 CET] <inflex> Hello all.   Is there a way to analyse a H264 (or other) file and make a log of activity per unit of time (ie, per second)?  I'm after the opposite of what most DVR/security people want,  I'm actually looking for points where there's no/low activity over a period of a few seconds or more
[05:08:39 CET] <friendofafriend> inflex: When you say "a lot of activity per unit of time", what activity do you mean?
[05:15:19 CET] <inflex> The difference between frames, though I supposed in terms of visuals,  motion / change of scene
[05:16:45 CET] <inflex> What happens is that I have a recording of my work, and there are times within a given job where I get up and go do other things, and the scene remains static,  I'd like to find those low-activity periods so I can cut them out in the NLVE.
[05:26:58 CET] <furq> if you just want to cut them out then look at the mpdecimate filter
[05:27:09 CET] <furq> i don't know of any filter that exports that information though
[05:39:25 CET] <inflex> np, appreciate the pointer, thanks
[06:14:31 CET] <KombuchaKip> DHE & Hello71: Thank you.
[12:24:26 CET] <zerodefect> Firstly, I'm using C-API interface to FFmpeg. Now Ubuntu 18.04 has FFmpeg 3.4.4 installed by default.  Since it's an older version, I went away and pulled down and then built v4.1 of FFmpeg.  Unfortunately, when I rewind back to 3.4.4, I cannot now find the H264 encoder when calling 'avcodec_find_encode(AV_CODEC_ID_H264)'. So I feel like I've butchered something :S.
[12:25:06 CET] <zerodefect> If I run `ffmpeg -encoders | grep x264`, I get:
[12:25:26 CET] <zerodefect>  V..... libx264              libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (codec h264)
[12:26:10 CET] <JEEB> I generally recommend requesting encoders by their name. also I hope you have re-compiled and linked your thing against the older ubuntu version + that the ubuntu version installed is one with libx264 enabled.
[12:26:32 CET] <JEEB> because requesting an encoder with the codec_id can give you literally any encoder doing that stuff :)
[12:26:42 CET] <zerodefect> Tried to retrieve by name too but to no avail.
[12:26:52 CET] <zerodefect> "libx264" ?
[12:26:53 CET] <JEEB> (and verify that the ffmpeg.c you're poking is built against the same FFmpeg that you're linking against
[12:27:01 CET] <JEEB> yes, libx264 or the rgb variant depending on things
[12:28:03 CET] <zerodefect> So I've actually gone a step further than what you suggest and completely deleted the 4.1 source and builds so that I don't trip over myself.
[12:29:32 CET] <JEEB> the source shouldn't matter as long as you no longer have the prefix around :)
[12:30:00 CET] <JEEB> (always set --prefix when building and then use PKG_CONFIG_PATH or PKG_CONFIG_LIBDIR to guide pkg-config to a specific thing)
[12:53:30 CET] <zerodefect> I wonder if during my install of 4.1, I corrupted something. I don't think I did install with `sudo`
[15:56:54 CET] <zerodefect> @JEEB, I got it to find encoder by name "libx264", but it won't find the encoder by id.  I find that quite interesting and unexpected.
[16:43:04 CET] <meiamsome> Hello, is there a limit to the amount of HTTP connections that can be made out of an FFMpeg process during its lifetime?
[16:45:57 CET] <pzy> I'm gonna go with... 65536
[16:46:32 CET] <kepstin> like, there's nothing hardcoded in ffmpeg itself
[16:52:48 CET] <meiamsome> We have a server nginx rtmp in docker that uses exec to run ffmpeg that takes the RTMP stream in and produces a HLS output to a URL of a Node server. It works but when we load test it (with 10 simultaneously) we discovered a pattern that the streams stop working after 6 minutes. It also seems to scale inversely with the number of streams, so maybe it's a limit on connections or something?
[17:06:43 CET] <pzy> you should be able to see that in your tests, right?
[19:51:28 CET] <tolszak> Hi, is it possible to decode single jpeg image using ffmpeg? E.g. to YUV? I google but I don't see any example
[19:53:05 CET] <tolszak> Ok seems I found something: https://shrex999.wordpress.com/2013/08/01/ffmpeg-yuv-to-jpeg-and-jpeg-to-yuv/
[19:53:43 CET] <JEEB> the usual examples under doc/examples should also be around :P
[19:53:59 CET] <JEEB> decoding JPEG is not any different from any other format in FFmpeg APIs
[19:54:29 CET] <DHE> or do you mean the CLI? that's what the example contains
[19:59:18 CET] <tolszak> JEEB: Yeah cli, I would like to test Intel HW decoder, and that's seems like a simplest solution
[20:50:58 CET] <nadermx> Hi, I'm having a issue trying to figure out how to send a cookie with ffmpeg input, I posted the question on superuser, https://superuser.com/questions/1394742/send-cookie-with-ffmpeg-input
[20:54:06 CET] <c_14> I believe you take the last 2 fields and separate them with an =
[20:57:25 CET] <nadermx> I'm not understanding, so you are saying I would do something along the lines of  "-cookies .youtube.com    TRUE    /   FALSE   1547252593  GPS 1=.youtube.com    TRUE    /   FALSE   1552434792  PREF    f1=50000000&hl=en" but only the last 2 lines?
[20:57:41 CET] <c_14> no
[20:57:47 CET] <c_14> you can throw away everything except for the last 2 fields
[20:58:23 CET] <c_14> so it'd be GPS=1; PREF=f1=50000000&hl=en etc
[21:00:47 CET] <nadermx> Ok, even when the there isn't a consitancy and the last two lines have a 0? ".youtube.com    TRUE    /   FALSE   0   YSC 9U9ILYfJDyA"  There I would still only do last 2 fields?
[21:01:05 CET] <nadermx> ie, YSC 9U9ILYfJDyA
[21:01:41 CET] <c_14> yeah
[21:02:26 CET] <c_14> first field is just the domain, you don't need that for ffmpeg (since it just hits that on everything), 2nd, 3rd and 4th I don't know, the 3rd is just a timestamp don't need that
[21:02:32 CET] <c_14> and the rest is the interesting bits
[21:03:10 CET] <nadermx> Thank you c_14, I will try this
[22:04:43 CET] <velix> I'm trying this: "-f segment -segment_time 7200 -start_number 1" "blah %03d.mp3", but %03d still starts at 0000
[22:05:05 CET] <velix> Can I also reply the last seconds after segmentising?
[22:05:37 CET] <c_14> start_number only works on input afaik
[22:06:17 CET] <velix> ppph segment_start_number !
[22:06:21 CET] <velix> oooh*
[22:07:00 CET] <velix> c_14: Thanks, works!
[22:08:23 CET] <velix> Perhaps segment_time_delta does rewinding?
[22:08:28 CET] <velix> let's try
[22:09:01 CET] <velix> ffmpeg is sooo grown up.
[22:10:13 CET] <velix> nope, that was wrong
[22:10:33 CET] <c_14> do you want an overlap between segments?
[22:10:52 CET] <velix> c_14: yeah
[22:11:06 CET] <c_14> don't think that's supported. not without really complex filtergraphs anyway
[22:11:40 CET] <velix> c_14: Actually, I could do it manually then.
[22:11:46 CET] <velix> segment_filename,segment_start_time,segment_end_time
[22:13:49 CET] <velix> c_14: maybe this one? https://stackoverflow.com/a/41812275
[22:15:00 CET] <velix> Man... Google search for "overlap" returns "overlay" all the time.
[22:16:33 CET] <c_14> that could work if you change the naming for the 2 instances to test_%d-1 and test_%d-2, that way it should sort correctly I think. You'd have to merge the m3u8s manually though
[22:16:47 CET] <velix> nah... then I'll go for the csv.
[22:16:52 CET] <velix> small bash script and done.
[22:17:00 CET] <velix> or even Excel :D
[22:22:19 CET] <velix> ah damn, it GENERATES a list.
[22:24:58 CET] <velix> seems like I need to do it oldschool:  -ss 00:03:00 -t 00:03:10
[22:26:28 CET] <velix> but sure, this doesn't seem to accept pure seconds :D
[22:27:07 CET] <c_14> -ss ? it should accept seconds just fine
[22:27:11 CET] <c_14> I do that all the time
[22:27:49 CET] <velix> c_14: oh, didn't find it. thanks. let me try
[22:28:07 CET] <Mortir> Hi
[22:29:06 CET] <Mortir> Anybody knows about a detailed tutorial on how to embed subtitles to a webm video?
[22:29:33 CET] <c_14> embed how, just convert to vtt?
[22:31:05 CET] <Mortir> Permanently sticking them.
[22:31:16 CET] <velix> c_14: It works. And that way, I also can do parallel processing in a queue.
[22:31:39 CET] <c_14> Mortir: https://trac.ffmpeg.org/wiki/HowToBurnSubtitlesIntoVideo ?
[22:35:11 CET] <poutine> Mortir, do you mean embed them in the video stream, like in the x264 SEI NALU as embedded 608/708, or do you mean burn in?
[22:35:47 CET] <poutine> I guess you did say webm, so you probably mean the latter
[22:46:49 CET] <velix> c_14: Yeah, good old "xargs" and -ss / -tt is way better than -segmentize.
[22:46:56 CET] <velix> c_14: I'm runing 8 parts in parallel
[22:47:12 CET] <Mortir> In such a way that conversion to other formats keeps them intact.
[22:47:13 CET] <velix> oldschool is always better
[22:47:16 CET] <velix> OIAB
[22:48:16 CET] <Mortir> c_14: I had found that link before but didn't try 'cause the example was with an avi. Thanks
[23:03:59 CET] <velix> Wow. 16 Threads also works on my i7 :D
[23:04:05 CET] <velix> More... Mooore. Moooore!
[23:10:21 CET] <fling> ffplay -pixel_format h264 -vcodec h264 -video_size 1920x1080 -i /dev/video3
[23:10:22 CET] <fling> [swscaler @ 0x7fc43cc1b820] deprecated pixel format used, make sure you did set range correctly
[23:10:32 CET] <fling> Which pixel format to use? ^
[23:12:09 CET] <velix> Can ffmpeg handle this crazy new 4K format from Sony Alpha cameras?
[23:38:39 CET] <fling> velix: how many cores?
[23:39:17 CET] <velix> fling: 16 (it's a virtual machine on a blade)
[23:39:29 CET] <velix> It's equal to i7 cores
[23:39:35 CET] <velix> about the same performance
[23:53:05 CET] <zerodefect> For members of structs in the C-API like AVCodecContext::codec_id, must they too be set with av_opt_set_xxx too?
[00:00:00 CET] --- Thu Jan 17 2019


More information about the Ffmpeg-devel-irc mailing list