[Ffmpeg-devel-irc] ffmpeg.log.20200105

Mon Jan 6 03:05:02 EET 2020

[00:47:50 CET] <BtbN> Can the loudnorm filter be used on a live stream, or is it useless without doing two passes?
[00:51:39 CET] <kepstin> it works ok on live stuff, but it's more likely to have to use the non-linear normalization mode.
[00:52:08 CET] <kepstin> er, i guess the linear mode only works in two-pass, so it has to use the non-linear mode
[00:53:28 CET] <BtbN> My intent is to apply it to a playlist of various unrelated videos, and prevent the usual "Holy shit why is this so loud suddenly" when it switches
[00:53:47 CET] <kepstin> yeah, it should work fine for that.
[00:58:32 CET] <noudle> im trying to get hwdec to work using mpv on a rpi3 and have compiled ffmpeg with '--enable-mmal --enable-omx --enable-omx-rpi' but mpv fails to use hwdec erroring out with the following,
[00:58:37 CET] <noudle> [ffmpeg/video] h264_mmal: MMAL error 2 on control port
[00:58:41 CET] <noudle> [ffmpeg/video] h264_mmal: Did not get output frame from MMAL.
[01:02:29 CET] <furq> noudle: error 2 is ENOSPC
[01:02:44 CET] <furq> maybe you're running out of gpu memory
[01:03:13 CET] <noudle> uhm
[01:03:23 CET] <noudle> sec let me try something
[01:03:52 CET] <furq> also make sure nothing else is using mmal
[01:04:24 CET] <noudle> furq: nothing else is running
[01:04:34 CET] <noudle> but it can be im using not the full gpu mem available
[01:04:40 CET] <noudle> need to test that sec
[01:13:21 CET] <noudle> furq: any idea how much mem should be needed?
[01:13:39 CET] <furq> no idea
[01:13:54 CET] <furq> test with 512 i guess
[01:14:31 CET] <noudle> 64 resulted in the same, testing 128 now
[01:15:08 CET] <noudle> oh wow 128 do work
[01:15:13 CET] <noudle> ty for the hint!
[02:55:57 CET] <Hello71> isn't 2 ENOENT
[02:59:18 CET] <furq> https://github.com/raspberrypi/firmware/blob/master/opt/vc/include/interface/mmal/mmal_types.h#L46-L66
[03:20:18 CET] <Hello71> right
[03:20:20 CET] <Hello71> but like
[03:21:22 CET] <nicolas17> ENOENT the errno is 2
[03:43:09 CET] <noudle> whatever increasing the gpu mem did the trick
[03:43:15 CET] <noudle> night
[04:48:38 CET] <Soni> can ffmpeg encode DSD yet
[04:51:11 CET] <nicolas17> what is DSD?
[04:52:12 CET] <Soni> effectively, a fancy name for delta-sigma modulation
[04:53:01 CET] <furq> an uncompressed audio format that there's no need to care about
[04:53:14 CET] <nicolas17> looks like it can only decode
[04:54:25 CET] <furq> dacs use delta-sigma internally so i think the point of it is to avoid the conversion from pcm
[04:54:32 CET] <furq> so it doesn't really make sense to have an encoder in ffmpeg
[04:55:04 CET] <furq> either way it doesn't sound any better unless you work for sony
[04:55:34 CET] <Soni> furq: all you care is about what you hear, but you forget that there are programmers here
[04:56:04 CET] <Soni> furq: ffmpeg is a highly modular synth and a DSD encoder would be a nice addition to its arse-anal
[04:56:26 CET] <nicolas17> make one :)
[04:57:02 CET] <furq> some dacs, rather
[04:58:12 CET] <Soni> furq: besides, I need something that's fast enough to use in real-time with a 155MHz 8-bit DAC
[04:59:14 CET] <Soni> (as it turns out, PCM + dithering is... not)
[08:31:11 CET] <MoziM> if a video with no audio is 1.9G then shouldn't all of its extracted keyframes be roughly equivalent to 1.9G as well?
[10:17:31 CET] <yang> Hello, what is the best way to capture Analog VCR video? I have a script that is capturing to .avi (windows format), but I would prefer to capture to mp4 instead - ffmpeg -f v4l2 -standard PAL -thread_queue_size 512 -i /dev/video0 -f alsa -thread_queue_size 512 -i hw:2,0 -vcodec libx264 -preset superfast -crf 25 -s 720x576 -r 25 -aspect 4:3 -acodec libmp3lame -b:a 128k -channels 2 -ar 48000 out.avi
[10:17:43 CET] <yang> Could I get an example for capturing .mp4 ?
[10:19:07 CET] <durandal_1707> .mp4 is awful for capturing
[10:19:26 CET] <yang> oh, which one is good and doesnt use much space ? .webm ?
[10:19:47 CET] <yang> .avi stream is about 200 MB / 10 min.
[10:20:34 CET] <yang> But since I'd prefer to play streams on Linux, .avi isn't optimal
[10:46:44 CET] <yang> Do you think .avi is ok to capture video ?
[11:39:11 CET] <MoziM> if a video is DTS_FGT what sort of audio codec should i remux it with to be as lossless as possible?
[15:03:11 CET] <utack> i did not quite manage to figure it out from the manual, but is there a way to use ffprobe to find the maximum number of consecutive bframes used in a stream?
[15:05:14 CET] <DHE> yes, but you have to actaully go and count them yourself. maybe some kind of awk script on the output with -show_frames
[15:07:10 CET] <utack> DHE i am trying to piece that together, how would one tell show_frames to only output pict_type?
[15:08:47 CET] <cehoyos> The option loglevel verbose is supposed to tell you the number of reference frames, it may help
[15:10:39 CET] <utack> i have added a grep now around it to find the number of bframes, not elegant but works :D
[16:37:23 CET] <yang> I am interested to record with Matroska open codec
[16:37:29 CET] <yang> The current ffmpeg command line for .avi is like this:
[16:37:31 CET] <yang> ffmpeg -f v4l2 -standard PAL -thread_queue_size 512 -i /dev/video0 -f alsa -thread_queue_size 512 -i hw:2,0 -vcodec libx264 -preset superfast -crf 25 -s 720x576 -r 25 -aspect 4:3 -acodec libmp3lame -b:a 128k -channels 2 -ar 48000 out.avi
[16:37:36 CET] <yang> How do I make the command optimal to be used with matroska ? Just replace out.avi with out.mkv ?
[16:38:10 CET] <yang> And secondary, is there a program/tool that reveals which format is the video file being recorded with ?
[16:38:25 CET] <cehoyos> What is "Matroska open codec"
[16:38:26 CET] <cehoyos> ?
[16:38:49 CET] <cehoyos> If you don't want to write an avi file but an mkv file, use out.mkv instead of out.mkv
[16:38:57 CET] <cehoyos> This will not change the "codec" but the container
[16:39:45 CET] <cehoyos> FFmpeg reveals which format (and codec) a multimedia file has, yes: Just use "ffmpeg -i input"
[16:39:56 CET] <yang> ok
[16:39:58 CET] <yang> thank you
[16:49:10 CET] <yang> How could I replace the command to record audio in an open format like FLAC ?
[16:49:23 CET] <yang> or OPUS
[16:51:34 CET] <cehoyos> Replace libmp3lame with the codec of your choice, note that mp3 and opus have a very different behaviour than flac and that opus may not play with older players.
[16:53:21 CET] <MoziM> i have a video file that's an mkv and contains 1 english audio stream, if i want to add a secondary audio stream (russian dub) do i have to create a new file or can a second audio stream simply be "inserted" into an existing an mkv
[16:53:27 CET] <MoziM> *existing mkv file
[16:53:30 CET] <MoziM> ?
[16:53:37 CET] <yang> Yes, can i do "-acodec flac" ?
[16:58:02 CET] <cehoyos> mkvmerge may be able to "insert" the audio track (I don't know), FFmpeg always creates a new file, you can choose if you copy the audio track (you should always do that if possible) or if you want to re-encode
[17:01:08 CET] <MoziM> cehoyos: mkvtoolnix includes mkvmerge correct?
[17:02:51 CET] <yang> cehoyos: is it necessary to use -b:a with flac?
[17:03:27 CET] <cehoyos> You cannot use b:a with flac but as said, you should only use flac if you have a very specific reason, use mp3 if you have no cluse.
[17:03:31 CET] <cehoyos> You cannot use b:a with flac but as said, you should only use flac if you have a very specific reason, use mp3 if you have no clue.
[17:05:16 CET] <yang> cehoyos: I'd like to have audio and video codecs in an open format
[17:05:34 CET] <cehoyos> Why?
[17:06:09 CET] <cehoyos> And why do you think that mp3 which has a very old, known specification, is not old?
[17:06:13 CET] <cehoyos> And why do you think that mp3 which has a very old, known specification, is not open?
[17:06:19 CET] <yang> >> "hat should work, yes, though I would strongly recommend not to use MP3 nowadays. There are tons of better audio codecs out there (better in the sense of same quality
[17:06:24 CET] <yang>               for much lower bit rate). Use Opus for state-of-the-art or AAC if the resulting file must be playable on hardware devices (which often don't support Opus)."
[17:12:04 CET] <cehoyos> Why do you ask if you know better? If your player supports opus, opus is the best codec. AAC is not more free than mp3 but has better quality. Best quality is not to re-encode.
[17:12:16 CET] <cehoyos> Why do you ask if you know better? If your player supports opus, opus is the best codec. AAC is not more free than mp3 but has better quality. Best quality is always not to re-encode.
[17:15:53 CET] <yang> does flac require re-encoding when capturing stream ?
[17:16:02 CET] <yang> and mp3 doesn't ?
[17:16:56 CET] <cehoyos> When you capture, it depends on what format your capturing driver provides, most likely uncompressed audio that you should compress.
[17:17:05 CET] <cehoyos> If your player supports opus, opus is the best choice.
[17:17:18 CET] <cehoyos> If your player does not support opus but aac, aac is the second best choice.
[17:17:25 CET] <cehoyos> If your player supports only mp3, use mp3
[17:18:11 CET] <cehoyos> It is likely that you will not be able to hear a difference between the three as long as the bitrate is high enough.
[17:18:23 CET] <cehoyos> 128k is not a low bitrate
[17:42:38 CET] <kepstin> 128 might be kinda low for mp3, depending on content. it'd suggest going a *bit* higher there.
[18:17:10 CET] <yang> does flac require 128k setting
[18:19:45 CET] <furq> you can't set the bitrate of flac
[18:19:53 CET] <yang> ok
[18:53:13 CET] <kilobyte_ch> I try to optimize an mp4 for streaming with ffmpeg. The mp4 has 6 audio streams, 1 video stream and 1 subtitle stream. I use the following parameters `-movflags faststart -c:v copy -c:a copy -c:s copy -c:d copy -c:t copy`. I get an optimized mp4 out of it with all streams but I don't see the subtitle text in vlc on the optimized mp4. The entry is there but there is no text. On the source file it is working.
[18:55:44 CET] <furq> is the source mp4
[18:58:47 CET] <kilobyte_ch> furq: yes, source is mp4
[19:00:09 CET] <kilobyte_ch> Not sure if it maybe messes the subtitles up with the faststart movflag
[19:03:28 CET] <kilobyte_ch> Ok, even with just `-c:v copy -c:a copy -c:s copy -c:d copy -c:t copy -map 0` the subtitles are broken afterwards.
[19:04:29 CET] <ritsuka> does the subtitles play in another player?
[19:05:27 CET] <kilobyte_ch> ritsuka: On the source the subtitles work in vlc and gnome video player, on the output file the subtitles don't work on both players
[19:08:07 CET] <pink_mist> I didn't know mp4 actually supported subtitle streams
[19:08:43 CET] <ritsuka> kilobyte_ch: what's the format of the subtitles stream?
[19:08:57 CET] <kilobyte_ch> ritsuka: vlc says tx3g
[19:10:47 CET] <ritsuka> I can confirm there is something wrong, just tried to remux a mp4
[19:11:01 CET] <ritsuka> in quicktime player and mpv the subtitles work
[19:11:06 CET] <ritsuka> but not in vlc
[19:12:19 CET] <ritsuka> oh nv, they actually work
[19:12:37 CET] <ritsuka> it's just vlc being stupid when you seek around in the file
[19:15:57 CET] <kilobyte_ch> hmm, in the gnome video player the subtitles are coming up for a fraction of a second (output file)
[20:02:01 CET] <pink_mist> wouldn't surprise me if that was a bug in the gnome video player
[20:12:58 CET] <q3cpma> Hello, does anyone know if a lossless DSD -> PCM conversion is possible? I'd like to use sox afterward to get it to CD "quality".
[20:14:46 CET] <durandal_1707> sox is dead
[20:22:41 CET] <q3cpma> Doesn't really matter. I finally settled for DSDIFF -> pcm_s32le -> pcm_s16le
[21:03:35 CET] <void09> trying to join 5 pieces of a video and at the point where they are joined, the playback stutters a little. any idea how to fix this ? I used  ffmpeg -f concat -safe 1 -i 1.txt -c copy output.mkv
[21:04:08 CET] <void09> I cut the pieces with ffmpeg also, and the video part is frame accurate
[21:07:46 CET] <autofsckk> Hello, im trying to put subtitles into a movie, i found this  ffmpeg -i video.avi -vf subtitles=subtitle.srt out.avi   but the output video has a very bad quality, how can i keep the same size/quality of the output video? or at least a decent one?
[21:11:17 CET] <pink_mist> keeping it the same is impossible since you're outputting to avi and thus need to burn in the subs by reencoding the video, so you'll never get the same quality unless you want to have absolutely huge files
[21:11:49 CET] <autofsckk> pink_mist: i see, but is there a way to make it better?
[21:12:20 CET] <pink_mist> I mean, if you don't need avi, you could just remux it to mkv instead, then the subs don't need to be burned into the video frames
[21:12:39 CET] <pink_mist> but if you need avi, you'll need to figure out some codec and codec options that are acceptable to you
[21:13:06 CET] <autofsckk> pink_mist: i would like to put the movie in a usb so i can see it from a tv
[21:13:20 CET] <pink_mist> that doesn't really answer anything
[21:13:26 CET] <pink_mist> some tvs support mkv with subs
[21:13:29 CET] <pink_mist> others don't
[21:15:10 CET] <pink_mist> other tvs support putting an avi file next to a srt file, and it will do the right thing
[21:15:13 CET] <pink_mist> others don't
[21:16:14 CET] <autofsckk> this one does not support mkv files, im not sure if it reads srt files, but i think it doesnt
[21:16:56 CET] <pink_mist> then we're back to: 21:12 <pink_mist> but if you need avi, you'll need to figure out some codec and codec options that are acceptable to you
[21:17:32 CET] <pink_mist> it's a matter of trial and error, though I'm sure there are blog posts out there that can help you figure such things out faster
[00:00:02 CET] --- Mon Jan  6 2020