[Ffmpeg-devel-irc] ffmpeg.log.20190521
burek
burek021 at gmail.com
Wed May 22 03:05:02 EEST 2019
[03:52:33 CEST] <snap1> ffmpeg devs are not nice: they refuse to fix: https://trac.ffmpeg.org/ticket/7824
[04:05:53 CEST] <DHE> you failed to submit a ticket properly. full output of ffprobe, for example, is expected. especially the version information shown, etc
[04:07:13 CEST] <snap1> no i did it fine
[04:08:03 CEST] <snap1> for version, it does not let me choose 4.1.3
[04:24:59 CEST] <Lynne> snap1: its not a bug, its a feature
[04:25:10 CEST] <Lynne> that's what the decoder will output
[04:25:42 CEST] <Lynne> it'll convert dst to pcm and that's the samplerate you get for the pcm signal
[04:26:11 CEST] <snap1> that's not correct samplerate
[04:26:41 CEST] <snap1> Channel(s) : 2 channels
[04:26:41 CEST] <snap1> Channel layout : L R
[04:26:41 CEST] <snap1> Sampling rate : 2 822 kHz
[04:26:55 CEST] <snap1> mediainfo reports correctly
[04:30:28 CEST] <snap1> lynne nothing to say now huh
[04:35:00 CEST] <snap1> i just exposed lynne
[04:58:59 CEST] <steve___> can someone please kick this asshat?
[05:02:31 CEST] <snap1> steve___ we don't need to kick lynne
[08:27:33 CEST] <grosso> hi
[08:30:59 CEST] <grosso> somebody can answer a question about rtmp streaming using libav* libraries?
[08:32:50 CEST] <snap1> this is #ffmpeg
[08:32:59 CEST] <snap1> try #libav
[08:33:09 CEST] <pink_mist> snap1: what the fuck
[08:33:26 CEST] <grosso> I mean ffmpeg libraries
[08:33:28 CEST] <pink_mist> snap1: libav* is provided by ffmpeg you idiot
[08:33:35 CEST] <grosso> not the command line tool
[08:33:50 CEST] <snap1> pink_mist libav is separate project
[08:34:00 CEST] <snap1> #libav
[08:34:18 CEST] <pink_mist> ffmpeg provides libav* libraries. libav provides libav* libraries.
[08:34:37 CEST] <pink_mist> he's in #ffmpeg because he's asking about the ffmpeg provided libav* libraries
[08:34:51 CEST] <grosso> ffmpeg libraries are called libav. Libav project is other thing
[08:34:56 CEST] <snap1> why is ffmpeg using libav libraries
[08:35:00 CEST] <snap1> why not use ffmpeg libraries
[08:36:46 CEST] <grosso> anyway, I mean to ask something about develop with the ffmpeg libraries... as in ffmpeg-devel topic said "Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg"
[08:37:13 CEST] <pink_mist> grosso: yeah, just ignore snap1 .. he doesn't have any clue what he's talking about
[08:37:42 CEST] <grosso> ok, thank you, so I proceed to ask:
[08:37:46 CEST] <snap1> pink_mist are you familiar with libav vs ffmpeg conflict
[08:38:57 CEST] <snap1> pink_mist think of ffmpeg as Coca-cola and libav as Pepsi
[08:39:08 CEST] <pink_mist> snap1: please stop talking to me.
[08:39:10 CEST] <pink_mist> I don't care.
[08:39:29 CEST] <snap1> you should care
[08:39:33 CEST] <snap1> because you are confused
[08:39:35 CEST] <grosso> I'm streaming from a device to a media server, and then to another device.. in both ends I'm using ffmpeg libs
[08:40:57 CEST] <grosso> I'm sending video h.264 and audio AAC, through a rtmp (flc) protocol... so far, everything works fine
[08:42:39 CEST] <grosso> now, I want to send occassionally, i mean not continuous but rather at some points, some textual data... just a few bytes of data, for signaling something
[08:43:36 CEST] <grosso> I really don't know how to achieve that, but I've tryied some ways...
[08:44:09 CEST] <grosso> for example, I tryed to create a 3rd. stream with subtitles
[08:46:11 CEST] <grosso> then I face some problems: 1) there is no documentation on how to encode subtitles; 2) I found some non-official doc, then try some av_encode_subtitles2() function, without luck...
[08:47:17 CEST] <enelar> Hi there guys. Im trying to add reach realtime data to video stream. Right now after processing the stream with python, im append new data to sub.vtt file. Im trying to produce HLS stream with that realtime-generated subtittles
[08:47:26 CEST] <snap1> grosso encode subtitles? you mean you want computer to automatically translate audio into text?
[08:47:28 CEST] <enelar> ffmpeg -y -i rtmp://relay/live/test -f webvtt -i /sub.vtt -start_number 0 -hls_time 10 -hls_list_size 0 -f hls /out/index.m3u8
[08:48:10 CEST] <enelar> it produces result stream with video audio and subtitless. but the subtitles are empty
[08:48:22 CEST] <grosso> but, appart from that, there is another issue, worst than that: since I don't want to send it continously, then, when calling av_write_interleave_frame() i get an error that too many time difference between first and last element in the buffer
[08:50:17 CEST] <grosso> so, I conclude subtitles are not suitable for that "timed metadata" that I want... so there is more like sending raw rtmp packets what I need
[08:51:08 CEST] <grosso> but I can't figure out how to do it from "high" level av_format stuff...
[08:52:48 CEST] <enelar> grosso, missed the question, seems im facing similar issue
[08:54:15 CEST] <grosso> enelar... well, in my case I work with the libraries, I know nothing about command line tool
[08:56:13 CEST] <enelar> grosso, gocha my bad
[08:56:50 CEST] <grosso> snap1: encode subtiltes I mean provide some source text, and then the library put it some way on the stream and I can recover it at the other end... that way you can send id3 tags or some other timed metadata
[08:57:08 CEST] <snap1> webvtt
[08:57:20 CEST] <snap1> you can use that
[08:58:23 CEST] <grosso> how to use it?
[08:58:56 CEST] <snap1> webvtt is subtitle format best for your needs
[08:59:27 CEST] <grosso> ok, so I need to know how to implement that in my C code
[09:06:32 CEST] <grosso> well, ultimately, I opted for sending an av_packet with av_write_frame(), with stream_index=2 while there is not really any stream with that index in the av_format_context... surprisingly, nothing breaks, and the packet reaches the other end...
[09:09:45 CEST] <grosso> I think is maybe an unintended behavior... anyway, I try to attach any data to that packet, but the data is always lost... received packet have always the same contents of 213 bytes, with always the same values, no matter what I put on the sending end
[09:12:00 CEST] <grosso> I will research about webvtt, maybe I found some documentation
[09:12:08 CEST] <grosso> thank you
[09:19:35 CEST] <grosso> rtmp has something named "on
[09:19:52 CEST] <grosso> "onCuePoint" and "onCaptionInfo"
[09:21:11 CEST] <grosso> it would be nice to have a way to send it with ffmpeg libraries
[09:32:05 CEST] <furq> grosso: rtmp doesn't support subtitle streams
[09:35:43 CEST] <furq> it looks like the flv demuxer is at least aware of captions but i don't see anything in the muxer
[09:36:15 CEST] <furq> https://trac.ffmpeg.org/ticket/2933
[09:55:09 CEST] <enelar> grosso, yep, rtmp (flash video steam) uses separate link to fetch subtittles and as3 to display it over the video
[09:55:27 CEST] <enelar> guys, how to make ffmpeg reload subtittle file each second?
[09:55:46 CEST] <enelar> hm, ill try pipe file hope it helps
[10:01:52 CEST] <Gigabitten> I don't suppose there's a way to access this IRC's historical messsages?
[10:02:01 CEST] <Gigabitten> From like, a few days ago
[10:02:17 CEST] <Gigabitten> somebody answered a question I had and I lost the exact solution and never bothered to figure out how it worked
[10:02:21 CEST] <sgnls> ...not unless somebody has an idle backlog for you :(
[10:02:25 CEST] <Gigabitten> agh
[10:02:29 CEST] <pink_mist> Gigabitten: this channel is publically logged
[10:02:31 CEST] <pink_mist> like the topic says
[10:02:46 CEST] <Gigabitten> oh yeah right at the end of the... hmm
[10:02:48 CEST] <pink_mist> check on the bugreports page I think
[10:02:49 CEST] <TheAMM> If you remember specific words I can scroll up a bit
[10:03:23 CEST] <pink_mist> oh wait, here: https://lists.ffmpeg.org/pipermail/ffmpeg-devel-irc/
[10:03:57 CEST] <Gigabitten> yup found it myself! thanks for the help tho
[10:04:01 CEST] <Gigabitten> glad that exists
[10:08:14 CEST] <Gigabitten> Cool, I got my solution back! And this time I'm on Linux where I have .bash_history
[10:08:21 CEST] <Gigabitten> Windows plz
[10:08:38 CEST] <pink_mist> thanks to cygwin I have .bash_history on windows too :P
[10:09:02 CEST] <pink_mist> (and a much better shell than CMD.EXE)
[10:09:47 CEST] <snap1> powershell is more powerful than cmd and bash
[10:09:56 CEST] <Gigabitten> I haven't messed with cygwin, I have an SSD so generally I'll just bop into Linux if I want to do something shell-y and mount some drives
[10:10:25 CEST] <Gigabitten> and nah, I've found that for my random shenanigans bash is better than powershell simply because it's on linux and linux more often has the tools I want
[10:10:39 CEST] <Gigabitten> I was just dumb in this instance
[10:14:06 CEST] <pink_mist> https://github.com/Microsoft/Terminal <-- honestly looking forward to this being available
[10:15:38 CEST] <Gigabitten> I've heard about that. Hopefully it turns out well.
[10:26:03 CEST] <enelar> snap1 please compare powershell to zsh, not the deprecated tools
[10:26:32 CEST] <snap1> enelar fair enough can you run zsh on windows
[10:26:46 CEST] <enelar> snap1 you can bash on windows?
[10:26:52 CEST] <snap1> yes
[10:26:59 CEST] <enelar> so what's the problem then?
[10:27:13 CEST] <snap1> i never said i had a problem
[10:27:50 CEST] <enelar> didn't touched windows for a while. i would compile zsh.exe and expect it run
[10:28:12 CEST] <pink_mist> zsh is availabe under cygwin just as well as bash is
[10:28:50 CEST] <enelar> cool. i just dont like when you compare most powerfull from A to one of the weakest of B
[10:28:54 CEST] <enelar> makes wrong impression
[10:29:10 CEST] <snap1> enelar i see, okay, i won't do it anymore
[10:29:26 CEST] <pink_mist> well, me ang Gigabitten were specifically talking about bash though, so I think it was fine for him to comment about it
[10:29:29 CEST] <pink_mist> *me and
[10:29:48 CEST] <enelar> though bash as an language, not a terminal
[10:29:50 CEST] <enelar> sorry
[10:31:25 CEST] <enelar> that vtt issue make me nervous
[10:31:41 CEST] <enelar> have spent whole night already
[10:54:54 CEST] <Gigabitten> okay so my fundamental issue is that I want to pad a bunch of images of various dimensions to make them all the same dimensions and then put them all in a slideshow. I received some help a while ago but I realized I'd prefer the images to be padded out to squares. I'm really not totally sure how the command works but I took my best crack at modifying it to suit my purposes and it... did not work. https://pastebin.com/eua2ExtJ
[10:56:04 CEST] <Gigabitten> Oh, and the old approach also isn't working on linux?
[10:56:25 CEST] <Gigabitten> Maybe I should just hop on over back to Windows to fix this technical problem (never thought I'd say that)
[10:57:55 CEST] <pink_mist> \ works differently under windows CMD.EXE and bash
[10:58:13 CEST] <pink_mist> you may need to change the "" surrounding your text into '' instead
[10:58:25 CEST] <pink_mist> on linux
[10:58:57 CEST] <pink_mist> I don't know whether that is the issue, but it's what I'd start with trying
[10:59:11 CEST] <Gigabitten> no dice
[10:59:58 CEST] <Gigabitten> and what do you mean \ works differently? not gonna pretend to know what it does in the first place in this context
[11:00:52 CEST] <pink_mist> well, on windows CMD.EXE, \ should just be a normal character
[11:01:04 CEST] <pink_mist> but under bash, in ""-quoted strings, it's not
[11:01:25 CEST] <Gigabitten> oh yeah
[11:01:44 CEST] <Gigabitten> so under windows, in "" strings it's not trying to escape anything, but in bash it is?
[11:01:49 CEST] <pink_mist> yeah
[11:02:29 CEST] <Gigabitten> okay, well if you don't have any other immediate ideas I'm going to just reboot real quick but I'll be here on my phone
[11:04:00 CEST] <Gigabitten> alrighty then
[11:40:19 CEST] <Gigabitten> Okay, I got one to mostly work and now it literally won't work anymore given the same input. What the hell is wrong with this program?
[11:43:16 CEST] <Mysterytrain> if you're trying to reshape images, wouldn't imageMagik be better than ffmpeg
[11:43:47 CEST] <Gigabitten> oh, I think I've heard about imageeMagik in passing
[11:43:52 CEST] <Gigabitten> I should have thought about it
[11:43:57 CEST] <Gigabitten> lemme check that out
[11:45:14 CEST] <Mysterytrain> that's all I can say. being blind, I have little use for images lol
[13:06:38 CEST] <snap1> EVERYBODY, i am boycotting MPV due to the fact that people in #mpv are not very friendly
[13:07:30 CEST] <pomaranc> now I want to join them, thanks
[13:07:59 CEST] <durandal_1707> lol
[13:08:01 CEST] <snap1> pomaranc please do and you will see what i am talking about
[13:09:33 CEST] <pomaranc> I hope so
[13:27:29 CEST] <CounterPillow> snap1 is VERY smelly do not engage with him or you will also become STINKY
[13:37:51 CEST] <snap1> CounterPillow by that comment, i can tell you are a #mpv person
[13:40:23 CEST] <Mysterytrain> I'm a winamp person!
[14:12:59 CEST] <cehoyos> snap1: Then I fear you are wrong here;-))
[14:13:15 CEST] <snap1> cehoyos are you carl?
[14:13:39 CEST] <cehoyos> My first name is Carl Eugen if that was the question
[14:14:18 CEST] <snap1> cehoyos i am very confused with who are the ffmpeg devs in this channel
[14:14:32 CEST] <cehoyos> Me too but why is this relevant?
[14:14:43 CEST] <snap1> it's relevant to me
[14:16:37 CEST] <snap1> cehoyos who else is well known ffmpeg dev here?
[14:18:30 CEST] <DHE> looking over the recent chat history hasn't been very good for snap1
[14:18:46 CEST] <snap1> dhe what do you mean?
[14:20:42 CEST] <DHE> you've confused ffmpeg with libav* and really drilled on it. there's the whole ticket thing..
[14:21:06 CEST] <snap1> DHE yes that's correct, i know that there was whole ffmpeg vs libav drama
[14:21:33 CEST] <snap1> i was trying to explain that to pink_mist
[14:21:50 CEST] <DHE> not relevant. user was asking about using libavformat, libavcodec, libavutil, etc which is what the * meant. these are subdirectories under the main ffmpeg source
[14:21:53 CEST] <snap1> and i think ticket was just misunderstanding
[14:22:58 CEST] <snap1> why would libav ( who is rival or ffmpeg) be in ffmpeg
[14:23:00 CEST] <snap1> that makes no sense
[14:23:44 CEST] <snap1> PEPSI would not be advertising coca-cola products
[14:24:28 CEST] <DHE> other way around. libav named themselves after the family of directories/subprojects under the ffmpeg directory/project
[14:26:31 CEST] <snap1> and what did ffmpeg do about it, after?
[14:27:01 CEST] <DHE> objection, relevance
[14:32:16 CEST] <snap1> DHE cehoyos did not remove my last ticket, so i think previous ticket was just big misunderstanding
[14:32:43 CEST] <cehoyos> I haven't "removed" a ticket for several years...
[14:33:32 CEST] <snap1> cehoyos this is valid bug: https://trac.ffmpeg.org/ticket/7824
[14:34:27 CEST] <cehoyos> Yes, feel free to reopen if a file decodes incorrectly for you
[14:34:45 CEST] <snap1> ffprobe's job is not decode
[14:34:53 CEST] <BtbN> It very often is
[14:35:12 CEST] <DHE> it does do it as a means of getting information from a stream, or if the user asks for it (eg: show_frames)
[14:35:14 CEST] <snap1> ffprobe's job is show correct information on media file
[14:35:25 CEST] <BtbN> And to do that, it very often has to decode it
[14:36:05 CEST] <snap1> cehoyos fact is mediainfo shows correct info and ffprobe does not
[14:37:15 CEST] <cehoyos> There is no "correct" info: There is a whitepage and there a technical conditions that apply if you decode an audio file.
[14:37:31 CEST] <snap1> yes there is
[14:37:51 CEST] <snap1> i expect 48khz mp3 file to show up as 48khz file
[14:38:59 CEST] <snap1> DHE what do you think about this?
[14:39:17 CEST] <cehoyos> But there are files that do (or would) not show as 48kHz, since mp3 stores coefficients, not audio data, the information has (only) a technical background, also note that some codecs always change the input frame rate and FFmpeg will always show the actual output frame rate, not the one of the past input file
[14:39:39 CEST] <DHE> I think I'm spending far too much time about this and I'm busy for the first workday of the week
[14:40:26 CEST] <snap1> cehoyos please upload a mp3 file that would show up 48khz on mediainfo and something else on ffprobe
[14:40:41 CEST] <cehoyos> I don't have one but sbr in mp3 was specified
[14:41:48 CEST] <cehoyos> But you can see the issue with aac
[14:42:07 CEST] <cehoyos> And a different issue with opsu
[14:42:08 CEST] <cehoyos> opus
[14:42:31 CEST] <snap1> opus is different though because opus only support certain khz
[14:42:43 CEST] <snap1> but please upload aac file then
[14:43:23 CEST] <cehoyos> But the opus tools will show you a different (in your words "correct") sample rate...
[14:45:16 CEST] <snap1> cehoyos yes we had a whold drama with opus and khz
[14:50:37 CEST] <CounterPillow> as people who don't understand multimedia do
[15:46:58 CEST] <thnee> I am using the following command: cat compressedVideo450-1547475088.0363379.mov| ffmpeg -f mov -v 9 -loglevel 99 -i pipe: -b:v 1M -g 60 -hls_time 2 -hls_list_size 0 -y /tmp/platinum/videos/8cd8ade5-26eb-40ee-8152-ba4fbd8e916d/hls/index.m3u8, but am getting "Could not find codec parameters for stream 0 (Video: h264, 1 reference frame (avc1 / 0x31637661), none(bt709, left), 540x960, 1/1200, 459 kb/s):
[15:47:00 CEST] <thnee> unspecified pixel format" and "Cannot determine format of input stream 0:0 after EOF". Not sure what I am doing wrong?
[15:47:37 CEST] <furq> why are you piping it into ffmpeg instead of just letting ffmpeg open it
[15:47:37 CEST] <thnee> The same ffmpeg command works if I use the file directly, instead of piping it to stdin
[15:47:58 CEST] <thnee> furq: Because I am reading it in chunks from cloud storage so I don't have to hold the entire thing in RAM
[15:48:07 CEST] <furq> that won't work well with mov/mp4 etc
[15:48:21 CEST] <furq> you need to be able to seek to find the moov header if it's not at the start of the file
[15:48:40 CEST] <furq> which iirc it never is with mov, but i've not touched mov in years
[15:50:28 CEST] <thnee> Hmm ok this is news to me, that's unfortunate.. I wish there was some workaround so I don't have to copy the whole thing to disk
[15:51:07 CEST] <thnee> But thanks for the reply!
[15:53:16 CEST] <CounterPillow> >I wish there was some workaround so I don't have to copy the whole thing to disk
[15:53:16 CEST] <CounterPillow> FUSE
[15:53:43 CEST] <thnee> I mean there is gcsfuse, but that seems like a huge can of worms
[15:54:13 CEST] <furq> you could just read it over http
[15:54:35 CEST] <furq> it should still work if the server supports range requests
[15:55:09 CEST] <CounterPillow> Yeah, that'd work too
[15:55:44 CEST] <thnee> I kinda see your point, but how would the seek call in ffmpeg be translated to the range header?
[15:56:07 CEST] <thnee> Are you saying ffmpeg can take an http url and do that automatically?
[15:58:32 CEST] <cehoyos> thnee: the workaround is to increase a buffer in the source code (IO_BUFFER_SIZE in libavformat/aviobuf.c)
[15:59:06 CEST] <thnee> cehoyos: Well yeah, but I don't know when the moov header will appear, so I would have to read the whole file into RAM, kinda defeating the point
[15:59:28 CEST] <cehoyos> It's not just the moov header
[19:16:32 CEST] <dlb76> dlb76 #<THIS_CHAN-NAME> Cannot change nickname while banned/quieted on channel
[19:16:32 CEST] <dlb76> pls dont do this
[19:16:32 CEST] <dlb76> its very not handy
[19:23:04 CEST] <DHE> well clearly it's not this channel doing it
[19:25:32 CEST] <dlb76> well #ffmpeg-devel f.e. dont do this
[20:59:27 CEST] <tomb^^> hi, is there anyway to create an empty audio track (anullsrc) as pcm_s16le? (anullsrc=channel_layout=mono:sample_rate=48000 creates it as u8 channel, need it s16)
[21:05:10 CEST] <furq> tomb^^: that should be s16 by default
[21:05:27 CEST] <furq> you can use -c:a pcm_s16le to force it
[21:05:39 CEST] <furq> but if it's being converted to u8 then i assume your output format is demanding that
[21:06:01 CEST] <cehoyos> tomb^^: ffmpeg -f s16le -i /dev/zero
[21:06:51 CEST] <furq> actually nvm it is u8 here
[21:06:57 CEST] <furq> for some reason the null muxer converts it to s16le
[21:07:20 CEST] <tomb^^> so -c:a pcm_s16le is the right way to go?
[21:07:23 CEST] <cehoyos> That's because the default audio encoder for the null muxer is pcm_s16le
[21:07:36 CEST] <furq> well that was simpler than i expected
[21:07:40 CEST] <furq> tomb^^: yeah
[21:07:48 CEST] <tomb^^> thanks!
[21:08:15 CEST] <furq> the /dev/zero thing will work as well and i assume it's faster but probably not noticeably so if you're doing anything else at the same time
[21:08:25 CEST] <furq> if you're encoding video then it'll make no difference
[21:09:17 CEST] <tomb^^> i'm trying to combine 4 opatom files (1 video, 3 audio) but I need the destination file to be 4 audio tracks, so I thought to use "anullsrc" to add a silence track, but ffmpeg fails because the rest are s16 and the anullsrc comes out as u8
[21:18:23 CEST] <nadermx> I'm trying to input a video and audio but trying to cut parts of the video. I read that if that's the case I shouldn't use -c copy? My current command is ffmpeg -ss 10 -i video -ss 10 -i audio -c:v copy -c:a aac -movflags frag_keyframe+empty_moov -f mp4 -
[21:35:34 CEST] <tomb^^> furq, seems like it works - thank you very much
[21:50:06 CEST] <cehoyos> nadermx: You cannot use pipe output for mp4
[21:50:33 CEST] <nadermx> you can with -frag_keyframe++empty_moov
[21:54:02 CEST] <nadermx> The command works when I do not use a -ss, so I think it has to do something with the -c:v copy?
[00:00:00 CEST] --- Wed May 22 2019
More information about the Ffmpeg-devel-irc
mailing list