[Ffmpeg-devel-irc] ffmpeg.log.20170815

Wed Aug 16 03:05:01 EEST 2017

[00:00:48 CEST] <SpLiC3> I should note when accessing the resulting remuxed .ts over http vlc handles it ok with a 4 sec or so delay, the first video frame shows then it buffers? or waits for x new packets?
[05:44:17 CEST] <willbarksdale> Hey folks, I am currently decoding an RTMP stream over a wifi network using FFMPEG, I would like to also send this FTMP stream to the facebook live API. Is it possible to somehow send the stream directly to fb live without decoding it? I am trying to avoid decoding and re-encoding at the same time.
[05:51:11 CEST] <willbarksdale> I am using FFMpeg on iOS
[06:17:49 CEST] <Cracki_> stream copy
[06:18:04 CEST] <Cracki_> meaning -c copy
[12:48:56 CEST] <Eldarko> Hello, guys! Is it possible to send segments with meta-data (such as duration/time) via http? Lets say, i want to send .ts files from my home to server, store em there, and then build a playlist on a server-side.
[12:49:20 CEST] <BtbN> HLS
[12:52:35 CEST] <Eldarko> HLS is for the playlist. Playlist requires at least a duration and filename. How do i send this data via http by ffmpeg with each segment?
[12:57:41 CEST] <BtbN> it writes that to the playlist automatically if you use the hls muxer
[12:57:48 CEST] <BtbN> just need to run a HTTP server to serve it
[13:00:35 CEST] <Eldarko> can it send me duration as http header, or something like that?
[13:00:52 CEST] <BtbN> The duration is in the playlist
[13:01:19 CEST] <BtbN> HTTP servers usually don't parse media file to extract metadata
[13:03:11 CEST] <Eldarko> Yeah, well, ill have to parse each ffmpeg playlist in order to get duration of the segment. I was hoping there is a way to get it w/o parsing the string
[13:05:56 CEST] <Eldarko> i dont really need this playlist, i only need a .ts file itself and its duration since its too expensive to extract it later from .ts file.
[13:09:43 CEST] <Eldarko> Thak you, Timo )
[14:41:32 CEST] <erika> hi! quick question for devs: is there any plan for basic support of dolby AC4?
[14:42:01 CEST] <erika> lmk if it's better suited for #ffmpeg-devel
[14:42:34 CEST] <BtbN> if you manage to find a spec that's not tied to some NDA or crazy license fees, and someone willig to implement it.
[14:43:17 CEST] <furq> does anything actually use ac4
[14:43:41 CEST] <furq> now, i mean
[14:43:44 CEST] <furq> i'm sure it's only a matter of time
[14:44:03 CEST] <dystopia_> is dolby atmos ac4?
[14:44:23 CEST] <furq> no
[14:44:29 CEST] <erika> hmmm OK, thanks - not in the near future then
[14:45:06 CEST] <furq> ATSC 3.0 supports Dolby AC-4 and MPEG-H 3D Audio.
[14:45:08 CEST] <furq> oh wonderful
[14:45:36 CEST] <dystopia_> i live in dvb land
[14:45:44 CEST] <furq> same
[14:45:48 CEST] <dystopia_> things here are moving to aac
[14:47:01 CEST] <furq> yeah dvb-t2 is exactly the same as dvb-t as far as audio goes
[14:50:57 CEST] <furq> Currently a Dolby Digital 5.1-channel soundtrack can be encoded with AC-3 using 384Kbps but with AC-4 that same track can be encoded at only 96Kbps with no perceivable loss in quality. In fact the compression is so good that a stereo track can be encoded at just 32Kbps,
[14:51:01 CEST] <furq> what
[14:51:17 CEST] <furq> "can be" is doing a lot of work in that sentence
[14:52:38 CEST] <cryptodechange> Is a network drive output for ffmpeg a bad idea?
[14:52:50 CEST] <cryptodechange> Could the end result be corrupted, local network
[14:52:56 CEST] <furq> do you plan on outputting at more than 1gbps
[14:53:13 CEST] <BtbN> if it's not seekable, you're in for half-broken or entirely broken stuff
[14:53:36 CEST] <cryptodechange> It's pretty much locally connected via. a switch, literally just compressing video sources
[14:53:44 CEST] <cryptodechange> but in the event of hanging, packet loss, etc
[14:53:51 CEST] <BtbN> Using NFS is perfectly fine
[14:53:56 CEST] <BtbN> if it's on a LAN
[14:53:58 CEST] <furq> right
[14:54:09 CEST] <BtbN> that's why TCP exists
[14:54:10 CEST] <furq> obviously don't do that shit to sshfs over wan, but cifs/nfs on a lan will be fine
[14:54:22 CEST] <furq> packet loss won't be an issue unless you pull a cable
[14:54:49 CEST] <cryptodechange> Using an SMB share which I mounted
[14:54:53 CEST] <furq> smb is cifs
[14:56:40 CEST] <cryptodechange> Ok sure, but is there a slight off chance of packet loss (faulty cable, congestion, over worked drives), will FFMPEG wait until data is written?
[14:57:02 CEST] <BtbN> ffmpeg has no idea and doesn't care about the network part of it
[14:57:09 CEST] <BtbN> it just sees a local file
[14:57:32 CEST] <furq> right
[14:57:33 CEST] <cryptodechange> Say I am scrubbing my ZFS drive, whilst xfering a lot of data, and operations are waiting/hanging on the data
[14:57:53 CEST] <cryptodechange> ok interesting
[14:58:14 CEST] <furq> this is why we have buffers
[14:58:16 CEST] <cryptodechange> Is there any way to check for corrupted/bad frames, instead of using my eyes? :P
[14:58:44 CEST] <BtbN> what kind of hellish network are you on that you expect it to mangle everything that goes through it?
[14:58:53 CEST] <furq> ffmpeg -v quiet -xerror -i foo.mp4 -f null - || echo oh no!!
[15:00:17 CEST] <cryptodechange> \o/
[15:01:23 CEST] <cryptodechange> This is my setup
[15:01:24 CEST] <cryptodechange> https://imgur.com/a/6Bjxh
[15:01:39 CEST] <cryptodechange> Behemoth is my 48 TB NAS and plex server
[15:01:52 CEST] <cryptodechange> Little box is my seedbox/encode destination
[15:02:02 CEST] <cryptodechange> So: big nas - encode > little nas - rsync > big nas
[15:02:16 CEST] <cryptodechange> My line of thinking was to avoid fragmentation over long encode times, when other stuff is being xfered
[15:02:25 CEST] <cryptodechange> As ZFS can't be defragged in a traditional sense
[15:07:38 CEST] <BtbN> you're worrying about all the wrong things...
[15:07:56 CEST] <BtbN> Networks don't shred data passing through it, and modern file-systems have no need to worry about fragmentation
[15:09:22 CEST] <furq> nah zfs fragmentation is a pain
[15:09:36 CEST] <viric> btrfs fragmentation too
[15:09:37 CEST] <furq> i doubt that this is going to help matters though
[15:10:27 CEST] <furq> also i hope you budgeted 20% of that 48TB for leaving empty forever or else the zpool performance goes to absolute fucking shit
[15:10:42 CEST] <viric> :)
[15:10:43 CEST] <furq> because you'll need to do that
[15:34:39 CEST] <cryptodechange> It's 12 * 6TB, split into 2 raidz-2 vdevs
[15:43:56 CEST] <furq> that seems like an excessively large raid6
[15:58:38 CEST] <SoreGums> how do i get these 2 commands into a single command? I know i need to use filter complex however all combinations I've tried don't work. What I want is to downmix the 6ch opus to aac and have both 6ch & 2ch in the mkv
[15:58:42 CEST] <SoreGums> ffmpeg -i chapters.video.audio-6ch-opus.subtitles.mkv -map_metadata -1 -map_chapters -1 -vn -sn -c:a aac -af "pan=stereo|FL < 1.0*FL + 0.707*FC + 0.707*BL|FR < 1.0*FR + 0.707*FC + 0.707*BR" audio-2ch-aac.m4a
[15:58:42 CEST] <SoreGums> ffmpeg -probesize 150M -analyzeduration 150M -i audio-2ch-aac.m4a -i chapters.video.audio-6ch-opus.subtitles.mkv -map 0 -map 1 -c copy chapters.video.audio-6ch-opus.audio-2ch-aac.subtitles.mkv
[16:01:46 CEST] <SoreGums> as can been seen above, i'm getting my downmixed 2ch aac, then copy it into the original with all the other streams - it works, just from reading, i should be able to 1shot this
[16:17:02 CEST] <SoreGums> the reason i need the 2ch audio is for some reason device is not mapping the channels properly during the downmix and i have no control over it
[16:40:34 CEST] <pgorley> hi, is it normal that after a call to swr_convert_frame, the out AVFrame's nb_samples is still 0?
[16:46:51 CEST] <durandal_1707> pgorley: yes
[16:47:07 CEST] <durandal_1707> read docs and examples
[16:50:05 CEST] <Eldarko> Hello again, guys! :) Could anyone tell me why cant i select a name for a file that is going to be pushed to server via http
[16:50:11 CEST] <Eldarko> ffmpeg -i " + url + " -c copy -f hls -hls_time 10 -hls_flags second_level_segment_duration -use_localtime 1 -use_localtime_mkdir 1 -nostdin -loglevel 16 -hls_segment_filename %Y%m%d/%H/%M/segment_%Y%m%d%H%M%S_%%010t.ts -method PUT http://127.0.0.1:9090/live/stream.m3u8
[17:10:18 CEST] <Eldarko_> So, could anyone help with the question above?
[17:17:54 CEST] <Eldarko_> https://pastebin.com/uqHkAmmy
[17:28:29 CEST] <pgorley> durandal_1707: thanks, that helped me find what was happening
[17:28:41 CEST] <pgorley> i was ignoring an input changed error
[17:28:53 CEST] <pgorley> and swr_convert was never being called
[17:42:45 CEST] <momomo> anyone knows if a universal lnb can recieve ku band frequencies? c-band?
[17:44:39 CEST] <Pandela> Morning
[17:45:15 CEST] <Pandela> I was curious, is there a way to tell FFmpeg to process an entire file, but skip parts (say the beginning) to apply an effect?
[17:50:38 CEST] <pgorley> i keep getting "Input channel count and layout are unset", which options do i need to set on swr_context to fix this?
[17:51:44 CEST] <atomnuker> pgorley: you need to set both the input and output frame's sample format, samplerate and channel layout
[17:52:18 CEST] <pgorley> atomnuker: on the AVFrame, using av_opt_set_int, or both?
[17:52:47 CEST] <atomnuker> on the avframe
[17:52:56 CEST] <atomnuker> the opt set stuff is for the swr context
[17:53:50 CEST] <pgorley> are those options not set when decoding?
[17:55:27 CEST] <atomnuker> on the input avframe they are
[17:55:37 CEST] <atomnuker> on the output they're not, since you create the output avframe
[17:55:55 CEST] <pgorley> seems the channel layout on my input frame is set to 0
[17:58:42 CEST] <durandal_1707> pgorley: set channel layout based on channel count
[17:59:22 CEST] <pgorley> AV_CH_LAYOUT_MONO for 1 and AV_CH_LAYOUT_STEREO for 2, etc, right
[17:59:50 CEST] <atomnuker> pgorley: no, its AV_CH_LAYOUT_MONO for mono and AV_CH_LAYOUT_STEREO for stereo
[18:01:59 CEST] <Eldarko> Is it possible to set a name pattern for a .ts file when using PUT?
[18:02:25 CEST] <Eldarko> or even just a name
[18:03:15 CEST] <SpeakerToMeat> And, I have to run, duh, see you all soon :D
[18:03:33 CEST] <Eldarko> bye
[18:03:53 CEST] <pgorley> durandal_1707, atomnuker: thanks, turns out i just forgot to set AVCodecContext->channel_layout before decoding
[18:04:49 CEST] <SpeakerToMeat> I'll leave the question here, I want to combine 6 mono wav files into a single 5.1 track, using this guide https://trac.ffmpeg.org/wiki/AudioChannelManipulation but I have two doubts. A) I need to change the bit depth, do I need to do this for all inputs separatedly or can I do this to the post complex filter map output somehow? and B) in cases where channels are labelable is there any way to have them
[18:04:51 CEST] <SpeakerToMeat> rightly tlabeled
[18:05:29 CEST] <orbisvicis> how can I determine the actual bitrate of a vbr mp3 file ?
[18:05:43 CEST] <orbisvicis> well, average bitrate, min/max bitrate
[18:05:56 CEST] <BtbN> "actual" bitrate?
[18:06:04 CEST] <BtbN> It's variable, there is no actual bitrate.
[18:06:18 CEST] <BtbN> If you want the average, just divide the filesize by the duration
[18:09:57 CEST] <orbisvicis> BtbN: how can I get the data filesize, excluding headers/frames ?
[18:10:00 CEST] <orbisvicis> and min/max
[18:10:27 CEST] <BtbN> an .mp3 file is rather raw, the size of the header barely matter
[18:16:14 CEST] <orbisvicis> hmm so duration, according to ffmpeg, is "estimated" from bitrate
[18:17:06 CEST] <orbisvicis> is there any tool to tell me the duration, cbr/vbr, average bitrate (cbr/vbr) and min/max bitrate (vbr) ?
[18:25:07 CEST] <BtbN> the duration for mp3 should be accurate
[18:29:43 CEST] <DHE> that may happen if the input were a pipe or none-file. ffmpeg should seek to the end to read the timestamp on the last sample, right?
[18:29:55 CEST] <DHE> otherwise yeah, mp3 duration may be wildly inaccurate if using bitrate alone.
[18:30:08 CEST] <Lirk> Hi
[18:30:17 CEST] <orbisvicis> no, I get this using `ffmpeg -i` on certain mp3 files
[18:30:38 CEST] <Lirk> is there way to get help about ffmpeg postprocessing from command line?
[18:31:20 CEST] <Lirk> from source postprocess.c /* -pp Command line Help */ const char pp_help[] =
[19:21:26 CEST] <SpeakerToMeat> I want to combine 6 mono wav files into a single 5.1 track, using this guide https://trac.ffmpeg.org/wiki/AudioChannelManipulation but I have two doubts. A) I need to change the bit depth, do I need to do this for all inputs separatedly or can I do this to the post complex filter map output somehow? and B) in cases where channels are labelable is there any way to have them rightly labeled?
[19:23:39 CEST] <SpeakerToMeat> I'll try to do a global -c:a see if it works
[19:36:26 CEST] <SpeakerToMeat> Btw anybody here well experienced with quicktime? what's the right expected PCM type (alongside prores in case it matters any), signed or unsigned? little or big endian?
[19:38:34 CEST] <JEEB> lessee the thing I have around
[19:39:00 CEST] <JEEB> pcm_s16be
[19:39:08 CEST] <JEEB> is what I have, but I have a feeling it's pretty free
[19:42:04 CEST] <SpeakerToMeat> JEEB: I'm checking wikipedia, it seems there's fourcc for both signed le and be, and identifiers for 24 be and 32 be, but these do not denote whether they're signed or unsigned
[19:43:17 CEST] <SpeakerToMeat> Wellthey don't specify two's complement like 16 bits le/be so I guess unsigned
[19:43:40 CEST] <JEEB> well I just checked what I had in a prores mov file and that was what I just noted :P
[19:43:58 CEST] <JEEB> I would guess audio is signed unless specifically noted otherwise
[19:44:26 CEST] <JEEB> also you could check what FFmpeg writes with which identifiers in MOV
[19:44:35 CEST] <JEEB> wikipedia is not exactly the best of references
[19:44:51 CEST] <SpeakerToMeat> Yeah but it has more information than the apple developers document...
[19:45:12 CEST] <SpeakerToMeat> Which doesn't even specify whether the in24 fourcc identifier is for BE or LE pcm...
[19:45:31 CEST] <JEEB> also there's QTFF dot pdf
[19:45:36 CEST] <JEEB> from apple
[19:45:52 CEST] <JEEB> (also they seem to have the file format spec as HTML now, too)
[19:45:53 CEST] <JEEB> https://developer.apple.com/library/content/documentation/QuickTime/QTFF/QTFFPreface/qtffPreface.html
[19:47:04 CEST] <SpeakerToMeat> Yeah I should check, I'm here right now https://developer.apple.com/library/content/documentation/QuickTime/QTFF/QTFFChap3/qtff3.html
[19:47:06 CEST] <SpeakerToMeat> thanks
[19:47:25 CEST] <JEEB> https://developer.apple.com/library/content/documentation/QuickTime/QTFF/QTFFChap3/qtff3.html
[19:47:40 CEST] <JEEB> this page contains the definition for the identifier I have in my mov file's audio track 'twos'
[19:47:43 CEST] <JEEB> for example
[19:48:27 CEST] <SpeakerToMeat> Yeah it says twos and raw come from version 0, versions 1 and 2 extend that
[19:50:01 CEST] <JEEB> sound sample description box seems to contain formatSpecificFlags which then has endianness etc
[19:50:29 CEST] <JEEB> so I would guess the newer four-letter identifiers don't base on that thing itself too much for the endianness
[19:53:22 CEST] <SpeakerToMeat> That would make sense, I'm checking the sound descriptors for versions 1 and 2
[20:01:20 CEST] <SpeakerToMeat> Ok for labeling and layout with amerge, is there any way to manually specify a layout for inputs which do not have it?
[20:01:26 CEST] <SpeakerToMeat> audio I mean
[21:10:14 CEST] <Justanick> Hi, what is the right was if I want to cross compile ffmpeg 3.3.3 but I m not able to execute the cross compiled files on the build host.
[21:10:14 CEST] <Justanick> At the moment it fails with zlib requested but not found.
[21:10:14 CEST] <Justanick> --target-exec= is not set
[21:11:07 CEST] <Justanick> Is ther something like, "use the host compiler for dependency checking"?
[21:16:29 CEST] <DHE> well you don't need the host version of zlib. that doesn't actually do you any good
[21:18:37 CEST] <Justanick> DHE: I know that I don't need the host version. The target version of the zlib is installed but the test fails with collect2: error: ld returned 1 exit status
[21:30:35 CEST] <JEEB> actual error is most likely in ffbuild/config.log
[21:32:33 CEST] <Justanick> JEEB: I see only a config.log. There is no folder ffbuild
[21:33:36 CEST] <JEEB> ok, then that release is from before ffbuild subdirectory was added
[21:33:46 CEST] <JEEB> current master has it under ffbuild/
[21:35:46 CEST] <Justanick> I'm at the moment on 3.3.3
[21:37:26 CEST] <Justanick> JEEB: The config.log is at https://gist.github.com/D-Albers/4d93a769078ddff44fa7373cc91d5b7c
[21:41:19 CEST] <Justanick> Any ideas how I can fix this?
[21:48:04 CEST] <niklob> hello, " -i video.mkv -i sub.srt -c:v copy -c:a copy -c:s dvbsub output.mkv" returns:  [matroska @ 000001cd047cc700] Application provided invalid, non monotonically increasing dts to muxer in stream 2: 229503 >= 27794 av_interleaved_write_frame(): Invalid argument
[21:49:08 CEST] <niklob> ffmpeg can't convert srt to dvb ?
[21:54:38 CEST] <kepstin> Justanick: do a search for "-lz" in that output to find the linker error. Looks like there's some abi incompatibilities between your libraries and the compiler you're using.
[22:00:33 CEST] <Justanick> kepstin: The eroor message is: failed to merge target specific data of file /tmp/ffconf.uwSNapW1.o
[22:00:56 CEST] <kepstin> and the line before that, which complains about some different nan modes or something.
[22:01:07 CEST] <kepstin> you probably need arch-specific compiler knowledge to fix this :/
[22:02:07 CEST] <Justanick> kepstin: linking -mnan=2008 module with previous -mnan=legacy modules
[22:03:34 CEST] <kepstin> Justanick: you'll want to bring that compiler command and linker output to someone who works on the lede/openwrt project toolchain to get help, I suspect it's a bit out of the domain of what most people here know.
[22:03:42 CEST] <Justanick> kepstin: Coukd this problem been be triggered through changing the compiler option? If yes, I may have an idea.
[22:09:59 CEST] <niklob> How to soft-embed dvb-subs from a .srt-file?
[22:18:52 CEST] <Justanick> kepstin: Thanks for the help so far. :)
[22:24:29 CEST] <niklob> How can I tell ffmpeg to ignore wrong time value in my srt file?
[22:24:42 CEST] <niklob> time values*
[22:27:59 CEST] <niklob> nvm
[22:28:04 CEST] <niklob> it's a bug in ffmpeg
[22:28:07 CEST] <niklob> :(
[00:00:00 CEST] --- Wed Aug 16 2017