[Ffmpeg-devel-irc] ffmpeg.log.20160503

burek burek021 at gmail.com
Wed May 4 02:05:01 CEST 2016


[00:16:01 CEST] <hyponic> c_14 still nothing. here is the files i am using: https://goo.gl/g1mJSc https://goo.gl/Gf0d4q
[00:21:00 CEST] <c_14> Eh
[00:21:06 CEST] <c_14> literally none of those subtitle events has a duration
[00:21:10 CEST] <c_14> They won't be displayed
[00:21:12 CEST] <c_14> ever
[00:28:40 CEST] <devster31> hi, where can I get a complete list of the libraries I need to compile ffmpeg?
[00:29:12 CEST] <hyponic> c_14 00:00:08,359 --> 00:00:08,359 isn't that a duration?
[00:29:27 CEST] <hyponic> those are subs extracted from the stream itself.
[00:29:40 CEST] <c_14> Last time I checked 8.359 - 8.359 is 0
[00:30:14 CEST] <hyponic> c_14 hmm... good valid point i didn't see that. but those are extracted from the mkv stream.
[00:30:31 CEST] <hyponic> c_14 can i give you a test stream on pm?
[00:30:40 CEST] <c_14> sure
[00:33:50 CEST] <c_14> devster31: afaik the only library ffmpeg needs is a valid libc
[00:34:38 CEST] <c_14> everything else is optional
[00:36:22 CEST] <devster31> yes, I meant all the optional ones
[00:36:31 CEST] <devster31> the list of supported codecs and whatnot
[00:37:26 CEST] <c_14> decoders for most everything is standard
[00:37:35 CEST] <c_14> there's a few encoders that you need if you want those
[00:37:41 CEST] <c_14> check the output of ./configure --help
[00:37:45 CEST] <c_14> That lists everything
[00:37:50 CEST] <devster31> alright, thanks
[00:38:13 CEST] <furq> devster31: https://ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Features
[00:38:36 CEST] <furq> anything labelled with an E requires an external library
[00:38:54 CEST] <furq> or most things, anyway, since that wasn't 100% accurate last time i checked
[00:39:14 CEST] <c_14> Oh, I didn't know that page even existed
[00:39:22 CEST] <devster31> woa that's a long list
[00:39:26 CEST] <furq> it's tucked away pretty well
[00:39:35 CEST] <furq> i think i found it by chance through google
[00:40:06 CEST] <furq> ./configure --help is the right way to go if you just want a big list of supported libs
[10:13:33 CEST] <anddam> hello, I'm trying to merge a .srt file with a MKV, I'm doing     ffmpeg -i $f.mkv -i $f.srt -map 0:0 -map 0:1 -map 1 -c copy ../$f.mkv
[10:14:17 CEST] <anddam> where f is set to the name of the files without the dot and extension, I'm explictly using 0:0 and 0:1 since the input 0 has multiple audio tracks and I only want the 0:1
[10:14:48 CEST] <anddam> I remember merging srt in MKV in past, this time I'm getting "Codec for stream 0 does not use global headers but container format requires global headers"
[10:15:02 CEST] <anddam> that I think is a warning since it's yellow in terminal
[10:15:09 CEST] <anddam> and then [matroska @ 0x7fd614015c00] Can't write packet with unknown timestamp
[10:15:11 CEST] <anddam> av_interleaved_write_frame(): Invalid argument
[10:16:55 CEST] <anddam> what is this lack of timestamp due to?
[10:31:42 CEST] <halcyforn> hello
[10:33:24 CEST] <halcyforn> i cant find this on net how to configure ffserwer and ffmpeg to work like readyshare on router. is this possible?
[11:48:47 CEST] <Carlrobertoh> I'm trying to do a screen casting with GDIGRAB. Could anybody tell me which timestamp shall i use?
[11:49:04 CEST] <Carlrobertoh> input packet timestamps?
[12:13:14 CEST] <Sartsj> does anyone here have any experiences packing a MPEGTS into smoothstreaming? I have a question open on superuser that I could use some help on: http://superuser.com/questions/1072458/ffmpeg-packaging-h264mpegts-into-smooth-streaming
[12:16:25 CEST] <Fyr> guys, is it possible to mux DTS-HD into mp4 container?
[12:16:45 CEST] <Fyr> pediwikia says that MP4 supports DTS-HD.
[12:17:48 CEST] <Fyr> but, nowhere can it be found.
[12:21:06 CEST] <c_14> I'm pretty sure mp4 doesn't support DTS
[12:22:24 CEST] <Fyr> c_14, https://en.wikipedia.org/wiki/Comparison_of_video_container_formats#Audio_formats_supported
[12:22:43 CEST] <JEEB> there's a mapping for DTS in ISOBMFF
[12:23:09 CEST] <JEEB> although I must say that wikipedia isn't the best source for this sort of stuff
[12:23:24 CEST] <JEEB> I think L-SMASH supports it, but not sure about libavformat
[12:25:25 CEST] <Paranoialmaniac> libavformat wrongly supports DTS in MP4 which uses mp4a fourcc + ObjectTypeIndication
[12:27:08 CEST] <Paranoialmaniac> MP4 does not support DTS audio since there is no spec of the encapsulation
[12:28:16 CEST] <Paranoialmaniac> registration of ObjectTypeIndication for a codec does not mean that codec can be encapsulated in MP4
[12:28:22 CEST] <Fyr> ok
[12:29:49 CEST] <Fyr> why do people keep using DTS or Dolby, when FLAC has already been implemented and it's free for use?
[12:30:45 CEST] <Fyr> FLAC compression, I guess, is almost maximal.
[12:31:45 CEST] <Paranoialmaniac> channel layout? i think FLAC does not support all supported channel layout of DTS audio and DolbyDogital
[12:32:57 CEST] <Paranoialmaniac> or embedded downmix
[12:34:25 CEST] <CruX|> Fyr: hmm why flac why not ape
[12:34:31 CEST] <CruX|> both are free
[12:34:42 CEST] <Fyr> ok
[12:34:50 CEST] <Fyr> let's ape everything.
[12:35:10 CEST] <Fyr> why are there so many standards?
[12:35:43 CEST] <Fyr> blu-ray uses DTS-HD and TrueHD.
[12:36:23 CEST] <c_14> And occasionally pcm
[12:36:27 CEST] <Fyr> +some additional.
[12:36:27 CEST] <Fyr> a player should support all of them instead of one.
[12:38:21 CEST] <ritsuka> actually there are some specs for DTS in ISOBMFF, but I guess they aren't free or public domain
[13:29:44 CEST] <Fyr> guys, is it possible to convert DTS-HD to ALAC, keeping all the channels?
[13:30:53 CEST] <Fyr> Apple users are so pitiable, I want my video to be playable on every platform.
[13:33:53 CEST] <sasos90> hey guys
[13:33:57 CEST] <Fyr> hey
[13:34:04 CEST] <sasos90> i am developing an android app. and i really want to use ffmpeg
[13:34:12 CEST] <sasos90> but i get some output stream file error
[13:34:17 CEST] <sasos90> does anyone knows about that?
[13:34:27 CEST] <sasos90> more accurate: Output #0, mp4, to 'i /storage/emulated/0/Download/test.mp4': Output file #0 does not contain any stream
[13:35:50 CEST] <sasos90> full output: http://pastebin.com/pBsqmRnG
[13:47:28 CEST] <sasos90> Fyr: do you know something about it perhaps?
[13:47:40 CEST] <Fyr> no
[13:47:53 CEST] <Fyr> I've never tried to develop for Android.
[13:50:41 CEST] <jkqxz> sasos90:  What was the full command line you used to produce that output?
[14:19:25 CEST] <RoyK> hi all. I have a USB webcam connected to a banana pi (raspberry pi clone thing), and attempting to read from it with -vcodec copy works, but gives me a bitrate at 184522.9kbits/s, which I somewhat lower than the theoretical limit of USB2, but still a wee bit high
[14:20:45 CEST] <RoyK> hm... I thought this camera did some encoding in the first place - doesn't look that way :(
[14:22:21 CEST] <RoyK> well, it should support h.264, according to the specs
[14:22:33 CEST] <jkqxz> Webcams still give very high bitrates when encoding to MJPEG.  They really only do just enough to fit the stream inside the available USB bandwidth.
[14:23:40 CEST] <RoyK> any idea how I can tell the camera to store in h.264? it's a logitech C920
[14:27:38 CEST] <jkqxz> You do need to explicitly enable H.264 encoding, because IIRC it's one of the ones which embeds the H.264 stream inside a comment field in JPEG images.  (Yay for "compatibility".)
[14:27:55 CEST] <jkqxz> I'm afraid I don't know how to do that with ffmpeg, though.
[14:28:19 CEST] <RoyK> http://paste.debian.net/679470/
[14:38:14 CEST] <jkqxz> RoyK:  "ffmpeg -f v4l2 -video_size 1280x720 -input_format h264 -i /dev/video1 -c:v copy out.mkv" works for me with a C920 (gives me a bit over 3Mbps).
[14:38:42 CEST] <RoyK> I only have /dev/video0
[14:39:11 CEST] <RoyK> oh
[14:39:17 CEST] <RoyK> --input_format
[14:39:18 CEST] <RoyK> sec
[14:43:59 CEST] <RoyK> hm
[14:44:02 CEST] <RoyK> jkqxz: ffmpeg -f v4l2 -video_size 1280x720 -i /dev/video0 -input_format h264 -c:v copy -c:v copy output$i.mkv
[14:44:19 CEST] <RoyK> still no h.264
[14:45:58 CEST] <relaxed> RoyK: -input_format h264 needs to go before the input
[14:46:07 CEST] <RoyK> oh :)
[14:47:23 CEST] <RoyK> that's more like it :D frame=  934 fps= 36 q=-1.0 size=   11525kB time=00:00:30.90 bitrate=3055.0kbits/s speed=1.18x
[14:47:44 CEST] <RoyK> relaxed: thanks :)
[14:56:20 CEST] <Tuman> is there anybody here who speaks italian?
[15:08:43 CEST] <Fyr> Borkr, are you borkra?
[15:09:28 CEST] <Borkr> Fyr: Nope, sorry :)
[16:29:22 CEST] <Zitter> hi, I have a .mp4 container (with audio & video inside) and another file with audio only. Now I want to add second_file.m4a to the first_one.mp4 without losing audio from first_one.mp4. Best goal is to set first video's audio to a lower volume than second one.
[16:30:25 CEST] <Fyr> Zitter, don't you want to add another audio channel?
[16:30:50 CEST] <Fyr> or you want to get rid of the first audio and use only the second m4a file?
[16:30:54 CEST] <Zitter> I've tried ffmpeg -i first_video.mp4 -i audio.m4a -map 0:v -map 0:a  -codec copy output.mp4 but it is not what I need
[16:31:01 CEST] <Zitter> Fyr, no.
[16:31:35 CEST] <Zitter> Fyr, I don't want add another audio channel. I want to listen *both*, but first audio with a lower volume
[16:31:44 CEST] <Zitter> as a audio background
[16:32:13 CEST] <Zitter> it is a meeting in italian, the second audio is english translation.
[16:36:45 CEST] <DHE> well, with that map command you're omitting the second file entirely
[16:37:07 CEST] <DHE> that can be done, you'll need to use -filter_complex and assemble a multi-stage pipeline
[16:42:51 CEST] <Zitter> DHE, yes. But how to take only audio from first container?
[16:43:25 CEST] <Fyr> Zitter, http://pastebin.com/hyHyFDQR
[16:43:55 CEST] <Fyr> you set the desired volume.
[16:43:55 CEST] <kepstin> Fyr: that seems kind of silly, you can do it all in a single command :)
[16:44:05 CEST] <Fyr> yes, I know.
[16:44:24 CEST] <Fyr> kepstin, but you didn't write a thing.
[16:44:34 CEST] <Fyr> I wrote the solution.
[16:44:47 CEST] <Fyr> plain and simple
[16:44:57 CEST] <Fyr> without finery. ;)
[16:46:16 CEST] <kepstin> I just got here. But sure, "ffmpeg -i first_video.mp4 -i audio.m4a -filter_complex '[0:a]volume=-3dB[base];[base][1:a]amerge[mixed]' -map '[mixed]' output_audio.m4a" should do it in 1 command.
[16:46:46 CEST] <kepstin> add aac encoding options and adjust volume to taste.
[16:46:50 CEST] <Fyr> yeah, it looks like.
[16:47:04 CEST] <Zitter> kepstin, I need both audio/video as a result
[16:47:31 CEST] <kepstin> Zitter: then add a '-map 0:v -c:v copy' before the output file (and rename it to something.mp4)
[16:48:27 CEST] <Zitter> ok... well, ffmpeg is really nice :)
[16:48:54 CEST] <Fyr> yeah, but its developers are not.
[16:49:08 CEST] <Fyr> they are bad people and they know it.
[16:57:04 CEST] <durandal_1707> Fyr: what you are talking about?
[16:57:12 CEST] <Fyr> ?
[16:57:47 CEST] <Fyr> the developers are not going to fix the delay bug I posted.
[16:58:21 CEST] <durandal_1707> whats bug number?
[16:58:43 CEST] <kepstin> that doesn't make them bad people.. just, busy? unconvinced? unaware that an issue exists?
[16:59:25 CEST] <durandal_1707> Perhaps bug is hard to fix?
[17:00:45 CEST] <Fyr> durandal_1707, https://trac.ffmpeg.org/ticket/5290
[17:01:15 CEST] <Fyr> also, I'm very sad about absence of multithreading when converting FLAC.
[17:01:15 CEST] <DHE> ffmpeg -i sourcevid.mp4 -i alternate-audio.m4a -filter_complex "[0:a]volume=-10[srcvol];[srcvol][1:a]amix[audioout]" -map [0:v] -map [audioout] -c:v copy output.mp4
[17:01:20 CEST] <kbarry> Typicallu I use RTMP in either a "Pull from me" or "Pushed to you"  scenario.
[17:01:36 CEST] <kbarry> I can do rtmp "Push", and find example all over.
[17:02:00 CEST] <kbarry> Is ffmpeg capable of RTMP (pul), where the signal waits for connections to pick up said video signal?
[17:02:02 CEST] <Fyr> 16-core Intel Xeon cluster convert FLAC as fast as my 4-cored phone does.
[17:02:09 CEST] <DHE> oh wait that's not how the volume works
[17:03:26 CEST] <flux> fyr, I'm not saying that isn't annoying, but I'm certain there are more valuable things on the table than multicore flac conversions :)
[17:03:48 CEST] <flux> besides I would really expect the xeon to trump the mobile core even single threaded, did you really benchmark that?
[17:04:09 CEST] <Fyr> flux, try to compress 1-hour 24-bit/192 kHz audio. =)
[17:04:25 CEST] <kepstin> flac is sufficiently fast that it's really not something anyone cares about... particularly if they're also compressing video
[17:04:50 CEST] <flux> well, for one, multithreading the AAC converter would probably be higher priority
[17:05:13 CEST] <kepstin> that said, I'm pretty sure individual packets in flac can be compressed in parallel without issue, so if someone feels like it would be useful, it would be great if they implement that :)
[17:05:33 CEST] <Fyr> ok, but the delay bug is related to AAC.
[17:05:37 CEST] <durandal_1707> Fyr: I have patch for mt flac encoding
[17:05:49 CEST] <Fyr> *muxing mp4
[17:06:01 CEST] <Fyr> durandal_1707, where have you been all this time?????????????
[17:06:28 CEST] <durandal_1707> Thing is it is missing some stuff
[17:06:50 CEST] <Fyr> I wasted a hundred of Vooodooo dolls onto FFMPEG developers.
[17:07:01 CEST] <Fyr> some of them must be dead already.
[17:07:55 CEST] <flux> you must be in the misguided understanding that FFmpeg developers owe you something.. :)
[17:08:03 CEST] <Fyr> they do
[17:08:10 CEST] <Fyr> I made a donation.
[17:08:16 CEST] <flux> ah, right, a donation
[17:08:23 CEST] <fling> How do I properly create a multiangle video from a bunch of files?
[17:08:27 CEST] <flux> donation with strings attached?
[17:08:32 CEST] <durandal_1707> Of few bucks
[17:08:38 CEST] <kepstin> a donation is just that, a donation... it doesn't mean they owe you anything...
[17:08:45 CEST] <Fyr> KKKKKKKK
[17:09:00 CEST] <flux> next time fyr should make a deal instead of donation :)
[17:09:01 CEST] <Fyr> I need more Voodoo dolls.
[17:09:09 CEST] <Fyr> flux, been there, tried that.
[17:09:10 CEST] <fling> 2. How do I properly convert mjpeg to h264 or even better thing without a noticeable quality loss?
[17:09:19 CEST] <Fyr> posted bug, reported onto #ffmpeg-devel.
[17:09:28 CEST] <Fyr> they requested donation.
[17:09:35 CEST] <Fyr> I made the donation.
[17:09:40 CEST] <Fyr> nothing happened. =)
[17:09:52 CEST] <DHE> fling: multi-angle as in a video file with lots of distinct video sterams in it?
[17:10:10 CEST] <fling> DHE: right.
[17:10:13 CEST] <Fyr> they talked like 15 minutes about the bug, then the conversation flawed away. =)
[17:10:22 CEST] <flux> fling, I don't think there's a "direct" way avoiding lossiness, just find a nice crf argument for x264 and off you go..
[17:10:32 CEST] <durandal_1707> Fyr : donate until bug is fixed
[17:10:32 CEST] <flux> fling, probably something between 10 and 20 :)
[17:10:46 CEST] <flux> double the donation each time. that'll get 'em!
[17:10:50 CEST] <DHE> fling: ffmpeg -i inputfile0 -i inputfile1 -i inputfile2 ... -map [0:v] -map [1:v] -map [2:v] ... -c:v XX output.mkv
[17:10:50 CEST] <fling> DHE: also the video freezes when I capture from multiple webcams at once -> https://bpaste.net/show/6dfe631deddc
[17:10:57 CEST] <Fyr> ='(
[17:11:07 CEST] <fling> DHE: but it works just fine when I capture with separate ffmpegs.
[17:11:22 CEST] <durandal_1707> Fyr: they need to do job for starving family
[17:11:36 CEST] <kepstin> fling: hmm. yeah, multiple live inputs with ffmpeg is a bit of a problem, since the whole input code is single-threaded :/
[17:12:02 CEST] <fling> flux: I don't know how to find the proper value for crf, my eyes are not so good for the video quality comparison :>
[17:12:27 CEST] <fling> kepstin: do I need to use multiple ffmpegs with pipes then?
[17:12:27 CEST] <durandal_1707> Fyr: also spamming devel irc channel now and then
[17:12:39 CEST] <kepstin> fling: there is no "proper" value, it's all up to "what quality loss vs filesize tradeoff can you tolerate"
[17:13:06 CEST] <Fyr> durandal_1707, if they were good developers, they would not have starving families.
[17:13:13 CEST] <fling> kepstin: ahh it also explains why things go wrong when alsa overflows the buffer.
[17:13:41 CEST] <flux> fling, I'll give you a value: 15. go ahead and use it ;)
[17:13:45 CEST] <kepstin> fling: multiple ffmpegs with pipes might actually work better, since the pipes act as a buffer so the different parts can run asynchronously (note that the default pipe buffer on linux is a bit small for uncompressed video tho)
[17:14:00 CEST] <durandal_1707> Fyr: yeah nobody claim open source give you free lunch
[17:14:22 CEST] <Fyr> durandal_1707, well, it's obviously not free.
[17:14:26 CEST] <fling> kepstin: I want the smallest possible file for the quality I need. But I can't properly compare the video quality with my own eyes&
[17:15:13 CEST] <fling> kepstin: what if I will use mbuffers and not uncompressed video? I'm capturing with -c:v copy
[17:15:19 CEST] <kepstin> fling: then borrow someone else's eyes, or pick a metric to analyze quality by.
[17:15:34 CEST] <fling> what is metric?
[17:15:54 CEST] <durandal_1707> Fyr: I Will try to finish that flac mt encoding patch
[17:15:54 CEST] <fling> I hoped there is something for seeing the difference between different quality level encoded videos somehow
[17:16:35 CEST] <kepstin> fling: there's metrics like 'psnr' or 'ssim' that attempt to quantify quality differences, but neither completely accurately models the human visual system
[17:18:04 CEST] <furq> they're also not very useful if you have psy on, which you should
[17:18:19 CEST] <fling> kepstin: also I'm capturing with -c:v copy because of the alsa xruns appearing when capturing -c:v libx264
[17:18:38 CEST] <fling> durandal_1707: what is mt?
[17:18:49 CEST] <furq> multithreaded
[17:18:49 CEST] <fling> DHE: thanks.
[17:19:11 CEST] <fling> flux: should be useful.
[17:19:39 CEST] <fling> Now I need to align the files somehow&
[17:21:48 CEST] Action: kepstin is kind of surprised that aac isn't trimming the codec delay off properly; that's required (and works) for lots of audio codecs like mp3, opus, etc.
[17:22:21 CEST] Action: fling is now surprised too
[17:23:02 CEST] <kepstin> fling: that's probably not related to your issue :)
[17:24:06 CEST] <fling> kepstin: I know, but I also remember the nasty mp3 delays.
[17:25:14 CEST] <kepstin> yeah, it's basically fixed now in mp3 - at least if you use lame - since it throws a 'xing' or 'info' tag at the start with the number of samples delay
[17:26:21 CEST] <kepstin> but with aac it's still a mess - some early encoders/decoders just assumed that there was a particular delay that all codecs used, iirc there's some itunes specific metadata in mp4 files, and then there's some newer way of specifying it with edit lists or something in the mp4 format
[17:26:43 CEST] <flux> edit lists are fine, no?
[17:26:45 CEST] <fling> Needs to be fixed!
[17:27:10 CEST] <flux> and it's the standard (edit lists that is).
[17:27:28 CEST] <kepstin> All I know is that I couldn't get anything to sample-accurate trim any of the aac files that I have tried making with a few different encoders and decoders :/
[17:28:05 CEST] <flux> if you set timescale to something nice, you can get microsecond trims, in theory ;)
[17:28:06 CEST] <kepstin> hmm. Not sure I have tried ffmpeg's built-in aac encoder tho; I should give that a whirl.
[17:29:20 CEST] <kepstin> my usecase is that I'm making a synchronized audio/video thing that runs in browsers, and for audio to loop correctly it has to have the delay properly trimmed.
[17:29:28 CEST] <kepstin> Browser support for various formats kind of sucks :)
[17:29:42 CEST] <flux> amazing if they don't support edit lists :(
[17:29:54 CEST] <kepstin> I found that vorbis works best - chrome has an off-by-1 trimming bug there, and firefox is sample-accurate
[17:30:14 CEST] <Fyr> kepstin, if it's somehow related to
[17:30:14 CEST] <Fyr> https://trac.ffmpeg.org/ticket/5290
[17:30:14 CEST] <Fyr> then kick the developers in the nuts.
[17:30:28 CEST] <kepstin> and firefox can't trim mp3 properly, but chrome does. I ended up compiling mpg123 with emscripten to get mp3s working in firefox
[17:41:40 CEST] <A124> mp3 is inferior and outdated.
[17:42:26 CEST] <kepstin> it's a legacy format for which a *lot* of pre-encoded files exist. Sometimes you just have to support it.
[17:48:43 CEST] <RoyK> hm.... relaxed helped me finding a way to capture h.264 directly from my cam, but when I try to switch between using 'motion' (the program) and going back to ffmpeg (via a switch connected to the bpi), ffmpeg usually dies with a lot of these errors. any idea how I can reset the cam somehow? I tried an USB reset, but it doesn't work. It helps to wait a while, but I don't know what's really causing this http://paste.debian.net/679604/
[17:48:48 CEST] <Zitter> any hint on why offset doesn't work here: ffmpeg -i first_audio.m4a -itsoffset 00:08:22 -u second_audio.m4a -strict -2 -filter_complex amerge -c:a aac -f mp4 out_audio.m4a ?
[17:49:17 CEST] <Zitter> btw, it is a -i and not -u on second output
[17:51:21 CEST] <Fyr> I would use -ss and/or -to -t
[17:51:33 CEST] <Fyr> and I'm sure it will work.
[17:53:36 CEST] <kbarry> A124 What do you like to use instread of mp3?
[17:59:07 CEST] <kepstin> I use opus when I can, it's a pretty nice codec (helps that they put the sample trimming right into the container spec, too)
[18:14:14 CEST] <A124> kbarry AAC.
[18:14:28 CEST] <A124> OPUS is better, of course, but the support is not yet everywhere.
[18:14:54 CEST] <A124> And at higher bitrates AAC is pretty close to OPUS.
[18:15:48 CEST] <A124> Though I contemplate, kbarry, that OPUS should be supported in browsers already, as youtube switched from AAC to OPUS month or so back.
[18:22:05 CEST] <furq> they didn't switch, they use both
[18:22:39 CEST] <furq> vp8/9 and opus aren't supported in safari and probably never will be because they are deeply in love with google
[18:22:48 CEST] <furq> s/they/apple/
[18:24:36 CEST] <kbarry> A124 fruq I am relatively new to the ffmpeg/audio/video world. All my work currently is in audio, and as such, its been a bit harder for me to get information that is audio-specific. Not everything that applies to video, it would seem, applies to audio ie, keyframes?)
[18:25:29 CEST] <A124> kbarry Not at all. But... if you use that audio in video, it is called fragments.
[18:26:03 CEST] <A124> So if you use audio in video, you ideally want audio and video synchronized. Aka fps / audio chunks aligned.
[18:26:10 CEST] <kbarry> For the foreseeable future, i will be working with audio-only. For thanks for the information.
[18:26:24 CEST] <furq> fwiw there's nothing wrong with mp3 if you're not bitrate constrained
[18:26:25 CEST] <A124> Also about audio AAC is the choice of modern players, bluray, etc.
[18:26:38 CEST] <kbarry> (about what the pieces are called).
[18:26:53 CEST] <furq> there's very little to choose between any of the popular audio codecs at rates above 128kbps
[18:26:57 CEST] <kbarry> A124:  I have been picking that up as I am getting more into it.
[18:27:14 CEST] <furq> that's why all the listening tests you'll see are at 96k, so that it's actually possible to find differences
[18:27:18 CEST] <A124> I have to disagree with furq, mp3 at 128 is inferior to OPUS at 128
[18:27:24 CEST] <furq> i said above 128
[18:27:34 CEST] <A124> Even above.
[18:27:46 CEST] <furq> mp3 v2 sounds about as good as anything else
[18:27:49 CEST] <furq> lame v2, that is
[18:27:58 CEST] <kbarry> I guess the better question is, would toy "average" listener be able to tell te difference :)
[18:28:08 CEST] <kbarry> Depends on the audience, the gear they have avaliable, etc.
[18:28:12 CEST] <A124> Everyone has different hearing, and I can hear difference up to 320 kbps.
[18:28:41 CEST] <A124> Better chinese phone with headphones.
[18:29:03 CEST] <kbarry> One thing I was trying to do earlier was create an RTMP stream that I could pull.
[18:29:22 CEST] <A124> I did this to compare the bitrates of OPUS, but it gave also info about the other format qualities obviously.
[18:29:26 CEST] <kbarry> I was really trying to pull in my webcam with ffmpeg, then view that stream via ffplay
[18:30:14 CEST] <furq> i should probably test with a standard test corpus but i've never been able to hear any difference between lame v2 and flac
[18:30:15 CEST] <A124> kbarry mp3 is still inferior even above 128kbps. The other formats are very close.
[18:31:01 CEST] <A124> furq I was able to hear difference of 256kpbs OPUS and FLAC, so anothr person also.
[18:31:19 CEST] <kbarry> A124 furq do either of yuou have any recommendations for audio/streaming newbs trying to improve my base-knowledge?
[18:31:27 CEST] <A124> But the difference was so miniscule, that for almost all audieces 256 OPUS (or say AAC) should be ok.
[18:31:39 CEST] <kbarry> ie, sites, blogs, conventions, user groups, videos, etc.
[18:31:45 CEST] <furq> as far as i can tell most people can't tell the difference between 128k opus and flac
[18:32:01 CEST] <A124> kbarry IDK, read about opus, aac, do not use mp3 for (insert word here) sake
[18:32:13 CEST] <furq> there's no reason to use mp3 these days anyway
[18:32:24 CEST] <furq> but don't feel like you're doing something awful if you have to
[18:32:29 CEST] <kbarry> I use it, because i don't know any better :)
[18:32:33 CEST] <A124> furq Often the recording itself is done in lower quality, so depends on the record itself.
[18:32:35 CEST] <kbarry> or didn't know any better
[18:32:56 CEST] <furq> i still have plenty of music in mp3
[18:33:07 CEST] <A124> But if you got synthesized (real, not fake), at high sample rate that is multiple to the one you downsample, there really is difference.
[18:33:21 CEST] <A124> Yeah, I do have a lot in mp3, but when I can I move to FLAC.
[18:33:32 CEST] <furq> sure
[18:33:36 CEST] <A124> Why? Lossless, and can make any lossy I want, ie OPUS for phone.
[18:33:41 CEST] <furq> i can't hear the difference but i've got plenty of space for lossless
[18:34:01 CEST] <A124> If I was a "regular" person I would want 128 to 256kbps opus or aac.
[18:34:34 CEST] <kbarry> In an attempt to become more familiar with ffmpeg et. al, I have been trying to use it for my listening (ffplay)
[18:34:47 CEST] <kbarry> And for recording , ie, music from a video on youtube, etc.)
[18:34:48 CEST] <furq> but yeah in my experience the big difference between mp3 and newer codecs is below 128k
[18:34:58 CEST] <furq> there is a huge difference between lame 64k and opus/aac at 64k
[18:35:06 CEST] <A124> kbarry if you are on nix, opus tools are part of most distros.
[18:35:07 CEST] <furq> or even vorbis
[18:35:16 CEST] <kbarry> I'm on Windows.
[18:35:17 CEST] <A124> kbarry If you worry about compatibility, use AAC.
[18:35:31 CEST] <kbarry> (my shop is 60/40 linux/windows
[18:35:36 CEST] <furq> the only thing to choose between them is compatibility really
[18:35:59 CEST] <A124> Above 192kbps and AAC or OPUS, most people (taking general mass, not audio folks) do not hear difference much.
[18:36:24 CEST] <A124> kbarry Btw, you mentioned streaming, any more details?
[18:36:39 CEST] <A124> https://upload.wikimedia.org/wikipedia/commons/8/8d/Opus_quality_comparison_colorblind_compatible.svg
[18:36:43 CEST] <A124> https://upload.wikimedia.org/wikipedia/commons/e/ec/Opus_bitrate%2Blatency_comparison.svg
[18:37:09 CEST] <A124> Unless you want lower bitrate or low latency, AAC is today perfectly good.
[18:37:24 CEST] <furq> kbarry: http://listening-test.coresv.net/results.htm
[18:37:39 CEST] <A124> With opus, music is usable from 48kbps, while at 64kbps it becomes pretty ok
[18:37:47 CEST] <furq> that's the best listening test i'm aware of
[18:38:34 CEST] <A124> Oh a person I know that works with video and audio for production told me...
[18:39:04 CEST] <A124> To use itunes library for aac. (Exctracting installer of itunes and using just the encoder from there) as it has best quality of all encoders.
[18:39:21 CEST] <furq> yeah it'd be nice to find a listening test that compares fdk to qaac
[18:39:26 CEST] <A124> Given that itunes does use AAC extensively for loooong years and they got ton of money it is believable.
[18:39:31 CEST] <kbarry> furq:  Wow, that is great.
[18:39:36 CEST] <furq> there's one from 2011 which compares fhg to qaac, but nobody really knows how similar fdk and fhg are
[18:39:41 CEST] <furq> fhg comes out quite well though
[18:39:47 CEST] <kbarry> The comparisons make it a lot easier.
[18:39:58 CEST] <furq> do note that the bitrates are a bit different
[18:40:00 CEST] <furq> particularly mp3
[18:40:26 CEST] <A124> Some people just hate mp3, cause it sucks. Or some will love you for not using mp3:
[18:40:49 CEST] <A124> Just AAC when you can xD I do not know of any crowd that hates AAC below 96kbps.
[18:40:53 CEST] <A124> *above.
[18:40:58 CEST] <furq> also yeah that chart illustrates well that not all encoders are created equal
[18:41:11 CEST] <furq> 96k faac is much worse than 96k qaac or fdk
[18:41:25 CEST] <A124> Sadly, yeah.
[18:41:26 CEST] <furq> probably much worse than 96k lame
[18:41:33 CEST] <furq> if you can imagine such a thing
[18:42:21 CEST] <A124> kbarry Btw, we stream tons of audio daily, and use that both for voice and music, using OPUS, or older alternative (CELT MONO)
[18:43:18 CEST] <A124> 20ms latency, great masking of lost packets, decent data savings (10-12kB/s per client)
[18:43:52 CEST] <furq> this is all moot if you're using rtmp anyway, since your only real choice there is aac
[18:44:01 CEST] <A124> If latency was not concern and was music only, aac would be as great. with OPUS you can go down to 5ms for realtime stuff.
[18:44:14 CEST] <kbarry> furq see, there is a perfect example of a limit in my knowledge,
[18:44:24 CEST] <kbarry> IF one isn't using RTMP, what else would one use?
[18:44:31 CEST] <A124> kbarry What you said you are using?
[18:44:42 CEST] <kbarry> Well, currently we use RTMP (mostly)
[18:44:46 CEST] <kbarry> But are working towards HLS
[18:44:53 CEST] <A124> Use AAC, that works almost everywhere.
[18:44:57 CEST] <A124> And is great.
[18:44:58 CEST] <kbarry> For myriad reasons.
[18:45:11 CEST] <furq> latency obviously isn't an issue then
[18:45:14 CEST] <DHE> HLS officially only supports h264 and AAC
[18:45:30 CEST] <A124> Problem resolved. Solution found.
[18:45:31 CEST] <DHE> any other codecs are at the implementation's discretion and not reliable
[18:45:55 CEST] <furq> kbarry: you can do rtmp and hls at the same time with nginx-rtmp
[18:46:00 CEST] <furq> assuming that's not what you're using already
[18:47:11 CEST] <kbarry> typically we are dealing with incoming streams, some live broadcasts (say a concert), and static content (that concert last week),
[18:47:26 CEST] <kbarry> were allowing for the scheduling between these two "types" of content.
[18:47:35 CEST] <kbarry> i'm going to look up nginx-rtmp
[18:47:47 CEST] <furq> https://github.com/arut/nginx-rtmp-module/
[18:48:13 CEST] <kbarry> https://obsproject.com/forum/resources/how-to-set-up-your-own-private-rtmp-server-using-nginx.50/
[18:48:48 CEST] <kbarry> question,
[18:48:50 CEST] <furq> https://gist.github.com/fur-q/d7028f51c38f7d0bb56e
[18:48:54 CEST] <furq> that's more or less the setup that i use
[18:49:05 CEST] <kbarry> "Recording streams in multiple FLVs", in that contest, what is 'FLVs'
[18:49:13 CEST] <kbarry> context*
[18:49:22 CEST] <furq> https://en.wikipedia.org/wiki/Flash_Video
[18:49:30 CEST] <furq> that's the only format that rtmp supports
[18:49:40 CEST] <kbarry> OK, its just saying different bitrates/?
[18:50:11 CEST] <furq> maybe? i've never used the recording stuff
[18:51:06 CEST] <kbarry> furq, you are then acting as your own CDN?
[19:01:15 CEST] <devster31> is yasm supported on ARM? it's suggested in the compilation build for ubuntu but not for the rpi
[19:05:15 CEST] <kbarry> furq why do you use nginx-rtmp instead of something OOTB?
[19:05:31 CEST] <kbarry> Maybe this is a newb question but, generally, how are you using nginx?
[19:05:38 CEST] <kbarry> (ie, are you providing VOD?)
[19:19:25 CEST] <kbarry> Ughhhhh,
[19:20:03 CEST] <kbarry> I have not been able to get a "stereo-mix" recoding device (Windows 7). I'm not sure that I have the technical chops to hack it together myself.
[19:20:19 CEST] <kbarry> Is there such a thing as ffmpeg capturing audi from a particular browse tab?
[19:20:30 CEST] <kbarry> (or maybe from a single application)
[19:20:39 CEST] <kbarry> as a sort of pseudo-stero-mix?
[19:24:11 CEST] <kbarry> Actually, Do any of you use a VAC? (virtual Audio Cable), There seem to be more than one option.
[19:25:16 CEST] <A124> https://i.imgur.com/Stt4MIP.jpg
[19:25:19 CEST] <A124> ehm bad chan
[19:49:50 CEST] <HeXiLeD> quick and fast question The V4L2 driver changed the video from 1024x720 to 640x480 anyone know what should i be looking for to get higher res ?
[20:04:17 CEST] <andrey_utkin> is there anything wrong with encoding h264 video using -pix_fmt yuvj420p? I see "deprecated pixel format used, make sure you did set range correctly" which is widely described as safely ignoreable if you use CLI util, but I also use API-based app. So yuvj is a setting from legacy codebase, should i transition to yuv420p (without j)? Thanks in advance.
[20:04:50 CEST] <Fyr> HeXiLeD, -vf scale=3840:2160
[20:05:00 CEST] <Fyr> make it bigger!
[20:06:22 CEST] <HeXiLeD>   
[20:06:23 CEST] <Fyr> andrey_utkin, Google says it's a deprecable format.
[20:06:23 CEST] <HeXiLeD>  
[20:06:29 CEST] Last message repeated 1 time(s).
[20:06:29 CEST] <Fyr> use yuv420p instead
[20:07:28 CEST] <HeXiLeD> i am looking for something like this ffmpeg -f v4l2 -framerate 30 -video_size 1024x720 -i /dev/video0 -c:v libx264 -b:v 1500k -an -f flv
[20:07:31 CEST] <HeXiLeD> for youtube
[20:07:38 CEST] <andrey_utkin> Fyr, seems right, but i am searching for some justification of this fact, to report to seniors. And the problem is that i don't know what yuvj thing is.
[20:08:03 CEST] <drv> the "j" stands for JPEG as far as I know
[20:08:51 CEST] <drv> it has to do with fullrange vs not
[20:09:00 CEST] <jkqxz> andrey_utkin:  Use yuv420p.  If you want to give it specific colour information then set it in the AVCodecContext.colour* fields rather than changing the pix_fmt.
[20:09:36 CEST] <Fyr> HeXiLeD, does it work or it doesn't?
[20:10:07 CEST] <Fyr> it's not answerable until there's no question.
[20:11:00 CEST] <andrey_utkin> jkqxz, i don't know if i want to give special color info, i think no. Now please help me with ridiculous question: what could be a reason behind using yuvj420p earlier in our app :)
[20:11:04 CEST] <jkqxz> The effect of using yuvj*p is entirely <http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/libx264.c;h=5030d65f79ef07af6335f897df4000534a486b14;hb=HEAD#l768>, otherwise it is treated identically to yuv420p.
[20:11:50 CEST] <HeXiLeD> Fyr: My code no. it falls back to 640 res
[20:12:36 CEST] <jkqxz> (And you are better setting it by that AVCodecContext.color_range method if that is what you want.)
[20:13:25 CEST] <Fyr> HeXiLeD, what does "$ ffprobe -i /dev/video0" return?
[20:14:43 CEST] <jkqxz> HeXiLeD:  Are you sure that it supports 1024x720?  That's a strange resolution, and generally those things have to come from a fixed list.  (Try 1024x768 or 1280x720.)
[20:16:37 CEST] <jkqxz> andrey_utkin:  They copied it from somewhere on the internet and it worked at the time, maybe?
[20:17:34 CEST] <HeXiLeD> http://dpaste.com/3DVJZZG Fyr
[20:18:31 CEST] <Fyr> HeXiLeD, you've just answered youself.
[20:18:38 CEST] <jkqxz> andrey_utkin:  To check, you could just look at the pixels of your input and see if any of the chroma values ever lie outside 16-235 or whatever the range is...
[20:18:39 CEST] <Fyr> only through -vf scale.
[20:18:53 CEST] <kepstin> HeXiLeD: if you want to see what resolutions that input supports, try running "ffmpeg -f v4l2 -list_formats all -i /dev/video0"
[20:19:22 CEST] <andrey_utkin> jkqxz, very probable. I wonder can it have something to do with compatibility with some player vendors, like apple devices or browsers
[20:23:59 CEST] <HeXiLeD> http://dpaste.com/3JCNDQ2
[20:24:01 CEST] <HeXiLeD> new cam
[20:24:19 CEST] <HeXiLeD> it falls back to the highest compressed
[20:24:54 CEST] <HeXiLeD> so if i want  1600x1200 wnat can it be done ?
[20:25:10 CEST] <Fyr> HeXiLeD, http://trac.ffmpeg.org/wiki/Capture/Webcam
[20:25:18 CEST] <Fyr> "Encoding example"
[20:25:36 CEST] <jkqxz> Remove the -framerate 30.  You can't carry 1600x1200 raw at 30fps over USB (I assume this is a UVC webcam).
[20:25:36 CEST] <kepstin> HeXiLeD: to use the compressed (mjpeg) format, you need to use the "-input_format mjpeg" option, then set the resolution
[20:25:56 CEST] <kepstin> that camera doesn't even support 1600x1200 raw at all
[20:26:22 CEST] <kepstin> note that using mjpeg mode, even at lower resolutions, will probably enable higher framerates over using the (default) raw
[20:28:35 CEST] <HeXiLeD> ummmm... i see that now... what if i use an ip cam ?
[20:28:45 CEST] <HeXiLeD> will ffmpeg work well with it ?
[20:28:52 CEST] <HeXiLeD> 1080P res cam
[20:29:28 CEST] <jkqxz> Why not?  It's on the list of resolutions.  It'll probably only go at 5fps, though.
[20:29:41 CEST] <HeXiLeD> -i http://camera1/video ? right ?
[20:29:50 CEST] <HeXiLeD> 5 ? ffmpeg limitation ?
[20:30:33 CEST] <jkqxz> In response to kepstin, sorry.
[20:30:37 CEST] <HeXiLeD> here is my problem. stream to youtube using linux because windows machines i have around cannot handle the workload with software being used. ie xsplit and others
[20:41:14 CEST] <kepstin> jkqxz: HeXiLeD: oh, I got the two mixed up.. the 1600x1200 mode is only on the raw, not mjpeg
[20:41:15 CEST] <kepstin> how odd
[20:41:32 CEST] <kepstin> I guess the jpeg encoder chip in the camera doesn't support frames that big :/
[20:41:40 CEST] <furq> do you really need more than 960*720 for youtube
[20:41:54 CEST] <furq> that's enough to trigger youtube's HD detection, and at 1.5mbps it'll probably look nicer
[20:43:06 CEST] <jkqxz> "uvcdynctrl -f" will give you a full list of formats/resolutions/framerates, so you can precisely specify something which is exactly supported without guessing.
[20:51:22 CEST] <t4nk858> hi, I'm generating a file with specific requirements using this command "ffmpeg -i recording.mp4 -vsync vfr -q:v 9 recording.mjpeg". This works great, now I want to replicate the same within my own program using libffmpeg. I have never used the library, how might I do that?
[20:51:40 CEST] <JEEB> there's no libffmpeg
[20:51:45 CEST] <JEEB> there's only the separate libraries
[20:51:59 CEST] <JEEB> and you can see how to use them by looking at the examples in the repository
[20:52:11 CEST] <JEEB> there are examples on anything from demuxing to decoding to filtering to encoding
[20:52:34 CEST] <t4nk858> hmm
[20:52:36 CEST] <JEEB> https://github.com/FFmpeg/FFmpeg/tree/master/doc/examples
[20:52:40 CEST] <t4nk858> thanks
[20:52:55 CEST] <JEEB> the actual repo is on git.videolan.org, but the github preview is just prettier
[20:53:05 CEST] <JEEB> that's why I web link there usually when pointing towards code
[20:53:31 CEST] <t4nk858> is there an easy way to look at that the ffmpeg cli source code to get a more or less ready chunk doing what the command I pasted does?
[20:53:57 CEST] <t4nk858> I guess that wouldn't be too straight-forward...
[21:05:18 CEST] <t4nk858> is this the CLI source code? https://github.com/FFmpeg/FFmpeg/blob/master/ffmpeg.c
[21:09:31 CEST] <HeXiLeD> jkqxz:
[21:09:51 CEST] <HeXiLeD> luvcview ?
[21:56:55 CEST] <HeXiLeD> -g brings me help
[21:57:34 CEST] <Fyr> help is good.
[21:57:54 CEST] <HeXiLeD> i mean no help
[21:59:14 CEST] <HeXiLeD> https://bpaste.net/show/d3c9445c5c45
[21:59:27 CEST] <HeXiLeD> that is what i have now
[22:00:44 CEST] <fling> https://bpaste.net/show/2211aee26a44
[22:14:58 CEST] <llogan> HeXiLeD: you need to show the complete console output from your command and describe what is wrong
[22:17:18 CEST] <llogan> HeXiLeD: why are you using -c:v flv instead of libx264? also, you are using some audio options but using -an
[22:40:28 CEST] <HeXiLeD> connects llogan because currently i don't have that support in ffmpeg. compiling at the moment
[23:03:16 CEST] <brontosaurusrex> how would I hide libx264 info from stdin? (and just keep the framecounter, ETA stuff)
[23:06:47 CEST] <furq> brontosaurusrex: -v quiet -stats
[23:09:13 CEST] <brontosaurusrex> furq: thanks
[23:20:25 CEST] <brontosaurusrex> another one: do i need any atom movers this days?
[23:25:15 CEST] <brontosaurusrex> like qt-faststart
[23:31:11 CEST] <llogan> brontosaurusrex: you can just use "-movflags +faststart" option
[23:31:35 CEST] <llogan> generally only needed for playbkac via progressive download
[23:33:16 CEST] <brontosaurusrex> llogan: and those are not default?
[23:34:49 CEST] <llogan> it is not set by default
[23:34:58 CEST] <brontosaurusrex> ok, thanks
[23:49:58 CEST] <xxDCDxx> hi, how can I compile the ffmpeg CLI from source?
[23:50:07 CEST] <c_14> ./configure; make
[23:50:31 CEST] <c_14> https://trac.ffmpeg.org/wiki/CompilationGuide
[23:51:03 CEST] <xxDCDxx> thanks! I'll give it a try. I'm trying to cross compile it for iOS, using this: https://github.com/kewlbear/FFmpeg-iOS-build-script  But it only built the libraries
[23:52:57 CEST] <c_14> https://github.com/kewlbear/FFmpeg-iOS-build-script/blob/master/build-ffmpeg.sh line 16
[23:53:03 CEST] <c_14> remove the --disable-programs switch
[23:54:16 CEST] <xxDCDxx> trying that now! c_14, you're a lifesaver, thank you very much!
[00:00:00 CEST] --- Wed May  4 2016



More information about the Ffmpeg-devel-irc mailing list