[Ffmpeg-devel-irc] ffmpeg.log.20150304

burek burek021 at gmail.com
Thu Mar 5 02:05:01 CET 2015


[00:30] <ppw> hey
[00:30] <ppw> is fdk-aac now included as standard in ffmpeg?
[00:31] <c_14> You can compile it in, but distributed with any builds.
[00:31] <c_14> (Licensing reasons)
[00:31] <c_14> *but it's not
[00:32] <ppw> so an eventual 'git pull' will still give me the source?
[00:32] <c_14> hmm?
[00:32] <llogan> it's not a native encoder. it's an external library that ffmpeg can use to encode with
[00:33] <ppw> ah.
[00:34] <llogan> it seems to only be nonfree with gpl
[00:34] <llogan> so i suppose it could be distributed with a build that lacks --enable-gpl (and --enable-nonfree)
[00:51] <c_14> Hmm, so I did some quick searching about libfdk-aac and (l)gpl compatibility, and nobody's quite sure whether it's only gpl incompatible or also lgpl incompatible. So if you are thinking of distributing such a build, you might want to ask a lawyer...
[01:02] <JEEB> IIRC it was incompatible with both, someone actually poked a lawyer for it
[01:02] <JEEB> you might want to poke on #ffmpeg-devel for gritty details if you need those
[03:41] <ironcolt> Hello room, good evening
[07:22] <ac_slater> hey all. I've been battling libavformat for a bit. I have some weird data stream (that ffmpeg supports). Im muxing the data into an mpegts format. The data makes it there (and works), but the logs spits out '... non monotonically increasing dts ...'. Any clues?
[09:18] <hendry> ttps://lists.ffmpeg.org/pipermail/ffmpeg-user/2015-March/025562.html <--- any ideas guys ?
[09:49] <relaxed> hendry: does it happen with ffplay?
[09:51] <hendry> relaxed: no... it doesn't
[09:51] <hendry> relaxed: playback doesn't stutter
[09:53] <relaxed> Then I recommend you start using mpv instead of mplayer
[09:53] <relaxed> http://mpv.io/
[09:59] <hendry> relaxed: mpv uses ffplay ?
[10:00] <hendry> relaxed: thanks btw
[10:00] <hendry> relaxed: is there a better intermediate format that what I'm using? Something takes less space?
[10:01] <hendry> relaxed: and any ideas about the decomposed colours ?
[10:01] <relaxed> it's probably from color space conversion
[10:02] <relaxed> look at "Lossless recording" from https://trac.ffmpeg.org/wiki/Capture/Desktop
[10:03] <hendry> relaxed: oooh, that looks good !
[10:03] <relaxed> mpv uses ffmpeg's libs to demux/decode like ffplay does
[10:30] <hendry> can anyone recommend a HDMI -> USB3 -> UVC -> FFMPEG capture device?
[17:12] <sg90> Hi, I am having some issues regarding ffmpeg terminal output. I am receiving lots of "ignoring invalid SAR: 0/0" messages, which repeatedly split out the bottom line of the output, which I am intending to parse out things like frame, fps etc. My ffmpeg command is here: http://pastebin.com/EjpjtXjk , I am piping video input into ffmpeg via nut. Thanks!
[17:27] <ychaouche> Hello guys
[17:27] <ychaouche> I want to add some background music to a video already containing audio (speech)
[17:28] <ychaouche> The background music should be very low
[17:28] <ychaouche> How do I do that ?
[17:33] <c_14> Pass the background song through a volume filter and then amix it with the other track
[17:35] <ychaouche> for example -af 'volume=0.2' ?
[17:37] <ychaouche> Is it true that ffmpeg changed its name ?
[17:37] <Mavrik> no.
[17:38] <ychaouche> I've been told that in #linuxmint-help on their IRC server
[17:38] <ychaouche> because I didn't find the ffmpeg package
[17:38] <ychaouche> Maybe it's in a "tainted" repo or something
[17:38] <c_14> ychaouche: something like -filter_complex '[0:a]volume=0.2[b];[1:a][b]amix'
[17:38] <kepstin-laptop> ubuntu (and thus mint) doesn't include the ffmpeg package.
[17:38] <ychaouche> oh really ? I didn't know that.
[17:39] <ychaouche> thanks for the info kepstin-laptop
[17:39] <ychaouche> So I guess I'll have to compile it myself
[17:39] <kepstin-laptop> this'll probably change at some point, now that debian has ffmpeg again, but not for a while until it trickles through :)
[17:39] <c_14> ychaouche: you can also use the static build
[17:40] <c_14> (in most cases)
[17:40] <ychaouche> Ok I'll try that, although I don't know what it is exactly :-(
[17:40] <ychaouche> like is that pre-compiled binaries and libraries that will directly be copied to their usual locations ?
[17:40] <ychaouche> no compilation required ?
[17:41] <c_14> ye
[17:41] <c_14> http://johnvansickle.com/ffmpeg/
[17:41] <c_14> Just place the ffmpeg binary wherever you want and use it.
[17:42] <ychaouche> thank you c_14
[17:42] <ychaouche> is avconv a sister project or totally unrelated ?
[17:42] <ychaouche> linuxmint advises to use that instead of ffmpeg.
[17:43] <ychaouche> I have never heard of it before I think.
[17:44] <ychaouche> oh
[17:57] <sg90> I've resolved my invalid SAR issue, as well as setting the -s flag I also had to specify the -aspect flag, my console output it much clearer. Quite surprised ffmpeg does not pick up the aspect from input files. Is there a loglevel that will collapse the "frame" line output into a single line?
[18:03] <kepstin-laptop> hmm? ffmpeg should pick up the aspect from input files. It obviously can't do this if the input files are wrong or are missing the information, of course.
[18:06] <sg90> the output files were being written correctly at the right aspect ratio, it was just the repeated messages to the terminal that were the main issue for me.
[18:34] <cortexman> does ffmpeg have a way to do vad and silence removal?
[18:36] <cortexman> nm
[18:57] <cortexman> i can't get silenceremove to work
[18:57] <cortexman> i have a pretty clean audio file with no noise during silence
[19:16] <JoshX> Hi, has anyone here ever successfull streamed anything to the youtube rtmp live streaming service?
[19:16] <JoshX> ffmpeg just starts, no errors, sometime the youtube streaming site tells me a few seconds the stream is good
[19:16] <JoshX> and then is tells me my encoder is not sending data
[19:33] <rmbeer> hello, i have a problem
[19:35] <rmbeer> if used the ffmpeg with -ss, the ffmpeg process all of video from 0 to -ss. i have a one way for avoid this?
[19:35] <kepstin-laptop> rmbeer: it depends where you put the -ss in the command line
[19:35] <kepstin-laptop> if you put it before the input file it will seek, which should be faster
[19:36] <rmbeer> i used this command: "ffmpeg -i out4.mp4 -r 1 -ss 00:55:49.000 -t 1 out5.mp4" , but this process all video from 0 to -ss for none, and -ss to -t for get 1 image.
[19:36] <c_14> move -ss before -i, and use -frames:v 1
[19:37] <rmbeer> thanks, i testing with this.
[19:39] <rmbeer> yes, this function very fast. :D thanks.
[19:39] <rmbeer> this parameters not have finder in the web. :S
[19:41] <rmbeer> goodbye :)
[21:21] <msmithng> what is the best way to apply watermarks to multiple output streams, with -filter_complex, the mapping gets a little wonky
[21:21] <msmithng> as in if you try to map it to another variant output, it will bail
[21:21] <msmithng> with Could not write header for output file #0 (incorrect codec parameters ?): Invalid argument
[21:27] <msmithng> easy enough. thanks llogan
[21:27] <llogan> i am an automaton summoning automaton
[21:30] <JoshX> :)
[21:30] <JoshX> lets see
[21:35] <JoshX> http://pastie.org/10000367
[21:36] <JoshX> that is my script and the output (ip's and keys removed of course)
[21:36] <JoshX> and youtube tells me: We are not receiving data from your encoder. Please make sure it is configured correctly in the Ingestion Settings page.
[21:37] <JoshX> but if i only change my destination to a rtmp server using nginx
[21:37] <JoshX> on my home server
[21:37] <JoshX> it streams flawlessly
[21:37] <klaxa> that paste is empty for me
[21:38] <c_14> It's fine here.
[21:38] <c_14> JoshX: -deinterlace is deprecated
[21:38] <c_14> use the yadif filter (or similar)
[21:39] <JoshX> oh this particular stream is not even interlaced so i can remove it
[21:40] <JoshX> removed it.. so that cant be it
[21:40] <JoshX> youtube every now and then tells me that the stream is ok.. (for about 2 seconds) and then its gone again
[21:41] <c_14> try -force_key_frames 2
[21:41] <c_14> Technically the gopsize should deal with that, but I've had people with similar issues
[21:41] <c_14> And since it works for 2 seconds that could be the issue
[21:44] <llogan> -maxrate should be added too
[21:45] <JoshX> llogan set to what?
[21:45] <llogan> some basic info here http://trac.ffmpeg.org/wiki/EncodingForStreamingSites
[21:45] <smo_> hi, if i use ffmpeg to stream http links is there a way to know the "download" speed ?
[21:45] <JoshX> yeah seen that page
[21:49] <msmithng> sorry mine took a bit longer to sanitize.
[21:49] <msmithng> http://pastebin.com/Rk1A8HW7
[21:50] <msmithng> I must not understand the mapping when it comes to scenarios like this
[21:50] <msmithng> thinking I might actually remove -bt and use -crf instead
[21:52] <JoshX> llogan, c_14 : http://pastie.org/10000410
[21:52] <JoshX> it's still not showing up in youtube
[21:53] <c_14> msmithng: can you change all your implicit maps to explicit ones? It makes it easier to see what's happening
[21:57] <c_14> JoshX: what resolution is your content?
[21:57] <msmithng> c_14: gladly& can you give me an example?
[21:57] <c_14> msmithng: add [padname] to the end of your filter_complexes and -map '[padname]' before the output
[21:57] <klaxa> c_14: it says 1920x1080 in the paste
[21:57] <c_14> Also make sure to map the audio stream[s]
[21:58] <c_14> so it does
[21:58] Last message repeated 1 time(s).
[21:58] <c_14> The youtube live encoder settings page says the minimum bitrate for 1080p video is 3000kbps, your maxrate is 1000k
[22:00] <c_14> JoshX: you might also have to switch to cbr encoding (use -b:v) according to the page https://support.google.com/youtube/answer/2853702?hl=en
[22:00] <JoshX> ok.. hang on
[22:01] <JoshX> -b:v == vbr right?
[22:01] <klaxa> funniliy enough that's also in the command line already, maybe just drop the -maxrate and adjust your $VBR variable
[22:01] <kepstin-laptop> JoshX: no, that's "bitrate (for video)"
[22:01] <JoshX> ah ok
[22:01] <msmithng> c_14: and then use the [padname] to apply in the outputs? and does the -filter_complex need to be applied to each output variant?
[22:01] <c_14> msmithng: if it does the same stuff, no
[22:02] <c_14> (iirc)
[22:02] <JoshX> c_14: there is no audio.. source is an IP cam
[22:02] <JoshX> therefore the -an
[22:02] <JoshX> whould i just encode 'nothing' then?
[22:02] <c_14> JoshX: afaik youtube requires an audio stream, generate an empty one with -af aevalsrc=0 and feed it minimal bitrate
[22:03] <msmithng> hmm& in the working instances I can get the logo on the first output but then the second doesnt get it
[22:03] <c_14> msmithng: try adding a split as the last part of the filter_complex, and map the first pad in the first output and the second in the second
[22:06] <JoshX> c_14 what would my audio line be when i need an 'empty' stream with 128kbps mp3?
[22:06] <msmithng> and that can be done in a filter_complex? Ill give it a shot
[22:06] <c_14> -af aevalsrc=0 -b:a 128k -c:a libmp3lame
[22:07] <c_14> msmithng: "[1:v]scale=64:62[wat];[0:v][wat]overlay=main_w-overlay_w-25:25,split[v1][v2]"
[22:07] <c_14> then just map '[v1]' for the first and '[v2]' for the second
[22:07] <JoshX> i was close :)
[22:08] <msmithng> TIL about split.
[22:08] <msmithng> :)
[22:13] <JoshX> grrrr
[22:13] <JoshX> i'm even resizing to 720p with all recomended settings
[22:14] <JoshX> and it shows for a few seconds and drops again
[22:15] <JoshX> 'March 4, 2015 at 10:13 PM (GMT+1)
[22:15] <JoshX> The health is good.'
[22:15] <JoshX> 'sure' ... :-(
[22:21] <JoshX> i'm lost.. the stream looks good, but it's not working
[22:21] <JoshX> and streaming with exactly the same settings to a nginx server with rtmp plugin works fine
[22:21] <c_14> try adding -level 4.1 ?
[22:21] <JoshX> in the vcodec section?
[22:22] <c_14> ye
[22:22] <llogan> did you try adding an audio stream as mentioned by c_14? (although it appears you have no MP3 encoder enabled)
[22:25] <JoshX> oh not sure i have libmp3lame
[22:25] <JoshX> let me recompile
[22:25] <JoshX> one moment
[22:25] <c_14> You can probably also just use the aac encoder, since it looks like they'll eat aac
[22:26] <JoshX> just c:a libaac
[22:26] <JoshX> ?
[22:26] <c_14> -c:a aac -strict -2
[22:27] <JoshX> i dont think i have aac as well :-/
[22:27] <JoshX> lets fix this
[22:27] <c_14> aac should be internal
[22:28] <JoshX> ah yes i have aac
[22:28] <JoshX> i just don't see it in the output
[22:28] <c_14> Do you still have the -an ?
[22:29] <JoshX> nope
[22:31] <JoshX> compiling with lame
[22:33] <joeflateau> I have an hls live stream, using hls_list_size=0 it contains all segments of the stream so clients can seek all the way back to the beginning.
[22:33] <joeflateau> I want to be able to let them cut a clip out of that, so I have an interface that lets them choose an in and out point and the server fires up an ffmpeg process with -ss and -t
[22:33] <joeflateau> the problem is...
[22:34] <joeflateau> the -ss is "seeking" from live instead of from when the stream started
[22:34] <joeflateau> is there a way to force the apple http live streaming demuxer to seek from the start of the whole stream instead of now?
[22:42] <JoshX> hmm i'm getting tons of Past duration 0.699989 too large
[22:42] <JoshX> i should change my preset?
[22:44] <c_14> JoshX: can you pastebin?
[22:44] <JoshX> sure
[22:44] <JoshX> what do you want?
[22:44] <JoshX> everything again?
[22:44] <c_14> Ye,
[22:44] <c_14> Whole commandline to see what's changed, and the output to see where the messages are coming from
[22:48] <JoshX> http://pastie.org/10000566
[22:48] <JoshX> its ffmpeg, just cloned from git and compiled about 10 minutes ago :)
[22:49] <JoshX> i changed the camera to 720p and changed the settings in youtube to be 1500-4000kbps and 720p
[22:49] <JoshX> i can't seem to see mp3 in the stream though?
[22:50] <c_14> try adding -f lavfi -i aevalsrc=0, then get rid of the -af aevalsrc=0
[22:51] <JoshX> the line now says: -f lavfi -i aevalsrc=0 -b:a 128k -c:a libmp3lame
[22:51] <JoshX> oh error
[22:52] <JoshX> Option force_key_frames (force key frames at specified timestamps) cannot be applied to input file aevalsrc=0 -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to.
[22:52] <c_14> eh, that has to be on the -i $SOURCE line
[22:52] <JoshX> Error parsing options for input file aevalsrc=0.
[22:52] <c_14> -i "$SOURCE" -f lavfi -i aevalsrc=0 \
[22:52] <JoshX> ok
[22:52] <c_14> Then shrink the other line to -b:a 128k -c:a libmp3lame \
[22:53] <JoshX> aha
[22:53] <JoshX> it seems i have an audio encoder too
[22:53] <JoshX> !!! We're preparing your live stream preview. Please wait a moment.
[22:54] <JoshX> w00t!
[22:54] <JoshX> this command line shold be written in the docs somewhere :)
[22:54] <JoshX> lets paste what it is now :)
[22:56] <JoshX> its not 'thrilling'
[22:56] <JoshX> https://www.youtube.com/watch?v=qlgXr0H_ZXE
[22:56] <JoshX> but it's live :)
[23:09] <JoshX> http://pastie.org/private/soplwzjspxh2nlrotwjnha
[23:09] <JoshX> that is what it is
[23:09] <JoshX> :)
[23:13] <JoshX> its just giving a lot of Past duration x.xxxx too large
[23:13] <JoshX> and the preset now is ultrafast
[23:13] <JoshX> and its a 4 core i7 with
[23:14] <JoshX> should i do something with -threads?
[23:14] <c_14> threads should be fine by default
[23:15] <JoshX> its not like its a heavy stream
[23:16] <JoshX> 2.5mbit 720p
[23:17] <JoshX> not sure whats causing the message
[23:17] <JoshX> vmstat shows ~95% idle
[23:17] <JoshX> cpu load is 0.3 on a 4 core (8 with hyper threading)
[23:18] <JoshX> not like it's stressed
[23:19] <JoshX> do you have any idea? c_14?
[23:23] <c_14> It's just a warning, so it shouldn't be fatal. Not sure why it's happening though.
[23:25] <JoshX> i see
[23:32] <JoshX> well it seems to work
[23:32] <JoshX> now lets test stability
[23:32] <JoshX> thanks for your help!
[23:33] <JoshX> and the script should be stored in a wiki somewhere ;)
[23:41] <sagax> hi!
[23:42] <sagax> how to make sound analysis on the signal / noise ratio in ffmpeg?
[23:42] <sagax> what read in docs?
[23:56] <sagax> analysis for sound
[00:00] --- Thu Mar  5 2015


More information about the Ffmpeg-devel-irc mailing list