[Ffmpeg-devel-irc] ffmpeg.log.20150127

[burek]

[01:09] <dvk> i get  warning: first frame is no keyframe, how can i make sure that it is a keyframe?
[01:09] <dvk> i'm using -ss time
[01:41] <MaudeLebowski> Hey there.  I am trying to create a video out of a set of jpgs that are named by unix timestamp.  The pics arent one after another, i.e. 1.jog, 2.jpg etc so it breaks after just 1 image.  Without renaming the files, how would you make a video starting with 1.jpg for example if the next image were 3.jpg, 7.jpg, and so on.
[01:43] <MaudeLebowski> I have looked everywhere for help, and all I get is rename the images which I really dont wanna do for every video I make.
[01:43] <MaudeLebowski> I have heard of piping images into it.
[01:43] <MaudeLebowski> But cant seem to get the right list of images on the fly.
[01:43] <c_14> look at the glob part of image2
[01:43] <MaudeLebowski> Like %
[01:44] <MaudeLebowski> I did do that.
[01:44] <c_14> https://ffmpeg.org/ffmpeg-formats.html#image2-1
[01:44] <c_14> -pattern_type glob
[01:44] <c_14> *.jpg
[01:44] <MaudeLebowski> But it stops after 1 image
[01:44] <MaudeLebowski> if they arent sequential in number
[01:44] <c_14> not glob
[01:44] <MaudeLebowski> Ok
[01:44] <MaudeLebowski> Will look into this
[01:44] <c_14> do ls *jpg
[01:44] <c_14> it does the same thnig
[01:44] <c_14> *thing
[01:45] <MaudeLebowski> Well *.jpg wouldnt be what I need unfortunately itd be like anything from 114.jpg to 489.jpg but it might not have 114.jpg or 489.jpg
[01:45] <MaudeLebowski> I know globs should be able to do that just cant figure out what pattern
[01:50] <MaudeLebowski> Globs is likely my solution though.  Ill keep working with it.  Thanks for the help.
[02:21] <coherence> i may simply be confused, but is it possible (or even reasonable) to try to set PTS in mpegts audio streams? (there is a tiny bit of silence at the start of the files after conversion from AAC - maybe it's the AAC priming frames)
[02:47] <pentanol> coherence may be -er flag?
[02:47] <pentanol> no, -re flag
[02:47] <coherence> pentanol: thanks, using it already
[02:48] <coherence> i suppose a command line would be in order
[02:50] <coherence> pentanol: are you suggesting that -er might be responsible for the artifacts?
[02:50] <coherence> i mean -re
[03:17] <ramp> hi hi
[03:17] <ramp> can anyone help me with using ffmpeg to embed a subtitle file?
[03:19] <relaxed> possibly, but we need more details than that.
[03:19] <ramp> okay
[03:19] <ramp> i have some mkv files, and some ac3 files and a .sub file
[03:20] <ramp> and i'm using ffmpeg to output an mkv file with the relevant audio streams and subtitles embedded
[03:20] <ramp> I'm just trying to figure out the format to embed the subtitles
[03:20] <relaxed> you want the subs included in the container or burnt into the video?
[03:20] <ramp> just included in the container
[03:21] <relaxed> ffmpeg -i video -i audio -i sub -map 0 -c copy out
[03:21] <ramp> there's only one -map command?
[03:22] <ramp> right now the thing I'm using looks like this
[03:22] <ramp> ffmpeg -i $input -i $a1 -i $a2 -map 0:v -map 1 -map 2 -metadata:s:a:0 language=eng -metadata:s:a:1 language=commentary -c copy -shortest $output
[03:22] <ramp> $input, is the input file name, $a1 is the first audio track, $a2 is the second audio track and $output is the output filename
[03:23] <ramp> sure, no problem
[03:23] <relaxed> yeah, using -map 0 was wrong on my part
[03:24] Action: relaxed just woke up
[03:25] <ramp> http://pastebin.com/iw32Rdtr
[03:25] <ramp> so i'm just wondering where to include $sub in that command
[03:25] <ramp> and what modifications I need to make
[03:26] <relaxed> why is -shortest in there?
[03:26] <ramp> in case there are slight differences in file lengths
[03:26] <ramp> it's just a precaution
[03:27] <relaxed> read fflogger's message again
[03:27] <ramp> what in particular would you like me to show you
[03:28] <relaxed> the bot was pretty specific
[03:30] <ramp> if you'd like to help me I'd be most appreciative, but if you want me to guess what information you require I might as well guess how to interpret the docs
[03:33] <relaxed> apart from the missing -i $sub in the script?
[03:34] <ramp> there also needs to be some accompanying -metadata I think?
[03:39] <relaxed> I think you can set the language but it can be omitted too
[03:40] <coherence> how would one extract just the audio stream from an AAC?
[03:41] <relaxed> are there multiple audio streams? you want raw aac?
[03:41] <coherence> relaxed: i *think* i want raw bitstream, just one stream in a file.  i want to remove the priming and remainder bits.
[03:41] <ramp> if I just include -i $sub the output file does not have subtitles
[03:42] <relaxed> you need -map 3
[03:42] <ramp> cool thanks, i'll try that now
[03:43] <relaxed> coherence: try, ffmpeg -i input -map 0:a -c copy output.aac
[03:43] <coherence> relaxed: thank you, i'll try now.  i'm clueless in audio codecs/formats, just learning and the ffmpeg docs are harder for me than man pages :)
[03:43] <ramp> i'm getting an error now
[03:43] <ramp> http://pastebin.com/0uhAqiTk
[03:44] <relaxed> ramp: "Subtitle codec 1833195076 is not supported."
[03:45] <ramp> what does that mean? the format my subtitles are in is not supported by ffmpeg?
[03:47] <relaxed> ramp: try converting them to the ass sub format first. ffmpeg -i input.sub out.ass
[03:48] <ramp> cheers
[03:51] <relaxed> coherence: check out the wiki too if you haven't already. https://trac.ffmpeg.org/wiki
[03:51] <coherence> thanks relaxed
[03:55] <coherence> relaxed: my questions stem from using ffmpeg to push AAC ADTS files to nginx-rtmp to be served using HLS, and the resulting mpegts files *i think* have a tiny bit of dead air at the start
[03:55] <coherence> i'm searching for the cause, and the priming bits in AAC files have my attention
[03:57] <coherence> although i'm also curious about the MPEG2-TS PTS delay as specified here: http://www.bretl.com/mpeghtml/timemdl.HTM
[03:58] <coherence> which is why i wondered if i could use ffmpeg to set the PTS in a given file
[04:19] <pentanol> coherence did you get wrong PTS\DTS?
[04:33] <coherence> pentanol: /bufbri
[04:33] <coherence> whoops
[04:35] <coherence> pentanol: the start_time shown for the audio stream when using ffprobe -show_streams is 0.7
[04:35] <coherence> but should be 0
[04:36] <coherence> so i *think* that means PTS is off
[04:36] <coherence> sorry, i'm guessing
[05:15] <{MiltonBerle}> Why would I get- image2pipe No such file or directory
[05:15] <{MiltonBerle}> When I am definitely piping in files
[05:15] <{MiltonBerle}> args=''; for x in `find . -type f -newer 1422284581.jpg ! -newer 1422284761.jpg`; do args="$args $x";done ; cat $args | ffmpeg -framerate 1/1 -i image2pipe -c:v libx264 -r 30 -pix_fmt yuv420p testing_movie.mp4
[05:15] <{MiltonBerle}> example
[05:15] <{MiltonBerle}> The commands before cat by themselves show files
[05:16] <{MiltonBerle}> If echoed
[05:32] <coherence> pentanol: http://pastebin.com/4ucZ3s9t
[05:32] <coherence> first packet from the corresponding audio stream in each file
[05:33] <coherence> the MPEG2-TS file shows a PTS of 63000 wihile the AAC source file has a PTS of 0
[05:33] <coherence> again could be my ignorance
[05:33] <coherence> but that seems wrong
[05:49] <pentanol> coherence what happen if you reencode this file with ffmpeg -i ... ?
[06:05] <coherence> pentanol: you mean the TS file back to AAC?
[06:05] <coherence> (interesting question anyway)
[06:09] <coherence> looks like the pts_times convert back to the original/like the source file
[06:10] <coherence> size and pos is off though
[06:25] <pentanol> coherence after convert with ffmpeg pts still 0.7 ?
[06:31] <coherence> pentanol: converting AAC -> TS using ffmpeg, pts is 1.4.  converting back TS -> AAC, pts is 0
[06:47] <pentanol> coherence .aac to .ts ? what's your command with ffmpeg?
[06:52] <coherence> pentanol: http://pastebin.com/rEj2cbVP
[06:53] <coherence> and some variations on that playing with -ar and -ab
[06:53] <coherence> /that/those/
[07:03] <Darby_Crash> hi to all
[07:04] <Darby_Crash> thanks for ffmpeg
[07:05] <brdxufan> hello
[07:05] <Darby_Crash> have someone a script for compiling ffmpeg/ffplay on nix?
[07:06] <Darby_Crash> hi brdxufan
[07:07] <Darby_Crash> i need a script for a static build of ffmpeg ffplay
[07:10] <brdxufan> av_new_stream(),   how to use?
[07:11] <brdxufan> who can help me?
[07:30] <pentanol> coherence I made the test out.ts with start_pts=126000 out.aac with start_pts=N/A
[07:31] <pentanol> You don't have wrong pts\dts when you paying the file?
[07:31] <pentanol> it not, that's okay
[07:31] <pentanol> of*
[07:31] <pentanol> if*
[08:54] <k_sze[work]> How does ffmpeg deal with files that have been incompletely written?
[08:55] <k_sze[work]> e.g. if I take an intact AVI file and chop off the last 9 kilobytes.
[09:34] <storms> Hi. I'm trying to scale videos to a 720 height, keep width but also use square pixels. -vf "scale=-1:720,setsar=1" makes a video with wrong aspect ratio in regards to the width vs height on a video with 4:3 SAR.
[09:34] <storms> A simple way to do this?
[09:39] <Vladimir_> HI to all
[09:40] <Vladimir_> can somebody help me..
[09:40] <Vladimir_> I need to mjpeg stream to mix in one
[09:40] <Vladimir_> http:mjpeg stream... 1280x720   ----> to be 2560x720
[10:02] <ArunC> Using ffmpeg cmd line, I'm trying to split a video at the same time, preserve the timestamps. At present I can split the video but I am not able to preserve the timestamp. Any idea which option I should be using?
[10:05] <ArunC> ie, if the video is 10 min long and if each video chunk is 5 min, the second chunk's starting timestamp should be 05:00 and not 00:00.
[10:20] <BtbN> how should that work? 5 minutes of nothing in the beginning?
[10:25] <ArunC> BtbN: Yeah, sort of. Just record the offset as the timestamp for the initial frame.
[10:25] <BtbN> I don't think that's possible. Might be possible for audio, but depends a lot on the container.
[10:27] <ArunC> BtbN: OK, primarily we deal with surveillance videos. So splitting is necessary to process the videos in parallel, without losing the frame timestamp.
[10:28] <BtbN> Sounds like bad processing software, which isn't able to offset an entire file
[10:28] <ArunC> BtbN: Actually, the processes reside on different boxes and not all can afford to download the entire video file (50+GB). :)
[10:29] <BtbN> And you can't just tell it "this file is offset by X minutes"?
[10:30] <ArunC> No, as that process is a 3rd party s/w and that the interfaces are fixed now.
[10:31] <BtbN> Well, you can insert black frames at the beginning, but you will have to re-encode for that.
[10:41] <Vladimir_> hi
[10:41] <Vladimir_> I tray to mix to stream mjpeg in on...
[10:41] <Vladimir_> what is best solution for that ?
[10:42] <Vladimir_> using filter ?
[10:45] <t4nk476> get_bits seeking functionality?
[11:00] <ArunC> Found what I wanted. We can use -itsoffset to specify the offset in the output video chunk.
[11:00] <ArunC> BtbN: ^
[11:02] <BtbN> Doesn't that add an offset to the entire input video?
[11:03] <BtbN> As it's an input parameter
[11:06] <ArunC> oh yeah, my bad..
[11:06] <ArunC> :(
[11:07] <ArunC> its an input param
[11:22] <ArunC> BtbN: Not sure if -start_at_zero will help.
[11:44] <cousin_luigi> What's the difference between using opus and libcelt to decode CELT?
[12:33] <D0L1k> hi all
[12:35] <D0L1k> We have a problem with timeshifting. We need to get concrete time sequence of file. We are using ffmepg 1.12, we are producing mpegts, we can connect and watch that file (streaming is allright). We can also start it from defined time/date, but we dont have duration of that file.. So in VLC/etc. we can't see duration
[12:36] <BtbN> I don't think there is a version 1.12
[12:37] <D0L1k> *1.2.11
[12:37] <D0L1k> sorry
[12:38] <BtbN> I'd recommend updating.
[12:39] <cousin_luigi> I'll rephrase my previous question: what codec overlap is there between libcelt and libopus and what happens when both are linked against?
[12:39] <D0L1k> In new version (for ex. 2.5.3) we can't use codec copy on ffserver
[12:40] <BtbN> With a recent enough ffmpeg, you can do ffmpeg -i someinput.mkv -ss start -t duration -c:v ... -c:a ... youroutput.mkv
[12:40] <D0L1k> but also there is not possibility of using "start from time for duration"
[12:40] <BtbN> You can't get a specific sequence from a file without re-encoding.
[12:41] <BtbN> ffserver is not exactly maintained anymore
[12:41] <BtbN> Depending on the codec you can start from an I frame, by putting -ss in front of the input
[12:43] <D0L1k> and is there any other software, where we can use timeshifting? .. in ffserver timeshifting is ok, but we can't select endtime/enddate, just startdate..
[12:48] <D0L1k> or how can we select sequence from ffm file? for example this works: "http://localhost:8080/nova.ts?date=2015-01-27T09:00:00" .. but we want something like "http://localhost:8080/nova.ts?date=2015-01-27T09:00:00&enddate=2015-01-27T09:30:00".. we can edit source code of ffserver, just need to know how set duration..
[16:55] <cousin_luigi> JEEB: Hello. Do you happen to be around?
[21:53] <Projectns> Hello i have a short quesition, is it posible to stream a video list (more than one video ) to a rtmp server ?
[21:53] <Projectns> *with ffmpeg
[21:54] <kepstin-laptop> you could use any of the supported methods of concatenating videos in ffmpeg to generate a single stream, which could then be sent over rtmp
[21:54] <Projectns> do u have some links to tutorials ?
[21:54] <kepstin-laptop> http://www.ffmpeg.org/faq.html#How-can-I-concatenate-video-files_003f
[21:55] <Projectns> or can i choose a directory with videos ?
[21:55] <Projectns> who play all videos from there
[21:57] <kepstin-laptop> no, but if you have a bunch of files that all have maching codecs and other parameters, then you might be able to put their names into a list file then use the concat demuxer/format to load them.
[21:57] <Projectns> ah cool , that is exactly what i want
[21:58] <kepstin-laptop> if they might have different codecs, it's more complicated since you have to decode them before concatenating (in this case, you'd use the concat filter)
[21:59] <Projectns> ok , i try to convert the videos what i want and add them to the ffmpeg
[21:59] <Projectns> is the file level with cat the command?
[22:00] <kepstin-laptop> file level concatenation is equivalent to using the cat command, yes.
[22:00] <Projectns> ok
[22:00] <Projectns> its sound hard... :D
[22:01] <Projectns> do u know some other broadcast tools for linux ?
[22:01] <Projectns> or some scripts or pgoramm
[22:01] <Projectns> progrma
[22:02] <kepstin-laptop> I think most rtmp clients handle streams being stopped/started just fine, so one thing you might try is just running ffmpeg separately on each file.
[22:02] <kepstin-laptop> e.g. using a shell script to loop over the files.
[22:03] <Projectns> mhh ok
[22:04] <Projectns> so my idea is to create a rtmp stream with some videos for the internet. And the server should running on a linux (Debian Server) without a gui...but the problem is to find a way to stream my video list to the rtmp server
[22:07] <Projectns> i found something https://trac.ffmpeg.org/wiki/Concatenate
[00:00] --- Wed Jan 28 2015