[Ffmpeg-devel-irc] ffmpeg.log.20151223

[burek]

[00:02:32 CET] <Guiri> Does the `-override_ffserver` flag override everything, including the format?  I'm using -filter_complex with libx264 so I'd rather not double-encode stuff.  But I can only get the stream to work using -format ffm http://.../.ffm
[00:02:47 CET] <Guiri> But the ffserver.conf is configured to use Format flv
[00:04:00 CET] <Guiri> Unless there's a way to do the -concat server side and simply pipe a series of `-i` inputs into ffserver.
[00:58:52 CET] <xintox> is it possible to get a 2 minute sample from an mp4 video file?
[01:03:45 CET] <prelude2004c> no nobody was able to figure out why vdpau is not using the cpu for h264 content ?
[01:48:00 CET] <E-TARD> ok i'm trying to get ffmpeg to hls but i'm doing something worng http://pastebin.com/6Hi4KgQc
[01:50:55 CET] <E-TARD> if someone could just look at my command -> http://pastebin.com/6Hi4KgQc
[01:55:56 CET] <Rokam> E-TARD: what error do you get?
[01:57:41 CET] <E-TARD> oh i forgot to remove the -use_localtime 1 in the pastebin. just over look that but i get this "Cannot use rename on non file protocol, this may lead to races and temporarly partial files av_interleaved_write_frame(): Unknown error"
[01:58:48 CET] <E-TARD> does the .ts file need a url as well
[02:05:01 CET] <derekr> I'm trying to do segmented transcoding of one muxed input to several outputs, and the video seems to be working fine, but when I do audio the playback seems to have gaps (small blips around the segment boundaries) - any ideas on what would cause that?  I've tried two different methods, using -f segment, and manual segmenting using -ss/-t, both have the same issue, here's some example scripts: http://pastebin.com/4vfHE4fg http://pastebin.com
[02:11:35 CET] <MachinaeWolf> What would cause a slight skip in recording from line in to mic jack?
[02:31:00 CET] <Rokam> E-TARD: sorry.. don't know whats wrong. I just know that you need to provide the full url
[02:31:27 CET] <E-TARD> i dont know how
[02:32:00 CET] <E-TARD> like this -hls_segment_filename http://192.168.1.2:8060/live%03d.ts
[02:32:31 CET] <Rokam> yes
[02:34:33 CET] <E-TARD> Could not write header for output file #0 (incorrect codec parameters ?): Error number -138 occurred
[02:34:50 CET] <Rokam> thats why I can't help you
[02:34:51 CET] <Rokam> :)
[02:34:59 CET] <E-TARD> oh ok well thanks for trying ;)
[02:46:50 CET] <Rokam> E-TARD: I'm able to encode without the listen option
[02:48:50 CET] <E-TARD> so what just remove that and it works?
[02:50:22 CET] <Rokam> yes. And you can use a http server to host it
[02:50:33 CET] <E-TARD> no
[02:50:39 CET] <E-TARD> that will not work for me
[02:53:51 CET] <C0nundrum> How do i save a file as flv but name it "filename.mp4" ?
[02:54:03 CET] <furq> -f flv
[02:54:16 CET] <C0nundrum> ah, thanks
[02:54:24 CET] <furq> i don't see why you would want to do that though
[02:58:10 CET] <C0nundrum> ugh now that i thnk about it flv is imcompatible with the default html5 player
[03:01:32 CET] <C0nundrum> o neat conversion doesn't take to logn when copying streams
[03:07:19 CET] <Rokam> E-TARD: try ffserver on a linux machine
[03:07:29 CET] <Rokam> that will work too
[03:07:49 CET] <E-TARD> i have
[03:07:51 CET] <E-TARD> but
[03:09:39 CET] <E-TARD> i can not get hls to work on it. i can get mpeg-ts to work but not hls because of the the hls_segment_filename or segment_list there is no way to tell ffserver in the config to use options like that
[03:10:00 CET] <E-TARD> i have not found a way
[03:11:07 CET] <E-TARD> i have looked at lines of code & don't see a way to do it other then avoptionvideo or avpresetvideo
[03:12:00 CET] <E-TARD> users in this chat room have told me to not do it that way
[03:12:09 CET] <furq> ffserver is an unmaintained mess and you probably shouldn't use it
[03:12:10 CET] <DHE> E-TARD: I use HLS a fair bit. hls_segment_filename is where the .ts is written, so if it's a URL then it must be writable
[03:12:27 CET] <E-TARD> i know hls
[03:12:28 CET] <DHE> far easier to just point it at something under your DocumentRoot and write to local disks
[03:12:38 CET] <DHE> (ie. apache)
[03:12:59 CET] <E-TARD> what i don't know is how to make ffmpeg to serve up the m3u8 and .ts
[03:13:11 CET] <E-TARD> i know it can without the use of apache
[03:13:12 CET] <DHE> it doesn't. use apache, or nginx, or microsoft IIS, or whatever you use
[03:13:36 CET] <E-TARD> it will out put http so why not
[03:14:00 CET] <furq> if you want to serve from a remote machine then just nginx-rtmp
[03:14:03 CET] <furq> +use
[03:14:09 CET] <DHE> well you don't write to http typically unless you have publishing enabled
[03:14:27 CET] <E-TARD> what I'm trying to do
[03:15:17 CET] <E-TARD> is have ffmpeg push a stream form my pc to a server that will take that stream & publish hls dash and flv
[03:15:32 CET] <furq> that's exactly what nginx-rtmp does
[03:15:38 CET] <furq> except the dash muxer is apparently broken
[03:15:42 CET] <furq> but you can work around that
[03:15:54 CET] <DHE> any reason not to just standardize on one format?
[03:15:55 CET] <E-TARD> i hate nginx
[03:15:58 CET] <E-TARD> whats next?
[03:16:04 CET] <furq> good luck then
[03:16:13 CET] <E-TARD> what ever
[03:16:29 CET] <E-TARD> JEEB help me
[03:17:43 CET] <derekr> python -m SimpleHTTPServer is another possibility :)
[03:18:49 CET] <E-TARD> so you guys want me to push ffmpeg from my pc to ffmpeg on a server that will out to file then server them with a webserver like apache or something like that?
[03:19:30 CET] <derekr> just run that on your pc?
[03:19:35 CET] <DHE> if it was a simple one-format for dash or hls then that's probably easiest. remote filesystem like cifs or ssh to push content directly onto the webspace
[03:19:36 CET] <E-TARD> no
[03:19:41 CET] <furq> i don't want you to do anything
[03:19:48 CET] <E-TARD> i have a server with a 1gbit
[03:19:53 CET] <furq> i encode on my pc and send it to an rtmp daemon on my server which serves rtmp and hls
[03:20:08 CET] <DHE> that works as well...
[03:20:10 CET] <furq> the rtmp daemon happens to be nginx
[03:20:21 CET] <E-TARD> i want the server to do the re-encode
[03:20:29 CET] <furq> you can do that
[03:20:59 CET] <E-TARD> so i can feed ffmpeg a stream of some kind & have ffmpeg re-encode & out to files
[03:21:01 CET] <E-TARD> ?
[03:21:09 CET] <furq> as long as the stream you send is compatible with rtmp
[03:21:36 CET] <DHE> hairpin it through nginx?
[03:22:16 CET] <E-TARD> something like this ffmpeg -i -f ffm rtmp://server
[03:22:32 CET] <furq> https://github.com/arut/nginx-rtmp-module/wiki/Directives#exec_push
[03:23:10 CET] <E-TARD> like this
[03:23:11 CET] <E-TARD> ffmpeg -f dshow -r 30 -video_size 640x480 -pix_fmt yuv420p -rtbufsize 2100M -i video="Logitech QuickCam Pro 9000":audio="Samson C01U (Samson C01U       " -map 0 -flags:v +global_header -flags:a +global_header -profile:v baseline -level 3.0 -maxrate 1024k -bufsize 2985k -codec:v libx264 -vb 896k -codec:a aac -ar 44100 -ac 2 -ab 128k -bufsize 2985k -tune zerolatency -bufsize 0 rtmp://server
[03:23:50 CET] <furq> if that's the command line you want to run on the client then sure
[03:24:19 CET] <spiderkeys> I'm currently working on a low-latency H264 live streaming solution that uses fragmented mp4 sent to the client browser over socket.io from a c++/node.js app and into the media source extensions, don't know if that is exactly what you want to do, but could helpful (although it is all done programmatically, no command line tools)
[03:24:20 CET] <E-TARD> is rtmp the best way to feed the server? what about others like udp
[03:24:29 CET] <furq> it's an rtmp server
[03:24:50 CET] <E-TARD> no to feed ffmpeg on the server
[03:24:51 CET] <DHE> that's how the nginx module works
[03:25:01 CET] <E-TARD> i'm not using nginx module
[03:25:14 CET] <E-TARD> stay on same page as me
[03:25:22 CET] <E-TARD> lol
[03:25:47 CET] <E-TARD> ffmpeg a stream to ffmpeg on server then out to files
[03:25:59 CET] <spiderkeys> piping h264 straight off of the USB driver, muxing it, and shooting it straight into the browser's media source api
[03:31:27 CET] <TD-Linux> yeah that's a pretty reasonable way to do it, just not off the shelf
[03:32:54 CET] <TD-Linux> you could probably even do webrtc, your webcam probably outputs H.264 baseline
[05:00:52 CET] <pinPoint> does new Grothendieck 2.6.6 bring any speed improvements to encoding times for libx265?
[05:14:04 CET] <acbsspdas> Using FFmpeg from git HEAD; ./configure is failing to find x265(tested both versions 1.7 and 1.8) using pkg-config
[05:14:15 CET] <acbsspdas> https://gist.github.com/anonymous/0b9b1041aedef64fa1b1
[05:22:08 CET] <xintox> howdy
[08:33:39 CET] <debianuser> MachinaeWolf: Those "ALSA buffer xrun" usually mean that the application (ffmpeg) don't read data from the buffer in time. When ffmpeg tells "I need 16-bit 48kHz stereo input" - alsa (actually the soundcard driver) fills a circular buffer with those, 192000 bytes per second no matter what happens. And ffmpeg must read them in time. If it doesn't - buffer overflows and you get overrun.
[08:35:33 CET] <debianuser> MachinaeWolf: So the real question is why ffmpeg doesn't read the data in time. There could be different reasons (video/audio sync, not enough cpu to encode, not enough bandwidth to write/send it, etc) and usually in the `ffmpeg -f alsa -i hw:0,0,0 ...` command line that's not because of the "-f alsa -i hw:0,0,0" part, but because of that "..." part. :)
[08:49:25 CET] <jonascj> Should a DVD hold 4.7*1024^3 bytes? That is 5153960755 bytes. My dvd writer says that 4500MB can  not fit the disk
[08:49:35 CET] <E-TARD> WoW i found this really cool app Open Broadcaster Software https://obsproject.com/index way better then what i was using. i just wanted to share that for anyone who does not know about it. going to bed night night
[10:22:17 CET] <dv_> correct me if I'm wrong. Group-of-Pictures does not really exist in h.264 as part of the format, right?
[10:22:21 CET] <dv_> unlike, say, MPEG4p2.
[11:00:54 CET] <JEEB> dv_: what specifics are you pointing towards?
[11:01:15 CET] <JEEB> I mean, you have the concept of GOPs due to IDR pictures (and with HEVC IRAP pictures)
[11:01:26 CET] <JEEB> also there's intra refresh and such
[11:04:28 CET] <dv_> I mean that gop as bitstream parameter isn't around anymore
[11:05:01 CET] <dv_> from what I know, you just have IDR pictures in h264, and users *can* manually implement GOP by periodically forcing I/IDR frame generation
[11:05:14 CET] <dv_> but it is not explicitely included by the spec
[11:11:36 CET] <JEEB> hmm, so MPEG-4 Part 2 had GOP length as a parameter in the bit stream? how did that work with adaptive random access pictures?
[11:29:18 CET] <dv_> well tbh I don't know. I just dimly remember somebody here telling me that.
[11:30:03 CET] <dv_> I need this info for documentation I am writing. it concerns a hardware encoder which has a "gop-size" parameter for all formats. mpeg2,mpeg4,h263,h264.
[11:30:41 CET] <dv_> I recall discussions here that "GOP" as a concept has problems and that h264 "learned from this"
[11:30:59 CET] <dv_> unfortunately, I lost the logs :/
[12:19:52 CET] <DHE> dv_: in theory you can run a single huge GOP and never send new IDR frames. in practice people want to be able to seek around videos and join midstream so something is used to bound the size
[12:29:32 CET] <BtbN> Certain Live-Streaming setups also never send new I/IDR frames unless requested, because a frame got lost or something
[12:29:45 CET] <BtbN> The WiiU gamepad for example works like that
[12:39:46 CET] <dv_> DHE: yeah, but with GOPs, there is a fixed interval. if you just have I/IDR frames, you can place them anywhere, not necessarily in a fixed interval.
[12:40:00 CET] <dv_> I guess this is what that person meant by "h264 learned from this"
[12:40:59 CET] <dv_> so the remaining question is whether or not older formats like mpeg4p2 and mpeg2 have some GOP-related difference compared to h264
[12:42:48 CET] <DHE> a few. with only 1 reference frame allowed it means any I-frame in MPEG2 and MPEG4-part2 is a GOP start point. also image quality without a new GOP tended to degrade over time. h264 has neither of these limitations and an I-frame isn't necessarily a GOP start position
[12:43:52 CET] <DHE> so for the mostpart IDR frames are arbitrary for the sake of having them. x264 will also insert one if it detects a massive image change and decides that the previous reference frames are no longer valuable ('scenecut' option)
[12:44:40 CET] <dv_> okay, thanks
[12:45:31 CET] <dv_> image quality without a new GOP doesn't degrade with h264 because it allows for intra refresh? or because it can have multiple reference frames?
[12:45:56 CET] <DHE> or the CABAC filters are pretty good
[12:46:33 CET] <dv_> is the in-loop deblocking another possible reason?
[12:47:23 CET] <DHE> probably helps a bit
[12:47:33 CET] <DHE> I don't know the codec in that level of detail to give an answer
[12:47:39 CET] <dv_> ok
[12:50:22 CET] <Mavrik> Of course, IDRs kinda get important as soon as you have randomly accessed streams :)
[12:55:57 CET] <DHE> at which point 'arbitrary amount of time' makes some sense
[13:45:17 CET] <Nivalis> hello i have trouble with record video from my ip camera with ffmpeg and this command http://pastebin.com/SG9Cehcq , then into vlc "video speed" is not constantly, sometimes it's fast
[13:48:05 CET] <Nivalis> how i can fix accelerate moment and have video 1:1, 1 hour in real time = 1 hour video time
[13:58:34 CET] <waressearcher2> Nivalis: hallo, wie geht's ?
[16:49:03 CET] <thomedy> i am trying to encode avc1 and mp4a i hade it at mp4 and aac
[16:49:19 CET] <thomedy> and im kind of in a hurry
[16:57:57 CET] <waressearcher2> thomedy: hallo, wie geh't es dir ?
[20:06:32 CET] <Properrr_> Hello to all. May somebody help me? I want get PSI tables from demuxer of ffmpeg
[20:06:41 CET] <Properrr_> What should i do?! ffmpeg 2.8
[20:19:07 CET] <waressearcher2> Properrr_: hallo, wie geht's es dir ?
[20:24:46 CET] <xintox> anyone know of a free cdn for streaming?
[20:27:02 CET] <waressearcher2> xintox: was ist fur "cdn" ?
[20:27:17 CET] <xintox> waressearcher2: huh?
[20:27:40 CET] <xintox> cdn. like peer5.com or something that will handle m3u8 streams
[21:11:51 CET] <skokkk> hello, I would like to convert thousands of .mp4's, .avi's, and more to .webm's. How would I go about this with minimal quality loss?
[21:19:25 CET] <JEEB> you will probably want to find the least bad -crf value for some clips with your otherwise selected parameters
[21:20:16 CET] <JEEB> also if you are planning to use those files over limited bandwidth, -maxrate and -bufsize are required
[21:24:08 CET] <c_14> Oh, and use vp9. Not vp8
[21:28:08 CET] <skokkk> cp9?
[21:28:11 CET] <skokkk> vp9?*
[21:31:43 CET] <c_14> libvpx-vp9
[21:31:50 CET] <skokkk> ah. thanks
[00:00:00 CET] --- Thu Dec 24 2015