[Ffmpeg-devel-irc] ffmpeg.log.20160322

burek burek021 at gmail.com
Wed Mar 23 02:05:01 CET 2016


[01:00:29 CET] <causasui> i'm using ffmpeg to livestream via twitch.tv. everything works great, except the video is from my secondary desktop not the primary (the secondary is to the left of the primary). can I add an offset to the frame size set via -s, or an offset somewhere else?
[01:46:32 CET] <llogan> causasui: i don't quite understand your question, but showing your command and the complete console output will be helpful
[01:53:42 CET] <jookiyaya> what is faster?  $10,000 computer in 2016 or  fastest supercomputer in year 2000 ?
[01:55:01 CET] <causasui> llogan: all right, please hold
[01:55:26 CET] <mrmenacex> does anyone know why a vp9 video in the regular html5 video tag would work on Android using chrome but not on ios 9.2.1 using chrome ?
[01:55:38 CET] <causasui> llogan: here's the script http://sprunge.us/ZISA # there aren't any errors in the console output, do you still need that?
[01:56:41 CET] <furq> mrmenacex: we might if you post ffmpeg/ffprobe output
[01:57:03 CET] <mrmenacex> what's ffprobe ?
[01:57:24 CET] <mrmenacex> there is no ffmpeg output really it's already encoded
[01:57:47 CET] <furq> does mobile chrome ship with video decoders
[01:57:53 CET] <furq> if it doesn't then it's down to iOS not supporting vp9
[01:58:36 CET] <mrmenacex> furq: it works in mobile chrome on android
[01:58:42 CET] <J_Darnley> Apple wouldn't allow a viable alternative to safari, would they.
[01:58:47 CET] <furq> well yeah but android supports vp9
[01:59:02 CET] <relaxed> causasui: the first -s should be -video_size
[01:59:21 CET] <furq> iOS chrome isn't listed on caniuse so i have no idea if it's supported
[01:59:23 CET] <relaxed> first -r should be -framerate
[01:59:38 CET] <furq> it definitely isn't in desktop or iOS safari though
[01:59:41 CET] <mrmenacex> it doesn't work in safari either
[01:59:45 CET] <furq> ^
[01:59:50 CET] <furq> http://caniuse.com/#feat=webm
[02:00:21 CET] <J_Darnley> Why isn't <video> just a wrapper around libavformat?
[02:00:38 CET] <furq> one would hope that google would bundle a decoder for their own codec, but i wouldn't put it past apple to have prevented them
[02:01:08 CET] <furq> either that or this is their way of forcing h.264 on youtube with iOS chrome so that it doesn't eat all your battery
[02:02:45 CET] <mrmenacex> furq: that sucks i just saw caniuse :(
[02:10:46 CET] <causasui> relaxed: ok I'll swich that. can you help me figure out how to set an offset, though?
[02:36:05 CET] <relaxed> causasui: pastebin.com the command and console output
[02:36:20 CET] <causasui> relaxed: command is just $scriptname $key
[02:37:23 CET] <relaxed> and console output is $nowheretobeseen
[02:38:01 CET] <relaxed> causasui: did you read https://trac.ffmpeg.org/wiki/Capture/Desktop ?
[02:38:40 CET] <furq> causasui: -i :0.0+1920,0
[02:39:09 CET] <furq> replace 1920 with the horizontal resolution of your left monitor
[02:39:45 CET] <llogan> causasui: for future reference we don't provide help for scripts here because sometimes the problem is with the script and not with ffmpeg, so we need unscriped commands and the complete console output. makes things easier to duplicate issues too.
[02:40:25 CET] <furq> that is true as a general rule but i don't think it makes a difference here
[02:40:36 CET] <causasui> relaxed: sorry, I had to trap it. http://sprunge.us/ecEW
[02:40:37 CET] <causasui> furq: that looks like what I wanted. i'll try it, thanks
[02:40:37 CET] <causasui> hmm, error: [x11grab @ 0x556ba38e6380] Capture area 1920x1080 at position 1920.0 outside the screen size 3600x1080
[02:40:37 CET] <causasui> :0.0+1920,0: Invalid argumenty
[02:40:37 CET] <causasui> i think it should be 1680 though?
[02:40:47 CET] <furq> 01:39:09 ( furq) replace 1920 with the horizontal resolution of your left monitor
[02:40:53 CET] <furq> so yeah
[02:40:56 CET] <llogan> also, apparently twitch now recommends aac instead of mp3
[02:41:24 CET] <causasui> furq: yep, 1680 did it, thank you!
[02:41:55 CET] <furq> i've had some issues in the past with browsers playing back mp3 in flv, so i'd recommend aac anyway
[02:42:05 CET] <llogan> http://trac.ffmpeg.org/wiki/EncodingForStreamingSites?action=diff&version=5
[02:42:49 CET] <causasui> llogan: i'll keep that in mind, thank you
[02:43:11 CET] <llogan> are you sure you need a -minrate?
[02:43:40 CET] <furq> https://trac.ffmpeg.org/wiki/StreamingGuide
[02:43:45 CET] <llogan> and that "-strict normal" looks superfluous.
[02:43:50 CET] <furq> should this page really be recommending ffserver
[02:44:09 CET] <llogan> that page should eb completely rewritten. it's a mess. mostly written by one guy.
[02:44:21 CET] <furq> yeah that whole second paragraph needs changing
[02:44:24 CET] <furq> at least
[02:45:02 CET] <llogan> too verbose. not enough structure or formatting. "typing as I think/notes to self" style.
[02:45:29 CET] <furq> does -vpre even exist any more
[02:45:42 CET] <llogan> IIRC, perhaps for custom presets.
[02:46:02 CET] <furq> "Here is what another person once did:"
[02:46:05 CET] <relaxed> it has -r 15 before the input :/
[02:46:06 CET] <furq> yeah this isn't good is it
[02:46:36 CET] <llogan> furq: that sentence always bothered me but i don't have the energy to dive into that rabbithole
[02:46:59 CET] <furq> if i streamed more often i might have a go at it
[02:47:19 CET] <llogan> please do
[02:47:43 CET] Action: relaxed votes furq should stream more
[02:47:47 CET] <furq> i have no idea about constraining bandwidth or anything other than rtmp though
[02:51:50 CET] <papamoose_> Anyone know how to also copy album art when copying streams to a different container? Right now I have this(http://pastebin.com/raw/PtgtWZsE) which works well, except it does not have the album artwork.
[02:52:15 CET] <llogan> causasui: "threads="2" # max 6" you may want to allow libx264 to automatically choose threads number. not sure why it says "max 6".
[02:55:34 CET] <c_14> papamoose_: ID3 tags? then not that I'm aware of
[02:55:57 CET] <c_14> ffmpeg converts the album artwork into a video stream internally
[02:56:48 CET] <papamoose_> fflogger: http://pastie.org/private/wie3muclridd2sn6dcgskg
[02:57:02 CET] <c_14> That's not the command output.
[03:00:03 CET] <papamoose_> c_14: i have tried -vcodec copy to copy the album artwork except that it returns an error when I add that option.
[03:00:53 CET] <c_14> what error?
[03:05:04 CET] <papamoose_> c_14: http://pastie.org/private/n9kvstqhrgh3mmj8swgow
[03:06:45 CET] <c_14> weird, and if you don't specify -vcodec copy?
[03:06:45 CET] <papamoose_> It seems to be on line 345: http://pastie.org/private/n9kvstqhrgh3mmj8swgow#345
[03:07:44 CET] <J_Darnley> ffmpeg is too clever for its own good.  No files have "embedded covers" but some files do have a 1 frame video stream.
[03:07:58 CET] <c_14> eh, it might be the j
[03:08:17 CET] <c_14> The pix formats with j are deprecated, it might just not be in the list
[03:08:35 CET] <c_14> The parser/demuxer should probably set range though...
[03:08:44 CET] <papamoose_> http://pastie.org/private/9tn68djrstdbh2daupg#1
[03:09:06 CET] <c_14> papamoose_: get rid of the -vn
[03:09:15 CET] <c_14> try letting ffmpeg pick the codec
[03:09:44 CET] <relaxed> papamoose_: -c:v libx264 -pix_fmt yuv420p
[03:11:29 CET] <papamoose_> c_14: getting rid of -vn did not seem to work. trying relaxed's method.
[03:11:30 CET] <relaxed> but that will create a video stream
[03:12:54 CET] <papamoose_> relaxed: hmm.. that is not ideal.
[03:14:02 CET] <relaxed> papamoose_: right, use -vn and then "mp4box -itags cover=cover.jpg input.m4a"
[03:16:06 CET] <papamoose_> mp4box is a different utility?
[03:16:25 CET] <relaxed> yes, it's from gpac
[03:18:25 CET] <papamoose_> relaxed: hmm.. so I know I can grab the cover like this: ffmpeg -activation_bytes ${ACTIVATION_BYTES} -i ${m} -an -vcodec copy cover.jpg
[03:24:32 CET] <papamoose_> relaxed: i have modified my script to include mp4box to add the cover after the conversion and it looks like it is working now.
[03:24:52 CET] <papamoose_> Any one know why ffmpeg cannot do this by default?
[03:25:02 CET] <J_Darnley> What did I say above?
[03:25:05 CET] <J_Darnley> ffmpeg is too clever for its own good.  No files have "embedded covers" but some files do have a 1 frame video stream.
[03:26:24 CET] <papamoose_> J_Darnely: sorry I didn't see your response. Would you mind clarifying: "ffmpeg is too clever for its own good"?
[03:26:39 CET] <J_Darnley> Read the second sentence in that line.
[03:26:47 CET] <TD-Linux> really I should probably file this as a bug
[03:26:50 CET] <TD-Linux> someone hits this once a week
[03:27:49 CET] <papamoose_> J_Darnley: I understand that, but I think I'm still missing something. If there is a 1 frame video stream, then -vcodec copy should do what I need..?
[03:28:11 CET] <J_Darnley> You need a way to tell ffmpeg that it isn't a video stream.
[03:28:27 CET] <papamoose_> Is there such a way?
[03:28:32 CET] <J_Darnley> No!!
[03:28:39 CET] <J_Darnley> That's half the problem!
[03:28:54 CET] <papamoose_> ahh. I see. :(
[03:29:00 CET] <J_Darnley> The other half is treating a binary data attachment as a stream.
[03:29:59 CET] <J_Darnley> Another half is the concept of embedded covers in the first place
[03:30:11 CET] <papamoose_> J_Darnley: Thanks for explaining that. :)
[03:30:13 CET] <J_Darnley> 150% problem.
[03:32:48 CET] <papamoose_> Well, thanks all for helping me solve this for now.
[05:25:24 CET] <seeit> hello, does anyone know how to speed up the conversion of a .mkv file on arm? right now its is about 0.5 which is too slow for streaming
[05:34:21 CET] <c_14> -preset ultrafast
[07:11:13 CET] <Mista_D> Hi, passing "filter_complex" to FFmpeg from a tmp.txt; can't seem to read command. [ "-filter_complex foo,bar" ] fails, while [ -filter_complex "foo,bar" ] doesn't. How would I store and pass the filter chain via variable in bash avoiding the quotations please?
[07:11:54 CET] <Mista_D>  [ -filter_complex "foo,bar" ] doesn't fail ( :
[08:48:26 CET] <pezet91> hello, I have encoded H264 stream from video grabber, but I can't restream this with avconv/ffmpeg. Maybe I can't set parameters to avconv/ffmpeg. Anyone can help me with this?
[13:04:59 CET] <Gringham> Hey :) Do you know a way, to pass a list of files to stream via ffmpeg, so that directly after one end the next one starts (without resetting the Timestamp) ?
[13:08:49 CET] <J_Darnley> Use one of the three concat options
[13:09:16 CET] <Gringham> I'll look after it, ty :)
[15:37:44 CET] <Gringham> Is there a way to concatenate via the concat demuxer and reading the files not from a text file but from a List ?
[15:38:10 CET] <J_Darnley> What is a "List" and how can ffmpeg read it?
[15:39:25 CET] <J_Darnley> Ultimately I will say the answer is no.
[15:39:32 CET] <J_Darnley> Write this List to a file
[15:54:04 CET] <Gringham> ffmpeg -f concat -i C:\Users\name\Desktop\int.txt -f mpegts udp://239.255.1.0:1239
[15:54:19 CET] <Gringham> this is what I tried to do now
[15:54:33 CET] <Gringham> the first video is playing, than it mutters about the format
[15:54:52 CET] <furq> what format are the videos
[15:56:16 CET] <Gringham> ts
[15:56:50 CET] <Gringham> [mpegts @ 0000002f72d30540] Non-monotonous DTS in output stream 0:0; previous: 2707320, current: 12848;
[16:08:50 CET] <leogps> I am using ffmpeg to re encode rtmp stream into mpegts segmented files on Ubuntu 13. The video gets stuck for every 5 secs with a glitchy sound in the audio.
[16:18:58 CET] <leogps> http://pastebin.com/SJNxbev7
[16:22:06 CET] <J_Darnley> Good to see people have no clue what they are doing.
[16:28:18 CET] <satinder___> Hi , any one there
[16:29:05 CET] <satinder___> if we want update over textfile for drawtext (overlay purpose) then how we can rewrite it
[16:29:21 CET] <satinder___> I using echo in system command
[16:29:42 CET] <satinder___> but I am not sure that is good or not but that is working
[16:30:28 CET] <satinder___> but i listen system command is not good , so please anyone give any opinion , what I can do for that
[16:30:31 CET] <satinder___> thanls
[16:30:36 CET] <satinder___> thanks
[16:35:24 CET] <petecout_> Is there a metadata value in ffprobe that shows a file is currently being encoded? I have an application that needs to monitor mpegts files as they are being generated to see if they are done.
[16:36:48 CET] <kepstin> petecout_: no obvious way with mpeg-ts. You'll want to just stat the file at interval and see if the mtime keeps updating or file size keeps changing.
[16:41:21 CET] <petecout_> kepstin: Thanks mate. I figured if there wasn't a flag I'd go by file count to see which one the system was currently one.
[16:41:27 CET] <petecout_> on
[16:41:37 CET] <petecout_> I move them to a different folder when they are done.
[16:46:09 CET] <Fyr> JEEB, hey, is there news on the delay bug?
[17:25:22 CET] <brick> ffmpeg -i try_%02d.png -vcodec mpeg4 try.avi    produces a movie one frame long when there are dozens of images there starting with try_01.png. what am i doing wrong?
[17:36:54 CET] <Gringham> when I concatenate two files like this :        "concat:file1.ts|file2.ts"    file2.ts has no audio anymore (file1 never had one) , is there a way to prevent this ?
[17:37:31 CET] <Gringham> when I run "concat:file2.ts|file1.ts" it still play the audio
[17:37:52 CET] <Gringham> plays*
[17:40:22 CET] <brick> Gringham, when you do a "mix" i think you get to specify which streams to use in the output
[18:15:08 CET] <explodes_> I'm trying to update my library by upgrading FFmpeg from 2.8.6 to 3.0. There are a few deprecations, and the app that uses the library now crashes. - Is it worth the upgrade process to use 3.0?
[18:18:18 CET] <thebombzen> brick: try_02d.png needs there to be try_01.png, try_02.png, etc.
[18:23:08 CET] <brick> thebombzen, OK, that might have been why. when i try this again i will make sure that there are no missing numbers.
[18:23:09 CET] <brick> thanks!Q
[20:44:19 CET] <petecout_> When your encoding mpegts segements for a HLS stream. The last segment is always 'inprogress' as it's encoded to. Is it best practice to prevent that working segment from being downloaded by the client or can HLS clients handle partial segemetns?
[20:53:08 CET] <pyro25> hey guys. I started playing with ffserver and I'm realizing how cool it would be to have a database-backed configuration. anyone has ever heard of this?
[20:55:12 CET] <blue_misfit> hey, so f4v is basically mp4, but is flv also comparable to mp4 or is it really its own thing?
[21:17:51 CET] <kepstin> blue_misfit: flv is basically a serialization of a flash command stream, so quite a bit different
[21:18:28 CET] <kepstin> (although there might be some cases where an flv stream contains an mp4 or mp4-like stream? I'm not sure entirely how that works)
[21:28:30 CET] <awefoijawef> hi
[21:28:57 CET] <Azrael_-> hi
[21:34:02 CET] <Fyr> hi
[21:36:01 CET] <petecout_> Ello
[21:44:04 CET] <petecout_> Has anyone ever run into this issue before an error happened: ffmpeg exited with code 1: public/sdps/video-audio3.sdp: Invalid data found when processing input
[21:44:20 CET] <petecout_> It seems to require a system restart when it occurs the first time.
[21:48:06 CET] <blue_misfit> kepstin, thanks :)
[22:05:43 CET] <Mista_D> what'
[22:06:06 CET] <Mista_D> s the recommended gcc version please?
[22:19:32 CET] <llogan> Mista_D: there is not a specifc recommended version
[22:20:33 CET] <Mista_D> Understood, will play with it...
[22:20:39 CET] <Mista_D> Thanks
[22:27:48 CET] <llogan> Mista_D: however, there may be some versions that could recommended to *not* use. https://trac.ffmpeg.org/ticket/3970
[22:28:14 CET] <llogan> i'm unsure how relevant that ticket still is
[22:29:13 CET] <llogan> also #1464
[22:32:53 CET] <Mista_D> allright, do not use gcc 4.2.  got 4.4.7 with Centos...
[22:40:43 CET] <squarecircle> heyo
[22:41:37 CET] <squarecircle> is there a way to dismantle a video in singular frames and write them to stdout?
[22:43:52 CET] <squarecircle> at the moment I'm calling ffmpeg very often to get frames at with increasing timestamps, but I would prefer to actually create a stream and parse this stream
[23:01:47 CET] <petecout_> Before I test this. Is it ok to use the same uri for input as output?
[23:02:45 CET] <Mavrik> squarecircle, dump raw frames to stdout?
[23:02:48 CET] <Mavrik> there's raw output for that
[23:07:19 CET] <kepstin> petecout_: depends entirely on what kind of uri it is.
[23:07:39 CET] <kepstin> petecout_: you'll have bad times if you use the same file:/// uri for input and output, for example :)
[23:07:49 CET] <kepstin> and with e.g. http, rtmp it depends on the server, not ffmpeg.
[23:13:18 CET] <squarecircle> Mavrik: sounds good, do you have a hook for me, where to look?
[23:15:41 CET] <Mavrik> ffmpeg -i <input> -codec:v rawvideo -f rawvideo -
[23:15:44 CET] <Mavrik> should work I guess
[23:15:55 CET] <Mavrik> Note that rawvideo will be whatever the frames in video are, usually YUV420p
[23:16:36 CET] <livingBEEF> Are frames counted from 1 or from 0?
[23:17:31 CET] <petecout_> kepstin: thnx mate. It's just a local file in path temp/vid.ts temp/vid.ts
[23:17:34 CET] <petecout_> It seems to be working ok
[23:18:43 CET] <kepstin> petecout_: hmm, i wouldn't expect that to work; when writing the output it should be truncating the file before writing...
[23:18:51 CET] <kepstin> maybe it's removing the file and creating a new one instead
[23:18:55 CET] <squarecircle> Mavrik: do you know how they are separated?
[23:19:03 CET] <Azrael_-> i'm streaming with ffmpeg a live-stream to an nginx-rtmp-server but the video occasionally just stops for a few seconds and then proceeds. the upload+download isn't overloaded, cpu also free. any suggestions how i could debug this?
[23:19:09 CET] <Mavrik> squarecircle, they're not afaik
[23:19:21 CET] <Mavrik> squarecircle, you need to know the size of the video :)
[23:19:36 CET] <kepstin> livingBEEF: I think in most places in ffmpeg frames are counted starting at 0.
[23:19:54 CET] <squarecircle> so I try to ffprobe the resolution and then read in the size of each frame
[23:19:55 CET] <squarecircle> ?
[23:20:02 CET] <livingBEEF> kepstin: thanks
[23:20:25 CET] <Mavrik> squarecircle, yes, and then calculate size in bytes for each frame
[23:20:35 CET] <Mavrik> and that's what you're gonna get in sequence
[23:22:36 CET] <kepstin> you have to have some knowledge of the pixel format being used to calculate the size correctly. (if you want to make it easy on yourself, you can use the -pix_fmt option to tell ffmpeg to convert everything to one format)
[23:23:57 CET] <petecout_> kepstin: Could be. I'm going to test and I'll let you know if there are any issues.
[23:27:56 CET] <squarecircle> Mavrik: that sounds a bit interesting
[23:30:49 CET] <petecout_> kepstin: You were right lol. Can't process same file on file system.
[23:40:10 CET] <squarecircle> Mavrik: Wikipedia tells me that yuv420 is "YUV420p   6 bytes per 4 pixels, reordered"
[23:41:01 CET] <squarecircle> Mavrik: the rest should be easy goung
[23:42:05 CET] <squarecircle> Mavrik: "-pix_fmt rgb24" should be working, or?
[23:42:32 CET] <Mavrik> if you want to convert it yes :)
[23:43:06 CET] <Mavrik> rgba is also common
[23:43:47 CET] <squarecircle> Mavrik: I want to do some magic with python
[23:43:48 CET] <squarecircle> ^^
[23:45:38 CET] <squarecircle> Mavrik: so it would be "ffmpeg -i testvid.mkv -codec:v rawvideo -pix_fmt rgb24 -f rawvideo -"
[23:46:56 CET] <squarecircle> Mavrik: nah: "ffmpeg -i testvid.mkv -codec:v rawvideo -f rgb24 -" ?
[00:00:00 CET] --- Wed Mar 23 2016


More information about the Ffmpeg-devel-irc mailing list