[Ffmpeg-devel-irc] ffmpeg.log.20150524

[burek]

[01:43:02 CEST] <joer1> Quick question incase anyone can help - I'm trying to publish an rtmp stream using ffmpeg. It seems to publish okay. Sets Stream #0:0: Video: h264, Stream #0:1: Audio: aac . However, running ffprobe on the published stream gives me Stream #0:0: Audio: aac  Stream #0:1: Data: none. Any ideas why?
[02:36:24 CEST] <v2px> Ohai. I'm trying to record my screen with x11grab on a second xsession using "-f x11grab -i :1.0" which works, but only records part of the screen. I can't seem to specify the size of the screen to capture with "-video_size 1680x1050" which yields "Capture area 1680x1050 at position 0.0 outside the screen size 1680x1050". Any ideas?
[03:06:01 CEST] <yumbox> v2px: try one pixel less
[03:06:22 CEST] <yumbox> at the moment fullscreen x11grab bugs out if you specify the full screen size
[03:07:14 CEST] <yumbox> v2px: see https://trac.ffmpeg.org/ticket/4574
[03:07:17 CEST] <v2px> ok cool, that works.
[03:08:05 CEST] <v2px> I have some problems recording fullscreen games though. I read somewhere that fglrx could be the culprit here, anything I could try?
[03:08:40 CEST] <yumbox> recording games is a lot more difficult, can't help you there.
[03:09:13 CEST] <v2px> it basically records mu desktop and all windows fine, including games.. unless they are in fullscreen
[03:09:23 CEST] <v2px> thanks for the help!
[03:10:31 CEST] <yumbox> well, as a workaround just play in maximized windowed if you want to record a game.
[03:10:53 CEST] <v2px> :p
[09:18:02 CEST] <ruby_on_tails> i am doing some screencasts from a mbp retina which has a huge resolution, i have no clue about the target resolution the screencast will be played on, whats the best strategy in this case? should i do a fullscreen screencast and reduce its dimensions or should i reduce my resolution to 1080p and then record the screencast?
[10:19:02 CEST] <jookiyaya> if i encode a same source video twice with exact same settings, would 2 video files be exactly same  bit by bit
[10:36:09 CEST] <maddogie> hello all
[10:37:35 CEST] <maddogie> was there something changed with the x and y-offset in the last version?
[10:42:39 CEST] <JEEBsv> jookiyaya: depends on the encoder and quite a few other things. also if you have stuff like containers and have lavf actually add real timestamps even that can change the file :P
[10:43:18 CEST] <JEEBsv> for example x264 in general should be deterministic by default, but VBV (maxrate+bufsize) and multiple threads is not deterministic for obvious reasons
[10:48:16 CEST] <maddogie> im asking because a script that i use for screenrecording from my second monitor now gives me an "outside the screen size" error. but it is absolutely not out of the screensize
[10:52:42 CEST] <maddogie> monitor 1 =1920x1 080 Monitor 2 1680x1050 . That makes 3000x 1080(1050)  overall. i use an x-offset of 1920 and try to record monitor 2 with a size of 1680x1050.
[10:55:40 CEST] <maddogie> if i use an x-offset of 1919 it works without any problems but them i  have a single column form monitor 1 in the recording
[11:13:57 CEST] <relaxed> maddogie: I belive this was fixed in git master
[11:14:02 CEST] <relaxed> believe*
[11:14:50 CEST] <maddogie> ok i hope
[11:15:28 CEST] <maddogie> is there a bug report? i couldn't find any
[11:17:38 CEST] <relaxed> I don't know. You can test with my build if you'd like, http://johnvansickle.com/ffmpeg/
[11:20:40 CEST] <maddogie> ok
[11:20:53 CEST] <maddogie> i will try your build
[11:21:02 CEST] <maddogie> thank you
[11:28:51 CEST] <ghartz> hi
[11:32:01 CEST] <JFQ> hello :)
[11:33:04 CEST] <JFQ> when I segment a file, encode each part and merge the encoded parts to one big file (aac / h264), I can hear "bumps" on the merge point
[11:33:36 CEST] <JFQ> I found out that there is "priming" and "remainder" sample (https://developer.apple.com/library/mac/documentation/QuickTime/QTFF/QTFFAppenG/QTFFAppenG.html)
[11:34:12 CEST] <JFQ> I guess that's what it's causing the bumps I hear but I can't find how to avoid this
[11:36:17 CEST] <BtbN> Why segment it when you just want to merge it again later?
[11:40:25 CEST] <JFQ> BtbN, distributed encoding
[11:41:04 CEST] <BtbN> Propably not worth it, encoding the whole thing can't take THAT long
[11:41:25 CEST] <JFQ> BtbN, In my case it will be really helpful
[11:41:35 CEST] <JFQ> can't be too deep
[11:41:43 CEST] <JFQ> but I need to segment :)
[11:42:00 CEST] <JFQ> I'm only in trouble with audio
[11:42:02 CEST] <JFQ> video is fine
[11:42:29 CEST] <JFQ> if I encode  audio and video separately (video in segment, audio in one row), the audio is out of sync
[11:43:11 CEST] <JFQ> and if I encode video + audio and merge them, there is this little bump
[11:46:54 CEST] <JFQ> BtbN, no other hints ? :)
[12:09:53 CEST] <maddogie> relaxed the git version of your build works. so i think in the next release my problem will disapper. thank you again
[12:17:40 CEST] <relaxed> maddogie: you're welcome
[15:39:17 CEST] <animax> a few questions. http://www.pasteall.org/58610. is there a difference between A) line 1 and line 3, B) line 5 and line 7? or is it sufficient to use the short forms of each one in each case? and which examples of usage are there for commands like the one in line 9? if I want to render an animation in mp4 format to use it for the web (website header). do I need to use a command like his?
[15:39:59 CEST] <animax> pardon, the smiley properly should be a capital B with a bracket ...
[16:03:25 CEST] <ChocolateArmpits> animax: the pix_fmt yuv420p will be required if you use filters and there's no format filter used, as filter output will be the default yuv422p. You can upload this to youtube, however crf will use more bitrate and some players won't be able to play the file
[16:04:44 CEST] <ChocolateArmpits> for example default configuration of WMP 12 (latest) with no additional codecs can only play h264 at main profile with yuv420p color subsampling
[16:05:08 CEST] <ChocolateArmpits> You might have similar limitations with mobile devices
[16:07:56 CEST] <JEEBsv> ok, let's begin with a) what filters output depends on the filter and the parameters for it b) MS's MF components can play high profile JustFine
[16:16:06 CEST] <animax> ChocolateArmpits: what would be an example for a format filter? an expression like in http://www.pasteall.org/58611 would make no sense?
[16:16:32 CEST] <ChocolateArmpits> animax:  literally a filter -vf format=yuv420p
[16:17:15 CEST] <Guest79271> Hi there. I need to stream .ts (H264 + AAC) files to stb that supports only raw/raw/usp transport. So, can I just use something like: avformat_alloc_output_context2(&m2ts, NULL, "mpegts", url) then read video frame and write it: av_write_frame(m2ts, &packet)? Or I need to use avcodec_encode_video2? Thank you.
[16:17:33 CEST] <Guest79271> raw/raw/udp*
[16:20:48 CEST] <Mavrik> uh
[16:20:55 CEST] <Mavrik> you kinda mixed stuff up a bit
[16:21:03 CEST] <Mavrik> Guest79271, is your video encoded?
[16:21:13 CEST] <Mavrik> and you want to stream what exactly? which container? do you need to transcode?
[16:22:23 CEST] <animax> ChocolateArmpits: so a command like -pix_fmt yuv420p always has to be used together with something like -vf format=yuv420p? I've seen expressions like in http://www.pasteall.org/58612. there is no filter.
[16:22:41 CEST] <ChocolateArmpits> animax: No, that would be the wrong approach
[16:24:05 CEST] <ChocolateArmpits> Either one or the other has to be used, but only if A) you are using filters B) you input is something other than yuv420p and the output video format supports more color pixel formats than yuv420p
[16:25:10 CEST] <ChocolateArmpits> Because png is rgb formatted, you should define what pixel format you want the video stream to be
[16:25:11 CEST] <Mavrik> animax, -c is ambigous, since ffmpeg doesn't really know if you mean codec for audio or for video track, so "c:v" is greatly preferred
[16:25:28 CEST] <Mavrik> also note that positioning of parameters matters :)
[16:25:46 CEST] <Mavrik> so keep encoding paramters after "-i" so you won't get funny behaviour
[16:27:23 CEST] <animax> Mavrik: Thanks.
[16:27:36 CEST] <Mavrik> An yeah, as ChocolateArmpits explained.
[16:27:52 CEST] <Guest79271> Mavrik: Yes, i know, i'm totally new to this stuff but i need to do this. Codec H264 (part 10), decoded format: 4:2:0 YUV. I need to use something that can play stb. mpeg transport stream definitely works on it.
[16:27:56 CEST] <Mavrik> For H.264, only YUV420p is required for players to support and by default ffmpeg will try to keep the same pixel format
[16:28:31 CEST] <Mavrik> So "-pix_fmt" will tell ffmpeg to always make sure that output is in YUV420p no matter the input (and thus ensuring compatibility)
[16:28:52 CEST] <Mavrik> Guest79271, uhm, you haven't really cleared up the situation
[16:29:03 CEST] <Mavrik> Do you have an already encoded stream? In which container?
[16:30:17 CEST] <Mavrik> animax, as for "-vf format" vs. "-pix_fmt"... "-pix_fmt" will insert that format filter implicitly, so you use the first version if you care about when the conversion in video filter chain happens. If you don't they're equivalent.
[16:36:54 CEST] <Guest79271> Mavrik: mpeg-ts
[16:37:30 CEST] <Mavrik> you can stream MPEG-TS directly to UDP socket without ffmpeg
[16:37:57 CEST] <Mavrik> 188B packets with 0x47 at the start, stuff them into 1500 UDPs and send
[16:38:42 CEST] <Guest79271> like all file with video and audio streams?
[16:39:29 CEST] <BtbN> mpeg-ts is especialy good for streaming, as it consists of a stream of very small packages
[16:39:37 CEST] <BtbN> so it can be cut into chunks very easily
[16:48:39 CEST] <animax> first of all thank you. have to think about it. afk
[16:57:14 CEST] <Guest79271> So i just need to cut every video/audio frame into 188B packets and thats all? I was working with mpeg4 H264 rtp stream and there some troubles with NAL units and sdp description. mpeg-ts seems to be more simple. Again sorry for my stupid questions.
[16:57:14 CEST] <Mavrik> Guest79271, MPEG-TS is built for streaming
[16:57:36 CEST] <Mavrik> Guest79271, so it's made to be just sent over wire/air/satellite
[16:57:54 CEST] <Mavrik> It's segmented into 188B packets with 0x47 start code
[16:58:32 CEST] <Mavrik> so as long as you send full 188B packets and don't cut them up (usually sent in 1316B batches since MTU is usually 1500) it works just fine
[16:59:08 CEST] <Mavrik> the data that's usually in SDP is here contained in special TS packets that are repeated over the stream :)
[16:59:16 CEST] <Mavrik> since the format was built to be broadcast over cable/air/tv
[17:06:32 CEST] <Guest79271> The last question. I just need to search file till 0x47 byte get next 188 repeat it 7 times then send 1316B packet and repeat again? Without any things with NAL units, timestamps and a\v sync?
[17:51:37 CEST] <Mavrik> Guest79271, well, you should probably send data at a rate the video is actually played
[17:51:55 CEST] <Mavrik> why not try using ffmpeg process?
[17:52:13 CEST] <Mavrik> ffmpeg -re -i <yourts> -codec:copy udp://<ipofstb>
[18:12:03 CEST] <Guest79271> Mavrik: yes, i will calculate delay. Because of custom firmware on stb i need to write own server. Thank you. You helped me a lot and save a lot of my time.
[18:12:21 CEST] <Mavrik> Guest79271, anyway, if you need exact timing, MPEG-TS packets have PCR
[18:12:29 CEST] <Mavrik> which is meant for sync and you can use that
[18:12:55 CEST] <Mavrik> and I strongly suggest you look at at least basic MPEG2-TS structure when doing something like that, it'll help you fix issues
[18:31:37 CEST] <capradmar> Hi! I'm unable to read a video file with ffmpeg. I have the following error message: Format avi detected only with low score of 1
[18:33:50 CEST] <capradmar> here is the full command and output: http://pastebin.com/0Ar00w23
[19:46:54 CEST] <jookiyaya> what fps does bluray video use
[19:58:41 CEST] <[-T-]> hi
[19:59:03 CEST] <[-T-]> have someone been using h264_qsv encoding ?
[20:01:11 CEST] <intracube> jookiyaya: http://en.wikipedia.org/wiki/Blu-ray_Disc#Video
[20:09:30 CEST] <BtbN> [-T-], ffmpeg has support for that?
[20:12:11 CEST] <capradmar> Hi! I'm unable to read a video file with ffmpeg. I have the following error message:
[20:12:24 CEST] <capradmar> http://pastebin.com/0Ar00w23
[20:17:41 CEST] <[-T-]> btbN yes
[20:17:50 CEST] <[-T-]> in the latest git
[20:18:02 CEST] <[-T-]> you have to download mediasdk from intel
[20:18:10 CEST] <[-T-]> and in conjonction with libmfx...
[20:22:33 CEST] <BtbN> 500$ for the SDK... nope
[20:23:17 CEST] <[-T-]> i am using the trial under linux
[20:23:22 CEST] <[-T-]> not sure if it ever expires ?
[20:23:38 CEST] <BtbN> 30 days
[20:24:18 CEST] <[-T-]> you mean the lib will stop working ?
[20:24:30 CEST] <[-T-]> under linux ?
[20:24:34 CEST] <BtbN> no idea
[20:25:13 CEST] <[-T-]> i had nvenc working in tvheadend btw
[20:26:35 CEST] <BtbN> It doesn't look like it's realy supported on linux, except for on some Xeon CPUs
[20:27:18 CEST] <[-T-]> on core i7
[20:27:21 CEST] <[-T-]> it's supposed to be
[20:27:27 CEST] <[-T-]> but on my side it hangs the gpu
[20:27:56 CEST] <[-T-]> i get this : https://bugs.freedesktop.org/show_bug.cgi?id=69330
[20:30:25 CEST] <BtbN> Wait, their QSV on linux sits on top of libva?
[20:30:50 CEST] <[-T-]> libmfxhw64.so
[20:30:53 CEST] <[-T-]> used by libmfx.so
[20:30:56 CEST] <[-T-]> used by ffmpeg
[20:31:06 CEST] <[-T-]> yes on top of libva
[20:31:48 CEST] <BtbN> what the hell...
[20:31:58 CEST] <BtbN> I'd have expected it talk to the hardware directly
[20:32:25 CEST] <[-T-]> libva.so is included in the package
[21:20:07 CEST] <t4nk810> Hi everyone! I have got some MP4 files (1280 x something, 25 fps) and some MOV files (full HD, 29.97 fps) and Blender handles the latter quite nicely which isn't the case for the MP4s. I would like to convert the MP4 files to the exact same format (MOV, video and audio codec, size, framerate) that the MOVs have. Does anyone know a ffmpeg convenience function for that? Thanks a lot!! :)
[21:22:27 CEST] <BtbN> Isn't mov and mp4 basicaly the same container?
[21:22:43 CEST] <JEEBsv> isobmff is based on mov, yes
[21:23:15 CEST] <BtbN> that framerate conversion most likely won't end up too great though
[21:24:29 CEST] <t4nk810> Is there any loss to except for the 25fps video simply being lower than the target 29.97fps?
[21:25:04 CEST] <BtbN> You can't pull new frames out of thin air
[21:25:17 CEST] <BtbN> And as it's not a multiple, you can't double the frames
[21:25:34 CEST] <BtbN> so it will end up with some weird thing in between, where it duplicated some frames
[21:26:00 CEST] <BtbN> I'd rather try to find out what blender doesn't like about the file, and fix that instead. In blender.
[21:27:50 CEST] <t4nk810> Well, Blender imports the file nicely and after applying a time-stretch effect to the video (not audio) the rendered output is also in sync. The issue is in the live preview where my machine is not fast enough to handle both live frame rate adaption and scaling
[21:28:21 CEST] <t4nk810> I'll try scaling up and using ffmpeg's -r option, maybe it will help
[21:29:49 CEST] <t4nk810> ffmpeg -i in.mp4 -r 29.97 -vf scale=1920:-1  out.mov   worked perfectly :) :)
[21:29:58 CEST] <t4nk810> thanks for the advice, guys!!
[22:57:12 CEST] <cbounce> hey, I'm having a wierd problem where ffmpeg is blindly overwriting an output file even when I explicitly set -n as a flag, I am guessing it has something to do with wierd characters in the file/folder, but can't narrow it down to a certain character (yet)
[22:57:26 CEST] <cbounce> is that a known problem?
[22:57:48 CEST] <cbounce> trying to find a minimal working example
[22:58:00 CEST] <cbounce> (or failing, for that matter)
[22:58:39 CEST] <c_14> Can you give any example?
[22:59:30 CEST] <cbounce> c_14: filename in question is: "./Rivo Drei/Yeah! Jawohl! Und Überhaupt/02 Was Ist Los.mp3'"
[22:59:44 CEST] <cbounce> let me try to piece together a more minimal example
[22:59:53 CEST] <c_14> might be the exclamation marks
[23:01:45 CEST] <intracube_afk> cbounce: are you quoting the filename? try single quotes if you're running on bash shell
[23:02:01 CEST] <cbounce> yeah, but a hasty test with random exclamation marks in a dummy file did not reproduce the issue
[23:02:19 CEST] <cbounce> intracube_afk: yes I am running in bash, and yes I am quoting correctly
[23:02:42 CEST] <cbounce> (at least I think I am)
[23:03:46 CEST] <intracube_afk> single quotes preserves literal value of the characters. there's some kind of interpretation with doublequotes IIRC
[23:05:35 CEST] <cbounce> intracube_afk: currently I am using double quotes, and would prefer to keep doing so - ffmpeg gets called inside a single-quoted string in the actual "production pipeline"
[23:06:00 CEST] <cbounce> at least as far as I can tell, there is no bash substitution going on - the new name has all the exclamation marks and Umlauts
[23:08:55 CEST] <intracube_afk> maybe post the full command line in #bash channel?
[23:11:07 CEST] <cbounce> intracube: yeah, let me try to figure out what is actually causing the problem - even if it IS a bash problem, I find it wierd that it can cause ffmpeg to clobber a file
[23:11:51 CEST] <c_14> If you can pastebin a full commandline (and the output) where ffmpeg clobbers a file, I could look at it.
[23:11:53 CEST] <intracube> maybe it's nullifying a load of the ffmpeg params, including -n
[23:12:42 CEST] <c_14> Maybe even with -loglevel debug
[23:13:09 CEST] <cbounce> c_14: give me a sec, have to strip a bunch of file/folder names preceding the actual path I pasted earlier and verify it's still broken
[23:13:27 CEST] <cbounce> don't want to paste my whole internal hierachy
[23:30:36 CEST] <cbounce> c_14: it's the umlaut: minimal working example: http://pastie.org/pastes/10205403/text?key=qgrgn2myrzbtxhii8qk1tq
[23:30:49 CEST] <cbounce> I set the -t 1 just to limit the amount of output
[23:31:09 CEST] <cbounce> result is the same - existing file get's clobbered
[23:31:16 CEST] <cbounce> gets*
[23:33:48 CEST] <intracube> cbounce: did single quotes work any differently? (just curious)
[23:34:17 CEST] <c_14> intracube: it's windows, single quotes won't work
[23:34:23 CEST] <c_14> cbounce: looks like a windows problem to me
[23:34:26 CEST] <c_14> (works under linux)
[23:34:48 CEST] <intracube> yeah, I'm confused. so not a bash issue after all :)
[23:35:05 CEST] <cbounce> intracube: no, same error without quotes or single quotes
[23:35:39 CEST] <cbounce> c_14: did you copy my umlaut from the dumb file? might be some wierd encoding
[23:35:40 CEST] <intracube> I thought windows used backslashes for paths?
[23:36:02 CEST] <cbounce> intracube: cygwin shell, ffmpeg win binary
[23:36:40 CEST] <jcay> cbounce surogat umlaut.. (theoretically)
[23:36:48 CEST] <intracube> ah ok
[23:37:45 CEST] <c_14> cbounce: copied your entire command (bar the exe, and I changed m4a to opus)
[23:38:33 CEST] <cbounce> jcay, c_14: just tested it with several hand-typed umlauts (stock win7 install, german keyboard), and as soon as the _output_ contains an umlaut, the clobbering is triggered
[23:39:02 CEST] <jcay> hmm, I see
[23:39:13 CEST] <c_14> cbounce: have you tried the command in the standard windows shell instead of cygwin?
[23:39:23 CEST] <cbounce> c_14: lemme do that
[23:40:29 CEST] <cbounce> c_14: same error
[23:40:42 CEST] <cbounce> (or no error, still clobbering)
[23:41:09 CEST] Action: c_14 is going to blame Windows being terrible at anything that isn't ASCII
[23:41:48 CEST] <c_14> Might have something to do with the locale, or other weird things
[23:42:00 CEST] Action: cbounce feels sympathetic for c_14's point of view, but still would like to keep his files from being clobbered :(
[23:42:29 CEST] <c_14> use linux? :p
[23:42:46 CEST] <cbounce> so, bug report for ffmpeg? is there a special place for win32 related bugs?
[23:43:11 CEST] <cbounce> c_14: well, would like to use my work PC for transcoding while working :P
[23:45:03 CEST] <c_14> You can open an issue on trac and mention it's windows-specific.
[23:46:55 CEST] <jcay> doesnt sound like a direct ffmpeg problem, but yeah you can try with the bug report
[23:46:57 CEST] <c_14> That or try the zeranoe forums maybe.
[23:47:16 CEST] <c_14> Might be somebody's had the problem before.
[23:48:11 CEST] <cbounce> ok, thanks for the input guys
[23:54:04 CEST] <jcay> bah, forgot --enable-libmp3lame... gotta recompile again :D
[00:00:00 CEST] --- Mon May 25 2015