[Ffmpeg-devel-irc] ffmpeg.log.20141218

burek burek021 at gmail.com
Fri Dec 19 02:05:01 CET 2014


[01:37] <agumonkez> hi there
[01:39] <agumonkez> I ripped some video stream (tv news) as a set of "mpeg transport stream data" (output of `file *`)
[01:43] <agumonkez> I could merge the chunks using cat, mpv can play it fine. I just wanted to cut the last 4min of the video, but I get aac muxing overhead errors http://pastebin.com/cKnTniX7
[01:43] <agumonkez> I tried to add the addtstoaac filter, same result (no audio under mpv)
[01:44] <agumonkez> hold on.. for a reason I get sound now u____u;
[01:44] <pzich> you're welcome
[01:44] <agumonkez> sorry for the vacuous rubber ducking session
[01:45] <agumonkez> I hope it was entertaining anyway
[01:45] <c_14> rubber ducky go quack quack
[01:45] <pzich> I've not dealt with that sort of muxing before, so it was at least somewhat interesting
[01:45] <agumonkez> what do rubber monkeys sound like ?
[01:46] <c_14> agoo?
[01:47] <agumonkez> I'll roll with that
[05:39] <relaxed> c_14: hmm, no. Let me look into that.
[05:43] <relaxed> Oh, I see the issue. It will be fixed in the next cycle. Thanks for the heads up.
[10:25] <arnaud> Hi
[10:26] <arnaud> I'm upgrading from an old ffmpeg version to 2.5, and I'm facing some problems. Hope someone here will be able to help me/give me some hints :)
[10:27] <arnaud> Some flags we used are now deprecated and don't exist anymore. e.g. CODEC_FLAG2_STRICT_GOP
[10:27] <arnaud> changed with this commit: http://git.videolan.org/gitweb.cgi/ffmpeg.git/?p=ffmpeg.git;a=commitdiff;h=a249f0cc23457bc73a249e1b1ca22adcfe5b0894
[10:27] <arnaud> So we should now use the encoder "private options"
[10:27] <arnaud> I'm not sure what it is, and how to change this.
[10:28] <arnaud> And I couldn't find information about this in the documentation (unless I miss something)
[10:30] <pzich> https://www.ffmpeg.org/ffmpeg-codecs.html#Codec-Options 'Options may be set by specifying -option value in the FFmpeg tools, or by setting the value explicitly in the AVCodecContext options or using the libavutil/opt.h API for programmatic use.'
[10:32] <arnaud> pzich: thanks :) I tried something like this, but it didn't work (I'm certinaly wrong, but not sure what I'm missing)
[10:32] <arnaud> I tried:
[10:32] <arnaud> av_opt_set_int(enc->priv_data, "strict_gop", 1, AV_OPT_SEARCH_CHILDREN)
[10:33] <arnaud> (enc being a AVCodecContext*)
[10:33] <pzich> I've never used ffmpeg programmaticly, so I'm afraid I can't help much further.
[10:34] <arnaud> ok, thanks anyway :)
[10:34] <pzich> what I'd do is look for a more common private option, then search around online for code or documentation that uses it
[10:34] <arnaud> Hope someone else can help me
[10:34] <arnaud> pzich: I tried, but didn't find anything :(
[10:34] <pzich> I doubt it, at this hour, but maybe.
[12:07] <tonnzor_> hi all, could anybody explain why nullsrc produce garbage?
[12:08] <tonnzor_> e.g. ffmpeg -y -f lavfi -i "nullsrc=size=100x100,drawtext=text=demotext:fontfile=/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf:fontcolor=red" XXX
[12:09] <tonnzor_> I want to generate a transparent input stream, put something on top (e.g. drawtext) and overlay this input over my file
[12:14] <saste> tonnzor_, nullsrc doesn't initialize memory
[12:14] <saste> if you want inited memory you should use for example color
[12:15] <tonnzor_> good to know :) how to produce transparent input using color? color=black at 0.0 will make whole layer transparent (i.e. including later operations)
[12:31] <tonnzor_> e.g. -vf 'color=size=100x100:color=black at 0.1,drawtext=text=demo:fontcolor=red at 1.0 [txt]; [in][txt]overlay' will produce invisible text
[12:33] <tonnzor_> @saste Any hint?
[12:39] <saste> tonnzor_, -filter_complex "color=...[background];[0:v] overlay" ?
[12:40] <saste> tonnzor_, also, why don't you directly use drawtext if you want semi-transparent text?
[12:42] <tonnzor_> I want it vertical on left side, i.e. -90 degree. it I draw directly on top of image -- I have to transpose clock, draw, transpose cclock
[12:42] <tonnzor_> i.e. convert whole stream 2 times. it sounds pretty heavy
[13:33] <mac_nibblet> Im getting this error (Command inlined) https://gist.github.com/macnibblet/3e6d77031058cde11188
[13:33] <mac_nibblet> Im guessing it's a missing dependency since i have compiled ffmpeg myself
[13:34] <mac_nibblet> Configuration used when compiling is: --enable-nonfree --enable-gpl --enable-libfdk_aac --enable-libx264 --enable-decoder=png --enable-encoder=png
[14:31] <Elirips> Hello. Whats wrong with the following call to ffmpeg: 'ffmpeg -i rtsp://webcam -c:v libx264 -x264opts keyint=15 -an -r 5 -f tee -map 0:v "[f=flv]rtmp://localhost:1935/hls/axis5fps|[f=flv]rtmp://localhost:1935/flash/axis5fps"' ?
[14:33] <Elirips> I get a 'av_interleaved_write_frame(): Broken pipe error
[14:33] <Elirips> But streaming to just one of the outputs, using -f flv rtmp://... works fine, I must be doing something wrong with tee
[14:39] <LigH> Greetings.
[14:40] <LigH> Would ffmpeg be able to pack a sequence of JPEG images into e.g. an MJPG AVI without reencoding?
[14:41] <LigH> Assumed that dimensions are identical and format is video compatible (4:2:0).
[14:52] <Elirips> Anyone an idea whats wrong with my 'tee' ? This line works: 'ffmpeg -i rtsp://webcam -c:v libx264 -x264opts keyint=15 -an -r 5 -f flv rtmp://localhost:1935/hls/axis5fps' but, not if I try the same using tee: 'ffmpeg -i rtsp://webcam -c:v libx264 -x264opts keyint=15 -an -r 5 -f tee -map 0:v "[f=flv]rtmp://localhost:1935/hls/axis5fps"' - fails with 'av_interleaved_write_frame(): Broken pipe'
[18:26] <ribasushi> are any developers awake who can update an ffmpeg ticket?
[18:26] <ribasushi> registration hoops make me unhappy ;)
[18:32] <ubitux> tell us more
[18:35] <ribasushi> ubitux: my email to the uvc mailing list isn't going through for some reason
[18:35] <ribasushi> so can't link to it yet :(
[18:36] <ribasushi> the root-cause of https://trac.ffmpeg.org/ticket/3956 is an (already bisected) kernel bug <my link will go here>
[18:38] <ubitux> ribasushi: so i can close? are you going to give me more information about the kernel bug?
[18:38] <ribasushi> ubitux: as soon as the stupid greylister takes my email :(
[18:39] <ribasushi> ubitux: will let you know in 5 mins
[18:39] <ubitux> sure
[18:40] <ribasushi> but the gist is - the same ffmpeg invocation works before a kernel commit, and stops working after, so in the spirit of "never break userspace" I think the kernel is to blame and not ffmpeg
[18:41] <ribasushi> ubitux: there we go: http://sourceforge.net/p/linux-uvc/mailman/message/33164469/
[18:41] <ubitux> sure; but if the kernel fixed that and has a better api usage to suggest, feel free to raise it
[18:41] <ubitux> ok, i'm closing the ticket
[18:41] <ribasushi> ubitux: my C is terrible, and the patch in question is humongous - I can't tell who/what is in play
[18:42] <ribasushi> ubitux: it looks like the uvc driver started calculating its own PTS values, and in the case of the variable C920 rate it calculates them wrong... or something
[18:42] <ubitux> ribasushi: you are gtzen?
[18:42] <ribasushi> no
[18:42] <ribasushi> I just ran across this ticket
[18:42] <ubitux> ok
[18:46] <ubitux> ribasushi: thank you, done
[18:46] <ubitux> if you need me to add or change something, do it now; i'm going afk pretty soon
[18:46] <ribasushi> ubitux: that comment is spot-on, thanks
[18:47] <ribasushi> hopefully there will be a usable response from the uvc folk soon ;)
[18:47] <ubitux> alright
[18:47] <ubitux> feel free to poke me again where there is an answer
[18:47] <ubitux> i'm probably not going to monitor the discussion
[18:47] <ribasushi> nod
[18:47] <ubitux> anyway, gtg, thanks
[18:47] <ribasushi> o/
[20:20] <targitaj> hello
[20:25] <targitaj> privet
[20:25] <targitaj> cd C:\Temp
[20:25] <targitaj> ffmpeg -i 1.avi -f mpegts "tcp://localhost:1234"
[20:25] <targitaj> pause
[20:25] <targitaj> http://andrey.mosalsky.com/20141218-965-63kb.jpg
[20:26] <targitaj> ne polu4ajetsa po instrukcii zapustitj strim :(
[20:26] <targitaj> pomogite pozalujsta
[20:27] <targitaj> mozet u gogo jestj sample
[20:28] <targitaj> http://a.mosalsky.com:2880/?source=C%3A%5Ctorrent14%5CDva.otca.i.dva.syna
[20:29] <targitaj> 2>B :>5 GB> @50;87>20; A <><>ILN ffpmeg
[20:29] <targitaj> teperj hotelosj bi strimitj nau4itsa
[20:45] <targitaj> jestj tut kto
[20:45] <targitaj> is anybody here
[20:45] <targitaj> ?
[20:46] <c_14> If you could ask your question in English, we could try and help.
[20:47] <targitaj> I'm trying to create stream using ffmpeg, but examples doesn't help me
[20:47] <targitaj> cd C:\Temp
[20:47] <targitaj> ffmpeg -i 1.avi -f mpegts "tcp://localhost:1234"
[20:47] <targitaj> pause
[20:47] <targitaj> http://andrey.mosalsky.com/20141218-965-63kb.jpg
[20:47] <c_14> Is there something listening on port 1234 or do you want ffmpeg to create a "tcp server" there?
[20:48] <targitaj> may be you have a sample
[20:49] <targitaj> how i can run stream with ffmpeg
[20:49] <targitaj> and play it with vlc
[20:50] <targitaj> my final wish it is playing it with jwplayer
[20:50] <c_14> use "tcp://localhost:1235?listen"
[20:50] <c_14> then with vlc play tcp://localhost:1235"
[20:50] <c_14> -"
[20:52] <targitaj> oooo big thanks
[21:03] <targitaj> if someone whants web site wich can convert video on your pc(windows server) and watch this video in browser, ask I'l giv it
[21:04] <targitaj> http://a.mosalsky.com:2880/?source=Dva.otca.i.dva.syna.%2801.seriya%29.2013.HQWEB-DLRip.Alexey724.mp4
[21:04] <targitaj> http://a.mosalsky.com:2880/?source=C%3A%5Ctorrent14%5CDva.otca.i.dva.syna
[22:38] <DrSlony> Hi, what flags can one enable during compilation to make ffmpeg faster?
[22:38] <pzich> --enable-fast and --dont-be-slow
[22:38] <c_14> make ffmpeg go faster
[22:39] <DrSlony> :)
[22:39] <DrSlony> i.e. does it auto-detect optimized instruction sets and so on?
[22:39] <DrSlony> do i have to enable sse2 manually?
[22:39] <pzich> pretty sure it should
[22:39] <DrSlony> it fails to see my opencl libs :(
[22:39] <DrSlony> "ERROR: opencl must be installed and version must be 1.2 or compatible"
[22:40] <DrSlony> everything's in /usr/lib64/OpenCL/global/include/CL/
[22:40] <pzich> are they in a path it's looking for?
[22:41] <c_14> is everything up to include/ in your include path?
[22:41] <DrSlony> which env var specifically?
[22:41] <c_14> CFLAGS -I
[22:43] <DrSlony> should it be CFLAGS="-I /usr/lib64/OpenCL/global/include/CL/" ?
[22:43] <c_14> without the CL
[22:43] <c_14> just /usr/lib64/OpenCL/global/include
[22:44] <c_14> as long as cl.h is at /usr/lib64/OpenCL/global/include/CL/cl.h or OpenCL/cl.h
[22:47] <DrSlony> yes its in CL/
[22:50] <c_14> then it should work with cflags like what you have above
[22:53] <DrSlony> same error :/
[22:53] <DrSlony> CFLAGS="-I/usr/lib64/OpenCL/global/include" PKG_CONFIG_PATH=$HOME/programs/vidstab/lib/pkgconfig $HOME/programs/code-ffmpeg/configure --prefix=$HOME/programs/ffmpeg --enable-nonfree --enable-frei0r --enable-gpl --enable-libfdk-aac --enable-libx264 --enable-libvidstab --disable-shared --enable-static --enable-encoder=ffvhuff --enable-encoder=ffv1 --enable-opencl
[22:53] <c_14> Can you pb config.log ?
[22:54] <DrSlony> sure
[22:54] <DrSlony> c_14 http://filebin.net/4zkbaepafo
[22:56] <c_14> What version of opencl do you have?
[22:57] <DrSlony> i dont know how to check
[22:57] <DrSlony> lets try cat
[22:57] <c_14> that is, is there a #define for CL_VERSION* in cl.h ?
[22:58] <DrSlony> damn, looks like 1.1
[22:58] <DrSlony> 1_0 and 1_1
[22:59] <c_14> You'll need a newer version of opencl then.
[22:59] <DrSlony> yes, though its a mystery to me how to update that
[23:00] <DrSlony> gentoo
[23:03] <DrSlony> jesus
[23:03] <DrSlony> dev-util/nvidia-cuda-toolkit is 950MB
[23:16] <DrSlony> instlaled it, is installed /opt/cuda/include/CL/cl.h but thats also 1_0 and 1_1
[23:22] <c_14> Hmm, it looks like current Nvidia drivers (for linux at least) only support up to opencl 1.1
[23:28] <DrSlony> bugger!
[23:28] <DrSlony> and im stuck with 340.* because 343 and newer dont support my GPU
[00:00] --- Fri Dec 19 2014


More information about the Ffmpeg-devel-irc mailing list