[Ffmpeg-devel-irc] ffmpeg.log.20130605

[burek]

[00:48] <llogan> bunniefoofoo: it is probably the "best" AAC encoder supported by ffmpeg, but it required --enable-nonfree, IIRC.
[00:48] <llogan> https://ffmpeg.org/trac/ffmpeg/wiki/AACEncodingGuide
[00:49] <llogan> *requires
[01:30] <jdolan> wow this is insanely frustrating:
[01:31] <jdolan> [AVFilterGraph @ 0x7fb1cac2c1a0] No such filter: ''
[01:31] <jdolan> i've got -v debug on, and it shows that it consumes the entire filter_complex value.
[01:31] <jdolan> yet i get that.
[01:34] <jdolan> http://pastebin.com/XNjmZAtG
[01:35] <jdolan> actually, here: http://pastebin.com/J6SZ75c9
[01:35] <jdolan> with output
[01:37] <ubitux> you can remove the escaping of the ',' since they're enclosed properly into a double quoting
[01:37] <ubitux> (what is your shell?)
[01:37] <jdolan> bash
[01:38] <ubitux> amerge output is not linked anywhere
[01:38] <ubitux> replace [video] with [out0] and add [out1] after amerge (and drop the trailing ';')
[01:38] <ubitux> btw, you can use -filter_complex_script for such filtergraph
[01:39] <jdolan> wow, the trailing ; was the killer here.
[01:39] <jdolan> you have to be kidding me.
[01:39] <jdolan> that's a brittle-ass parser :(
[01:39] <ubitux> ';' means you're expecting something after
[01:40] <ubitux> maybe we could improve that
[01:40] <ubitux> feel free to open a ticket
[01:40] <jdolan> everywhere else it means "end of instruction" :|
[01:40] <jdolan> _thank_ you tho
[01:40] <ubitux> not everywhere
[01:40] <jdolan> this is immensely helpful, all of your tips.
[01:40] <ubitux> not in camel for instance
[01:40] <ubitux> glad to see you happy
[01:41] <ubitux> again, feel free to open a feature request for supporting expression with trailing ';'
[01:45] <jdolan> sure thing, on the trac instance?
[01:45] <ubitux> yes, trac.ffmpeg.org
[01:58] <jdolan> https://ffmpeg.org/trac/ffmpeg/ticket/2641
[01:58] <llogan> jdolan: you will probably be asked for the complete console output
[02:46] <jdolan> weird.
[02:46] <jdolan> i should be able to filter pad with alpha fill, right?
[03:20] <KalDurak> @ubitux: You around?
[03:32] <defaultro> can ffmpeg overlap last 45 frames to the next file?
[03:58] <jdolan> hm. trying to stick a vorbis audio stream into an .mp4: nb_samples (512) != frame_size (64) (avcodec_encode_audio2
[03:59] <jdolan> that's the error message, which is only output at the very end of the encoding.
[03:59] <jdolan> i'm using -codec:a libvorbis
[05:28] <highgod> Hi, I want to ask a question, is there any example code to use ffmpeg device? I want to use ffmpeg dshow to capture audio, but did not find any code
[07:45] <meecola> hello! How i can stream the list of files without disconection on each file. i try to stream that:       while true; do; ffmpeg -re -i file -f mpegts udp://239.10.10.10:1234?pkt_size=1316; done. What i shod do for continuosly stream?
[08:02] <relaxed> meecola: that would keep streaming "file" one after another forever.
[08:05] <relaxed> you want to do that with multiple files?
[08:06] <meecola> relaxed: yes, but client drop frames after one ends and another starts. i need to stream it whithout any swiches, like live stream
[08:08] <meecola> i tryed concat feature, but its not worked for me
[08:09] <meecola> multiple files has different codecs
[08:09] <relaxed> oh, you need a for loop piping the multiple files into the while loop with one input
[08:12] <relaxed> you can probably drop while, unless you want to keep streaming forever.
[08:13] <meecola> thank you very much, i try it
[10:34] <ahuillet> hi
[10:37] <ahuillet> I'm using swscale to convert colorspaces from RGB to YUV420P, and I'm testing different flags. as far as I know, BICUBLIN does bilinear on U and V planes, so it should produce strictly the same output as BILINEAR on a RGB -> YUV420P conversion
[10:37] <ahuillet> it's not the case - is there something I misunderstood?
[10:43] <Narsilou> Hi, I'm looking for help on decoding h264 encoded packets (Basically the reverse of /doc/examples/decoding_encoding.c) But I'm having a segfault that i don't quite understand.
[10:51] <Narsilou> The problem is occurring around the following code http://pastebin.com/Eyc2JBSS. Debugger crashes within `avcodec_decode_video2`, and on variable inspection, it seems that it can't access my packet, any idea ? Do I need to initialize AVPacket with a function ?
[10:53] <ahuillet> Narsilou : looks good, but too short to tell
[10:53] <ahuillet> http://pastebin.com/AxEKxiRM
[10:53] <ahuillet> this is what I use, and it works
[10:53] <JEEB> Narsilou, av_init_packet()
[10:54] <Narsilou> I use av_init_packet before, just was not sure about assigning pointers and size afterwards.
[10:58] <Narsilou> Debugger points me to `  ret = avctx->codec->decode(avctx, picture, got_picture_ptr,  &tmp);` within the `avcodec_decode_video2`.  Could it be that I'm passing the address to a protected attribute (AVPacket m_pkt, not AVPacket* m_pkt) ?
[11:00] <ahuillet> your code is too short to tell.
[11:00] <ahuillet> please show whole code.
[11:00] <ahuillet> (or try with mine)
[11:04] <Narsilou> Here is the full code of this class http://pastebin.com/PxZBngTu (which is basically the whole code, the rest is juste some simple QtWindow). Looking at your code, the main difference I can see is `avcodec_get_frame_defaults(picture);` What does it do ?
[11:07] <ahuillet> see avcodec.h:  * AVFrame is typically allocated once and then reused multiple times to hold * different data (e.g. a single AVFrame to hold frames received from a
[11:07] <ahuillet>  * decoder). In such a case, avcodec_get_frame_defaults() should be used to
[11:07] <ahuillet>  * reset the frame to its original clean state before it is reused again.
[11:07] <ahuillet> I'm just following orders ;)
[11:17] <barneygale> Hi folks. I'm trying to generate a video out of a number of .tif files. My command is `ffmpeg -i screen-%04d.tif -vcodec mpeg4 lines.avi`. Is it possible to make the video repeat, and to set it to a certain length? I tried with `-loop_input -t 30` but it didn't seem to help.
[11:31] <Narsilou> ahuillet: Did not solve the problem, though now segfault happens in a more appropriate place. My packet->data seems to be screwed up. Why is this needed for decoding but not for encoding ?? (there is no such call to `avcodec_get_frame_defaults` in doc/examples/decoding/encoding.c)
[13:23] <JEEBsv> y
[13:41] <Novamedia> Hello we have a stange problem, trying to write avi container ( to be simple its just empty container )
[13:41] <Novamedia> Here is dump format Output #0, avi, to '/storage/nRita.read/build/nmi/storage/tmp/51_1370355861':
[13:41] <Novamedia>     Stream #0:0, 0, 1/90000: Video: mpeg4, yuv420p, q=2-31, 200 kb/s, 90k tbn, 0.04 tbc
[13:42] <Novamedia> and when we try write trailer for it with av_write_trailer
[13:42] <Novamedia> we got sigsegv  with Program received signal SIGSEGV, Segmentation fault.
[13:42] <Novamedia> [Switching to LWP 17700]
[13:42] <Novamedia> avi_write_trailer (s=0x44c11bb0) at libavformat/avienc.c:603
[13:42] <Novamedia> 603             if (avi->riff_id == 1) {
[13:42] <Novamedia> (gdb) bt
[13:42] <Novamedia> #0  avi_write_trailer (s=0x44c11bb0) at libavformat/avienc.c:603
[13:44] <ahuillet> have you tried valgrind? address sanitizer?
[13:44] <Novamedia> no because it is Tegra III board
[13:45] <Novamedia> currently we dont have such tools
[13:45] <ahuillet> and your "avi" pointer is NULL, on line 603?
[13:47] <Novamedia> looks like it from gdb output
[13:48] <ahuillet> I believe you should *check*, and work your way backwards from there, unless someone here has a better idea of what can be wrong
[13:48] <ahuillet> but really without a code sample and more information it's difficult to tell
[13:49] <Novamedia> hm how can present code sample
[13:49] <Novamedia> ?
[13:49] <ahuillet> we need to see what you are doing, right?
[13:50] <Novamedia> Moreover ther are some strange errors from ffmpeg in order of apperance
[13:50] <Novamedia> [IMGUTILS @ 0x4b4428a4] Picture size 0x0 is invalid
[13:50] <Novamedia> [NULL @ 0x45a0b9e0] Ignoring invalid width/height values
[13:50] <Novamedia> [avi @ 0x45a0b4f0] dimensions not set
[13:50] <ahuillet> doesn't sound like an error from ffmpeg, but you giving it wrong parameters...
[13:51] <Novamedia> there is open file function with initialization that looks like this
[13:51] <Novamedia>     av_log_set_level(AV_LOG_DEBUG);
[13:51] <Novamedia>     // presentation timestamp
[13:51] <Novamedia>     m_avpkt.pts   = 0;
[13:51] <Novamedia>     // decompresion timestamp
[13:51] <Novamedia>     m_avpkt.dts   = 0;
[13:51] <Novamedia>     // position in stream, -1 -> unknown
[13:51] <Novamedia>     m_avpkt.pos   = -1;
[13:51] <Novamedia>     // duration of this packet in AVStream->time_base units
[13:51] <Novamedia>     // 0 -> unknown
[13:51] <Novamedia>     m_avpkt.duration = 0;
[13:51] <Novamedia>     // Time diff in av stream units between pts of this packet and point at which
[13:51] <ahuillet> no no no
[13:51] <Novamedia>     // output from the decoder has converged independent from availability of previous frame.
[13:51] <Novamedia>     m_avpkt.convergence_duration = 0;
[13:51] <ahuillet> use pastebin.com
[13:51] <Novamedia>     // AV_PKT_FLAG
[13:51] <Novamedia>     m_avpkt.flags = 0;
[13:51] <Novamedia>     m_avpkt.stream_index = 0;
[13:51] <Novamedia>     m_avpkt.destruct = NULL;
[13:51] <Novamedia>     av_register_all();
[13:51] <Novamedia>     avcodec_register_all();
[13:51] <Novamedia> ahh sory
[13:52] <ahuillet> and paste the whole code while you're at it.
[13:55] <Novamedia> Ok here are our function http://pastebin.com/WmXpC4sT
[13:55] <Novamedia> we jus creatie such object and call openFile, then close.
[13:56] <ahuillet> Narsilou = Novamedia ?
[13:56] <ahuillet> same person, same company... ? or am I mistaken?
[14:00] <Novamedia> Hm dunno, maybe it was one of our previous dev.
[14:00] <Novamedia> but not me
[14:01] <Novamedia> Any way this code is working for us on desktop machines compiled with gcc4.7.3
[14:01] <Novamedia> But fails on gcc-arm
[14:01] <Novamedia> (based on 4.7.2)
[14:02] <Novamedia> after try to write header there is something like that in logs
[14:02] <Novamedia> in: -4713-0-1 0:0:0
[14:02] <Novamedia> [avi @ 0x44c11bb0] in: -4713-0-1 0:0:0
[14:02] <Novamedia> dimensions not set
[15:54] <GreatEmerald> Is there a way to tell ffmpeg to stop recording as soon as the pipe input stops transferring data?
[15:55] <GreatEmerald> That is, I have two inputs, a pipe and an a PulseAudio device, and I want it to stop recording as soon as the pipe is terminated
[16:45] <NovamediaDev1> Hello, i have a simple question, how can i list what was initialized with calls to av_register_all(); avcodec_register_all(); or in another words set ffmpeg  debug output tu such level?
[17:09] <NovamediaDev1> Maybe another question, is it wrong to call such function like av_register_all(); avcodec_register_all(); multiple times from code?
[18:34] <Narsilou1> ahuillet: Hi again, I've looked again at my problem seems avctx->codec->decode is 0 shouldn't it be a pointer to a function ? (`ret = avctx->codec->decode(avctx, picture, got_picture_ptr,&tmp);` is the incriminate lign within ffmpeg)
[18:44] <Narsilou1> I should use `codec = avcodec_find_decoder(codec_id);` instead of `codec = avcodec_find_encoder(codec_id);`
[18:44] <ahuillet> whow
[18:44] <ahuillet> yes, obviously.
[19:51] <killown> hi everyone, is that possible to split the output while ffmpeg is recording the screen?
[19:56] <saste> killown, "split the output"^
[19:56] <saste> killown, "split the output"?
[20:06] <killown> saste hi, thanks for the reply, is that possible to split while ffmpeg is recording the screen?
[20:07] <saste> killown, define "split"
[20:07] <killown> for example: I mount tmpfs with 2gbram and when the output reaches 500MB I want it to be spitted
[20:07] <killown> then a new output is used
[20:08] <saste> killown, check segment muxer
[20:08] <killown> I just need to free space in the tmpfs
[20:09] <killown> saste, thanks a lot, I will check it
[20:10] <saste> killown, right now there is no way to split on the base of the output size
[20:12] <killown> saste, do you know if such feature is being planned to be implemented?
[20:27] <bunniefoofoo>  is it a bad idea to have a frame size that is not a multiple of 16. I have tried this with mpeg4 asp and valgrind shows a buffer overflow
[20:33] <durandal11707> bunniefoofoo: please report bugs to bug tracker
[20:34] <durandal11707> its not bad idea to have frame size not multiple of 16
[20:58] <ayaka>  after read https://ffmpeg.org/doxygen/trunk/doc_2examples_2muxing_8c-example.html , I wonder can I use AVPicture to store a encoded frame like a encoded frame which codec is h.264
[21:03] <luc4_mac> Hi! After I build ffmpeg from git, should I expect to see the ffmpeg executable somewhere? I just find the libs in the prefix...
[21:07] <durandal11707> luc4_mac: you installed it too?
[21:07] <hackeron> hey, I have an ACTi IP camera, I'm trying to read an RTSP from it with ffmpeg, but it only reads 30 seconds and quits. I'm reading with: ffmpeg -i rtsp://admin:123456@192.168.1.38:7070/track -codec:v copy test.h264 -- it seems ffmpeg isn't sending the periodic "option" requests to keep the stream alive? - Is there some parameter I can add to get ffmpeg to send those?
[21:08] <luc4_mac> durandal11707: yes& but I see only the libs and the headers...
[21:08] <durandal11707> where?
[21:08] <luc4_mac> durandal11707: in the prefix
[21:08] <luc4_mac> durandal11707: I suspect& the binaries were placed in the same dir of the sources for some reason& is this possible?
[21:10] <luc4_mac> durandal11707: I followed the ubuntu compilation guide& but the "Finish" part is a little weird& it is written to try to run with "ffmpeg 2>&1 | head -n1"& shouldn't that run the ffmpeg bin in the path?
[21:10] <durandal11707> luc4_mac: its installend in prefix, which can be /usr/local
[21:11] <luc4_mac> durandal11707: I specified another prefix...
[21:11] <durandal11707> than another prefix is not in your path?
[21:11] <durandal11707> what "which ffmpeg" displays?
[21:11] <luc4_mac> durandal11707: no
[21:12] <luc4_mac> durandal11707: nothing& I didn't install ffmpeg in the path
[21:12] <luc4_mac> durandal11707: just running ffmpeg finds nothing, of course
[21:12] <durandal11707> so what you want, run ffmpeg or?
[21:12] <luc4_mac> durandal11707: run the ffmpeg I built building the sources& :-)
[21:13] <luc4_mac> durandal11707: after the build is completed& where is the binary located if not in the prefix I specified?
[21:13] <durandal11707> then run it with full path where you installed it....
[21:13] <luc4_mac> durandal11707: that is my question :-)
[21:13] <luc4_mac> durandal11707: I specified the path, I go in there, and I see everything but the executables...
[21:16] <luc4_mac> durandal11707: ok, problem solved rebuilding :-) for some reason, the build installs everything in the prefix I specified, except the binaries, which are placed in the sources directory. I couldn't find those because I followed the guide which was also clearing temporary files.
[21:16] <luc4_mac> durandal11707: a little weird anyway...
[21:16] <luc4_mac> durandal11707: I commonly see binaries placed in the prefix with the libraries, headers and so on...
[21:17] <durandal11707> luc4_mac: biaries are installed in prefix/bin
[21:22] <luc4_mac> durandal11707: not here...
[21:22] <jdolan> i've managed to put ffmpeg into an infinite loop :(
[21:22] <jdolan> i'm using an image as a background and overlaying video atop it.
[21:22] <jdolan> if i don't use -loop 1 -i background.jpg, i never see the movie.
[21:23] <jdolan> if i do use -loop 1 -i background.jpg, its an infinite loop.
[21:23] <jdolan> tho it does show the movie.
[21:25] <jdolan> durandal11707: is that a Marathon reference?
[21:26] <luc4_mac> durandal11707: that is what I was expecting& but no bin inside there...
[21:28] <durandal_1707> jdolan: what version?
[21:28] <jdolan> durandal_1707: 1.2.1 through Macports,.
[21:28] <durandal_1707> luc4_mac: no bin dir in prefix ? (that is really strange)
[21:29] <jdolan> the -loop option is actually not documented anywhere that i can find.
[21:29] <jdolan> i'm prefixing each background image input with -loop 1
[21:29] <jdolan> maybe i should only add -loop 1 once?
[21:31] <luc4_mac> with make install I see: "INSTALL	ffmpeg". In the prefix I find libs, presets, headers& but no ffmpeg binary. It is in the root of the source tree.
[21:34] <jdolan> does anyone know of documentation for the -loop option?
[22:55] <jdolan> assuming i need to encode video to some temporary file that i'm going to immediately re-encode, what is the fastest encoder available?
[22:55] <jdolan> in other words, i don't care about file size, i just want to dump the video to disk quickly and make it easy to encode again in the very next step.
[22:55] <jdolan> i was thinking AVI?
[22:55] <iive> avi is just file format (container)
[22:55] <bunniefoofoo> is it possible to create a pan&zoom effect for a still image, like for a slide show, using avfilters? I am looking at the scale filter and I don't see expression option
[22:56] <bunniefoofoo> jdolan, I like ffvhuf, don't know if its the fastest
[22:56] <bunniefoofoo> mpeg2 is very fast too
[22:57] <jdolan> nice, thanks bunniefoofoo
[22:57] <iive> mjpeg is usual trade off. it is like mpeg2 with keyframes only.
[22:58] <iive> you can try ultrafast on x264 and very low quant. qp=0 is looseless
[22:58] <bunniefoofoo> mpeg2 with -sameq (don't know if this option still works) is virtually lossless and decompresses *really* fast
[22:58] <iive> bunniefoofoo: no such thing as virtually lossless.
[22:59] Action: JEEBsv_ cries as someone mentions sameq
[22:59] <JEEBsv_> it
[22:59] <JEEBsv_> never
[22:59] <bunniefoofoo> I mean practically lossless?
[22:59] <JEEBsv_> was
[22:59] <JEEBsv_> "same quality"
[22:59] <JEEBsv_> asdf
[22:59] <iive> -sameq have no meaning when capturing, because it just uses same quantizer as the source
[22:59] <bunniefoofoo> I guess that would be "-qscale 0" if that's an option?
[23:00] <iive> only x264 supports quant parameter of 0
[23:15] <Moeabm> So ive installed ffmpeg from this guide https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide. Now i want to add --enable-libfreetype. Do I have to Uninstall ffmpeg completely and recompile?
[23:16] <bunniefoofoo> you don't have to uninstall, just reconfigure it and it will blast over what was there before (assuming it went in /usr/local)
[23:17] <bunniefoofoo> I guess in this example they have it going in $HOME/ffmpeg_build
[23:17] <bunniefoofoo> in short the "make install" command for most sw (including ffmpeg) replaces what is already there
[23:22] <Moeabm> Thank you!
[23:23] <bunniefoofoo> you are welcome, sir
[23:27] <CypherPunk38> Hello everyone. Is there a way to tell ffmpeg to process up to a specific timestamp in a video instead of a time span?
[23:31] <llogan> Moeabm: the guide does not install anything you compile to the system, so it will not interfere with any repositroy packages, including the so-called "ffmpeg" package.
[23:32] <llogan> maybe i misunderstood your quesiton. if you compiled ffmpeg, but now want ffmpeg to support libfreetype, then you will have to recompile.
[23:39] <CypherPunk38> Sorry everyone got booted. If anyone answered my question I didn't see it. My question was: is there a way to have ffmpeg work until a specific timestamp in a video instead of a time range?
[23:40] <bunniefoofoo> cypher, the file doesn't start at 00:00:00 ?
[23:55] <pyBlob> I want to call ffmpeg from a java application, but I don't know which command to use. I want to be able to give ffmpeg a video over stdin/filename, which then outputs the single frames as raw-color-arrays(argb) over stdout
[23:56] <pyBlob> -> what parameters do I need to specify?
[23:58] <bunniefoofoo> pyblob, I don't think you can do a frame sequence over stdout, would probalby have to be a container to hold those frames
[23:59] <pyBlob> ok, would it be possible to put a bmp/png-sequence into an .avi-container?
[00:00] --- Thu Jun  6 2013