[Ffmpeg-devel-irc] ffmpeg.log.20121020

[burek]

[00:07] <durandal_1707> needs_more_info
[00:07] <durandal_1707> does mkv file with aac have same duration ?
[00:07] <Endorgh> yes
[00:08] <Endorgh> mkv containing aac_he_v2 has the same duration as the original input file
[00:09] <durandal_1707> what you use to play mp4?
[00:11] <Endorgh> deadbeef
[00:12] <durandal_1707> does ffmpeg -i input.mp4 show correct duration?
[00:17] <Endorgh> durandal_1707: yes, it shows correct duration
[00:21] <durandal_1707> so it appears to be deadbeef problem and how it calculates duration
[00:22] <durandal_1707> how are deadbeef libav* plugins build?
[00:22] <durandal_1707> perhaps it uses older versions of library....
[00:24] <Endorgh> may be, I'll investigate about it...
[00:25] <Endorgh> otherwise, when I convert flac to aac_he_v2 into mp4, deadbeef, aqualung and VLC are having problems playing it
[00:26] <Endorgh> coding aac_he into mp4 works well on all of them, but not aac_he_v2
[00:26] <Endorgh> its confusing........ :S
[00:29] <durandal_1707> does aac_he_v2 decodes correctly with ffmpeg?
[00:29] <durandal_1707> (it may be that such profile is just broken)
[00:32] <Endorgh> yes, perfectly. If I convert from mp4 to mp3, it works perfect...
[00:32] <Endorgh> in any case, How can I check if 'aac_he_v2' profile is broken?
[00:35] <durandal_1707> hmm, dunno perhaps deadbeef use broken aac decoder
[00:35] <Endorgh> I understand
[00:37] <Endorgh> in last place and in order to discard any possible error made by myself, this command is 100% correct?
[00:38] <Endorgh> ffmpeg1 -i input.flac -acodec libfdk_aac -profile:a aac_he_v2 -ab 60k -ar 44100 -ac 2 output.mp4
[00:39] <Endorgh> it should give me an mp4 file containing high quality AAC-HE v2 audio?
[00:43] <durandal_1707> if you play it with quicktime and it plays correctly then yes
[00:46] <DelphiWorld> yo saste
[00:46] <saste> hi DelphiWorld
[00:46] <DelphiWorld> saste: do you know Sast-NG?
[00:46] <saste> saste.-ng?
[00:46] <Endorgh> durandal_1707: ok, I'll try it!
[00:46] <Endorgh> thanks!!
[00:46] <saste> an improved version of saste?
[00:47] <DelphiWorld> saste: first of all what is sast ?
[00:47] <saste> i have no idea
[00:48] <DelphiWorld> saste: saste-ng is a softcam ;)
[00:48] <DelphiWorld> saste: when i see you i remember it;-)
[00:50] <saste> can't find it on the internet
[00:50] <DelphiWorld> saste: please kill me:(
[00:50] <DelphiWorld> saste: my assness
[00:50] <DelphiWorld> saste: it's sasc:P
[00:52] <DelphiWorld> nighty
[12:14] <hendry> hi there, when I convert my OGV videos from recordmydesktop like so: ffmpeg -loglevel quiet -i out.ogv youtube-upload.avi
[12:14] <hendry> the avi files become "artefacted" and not as clear as the original out.ogv
[12:14] <hendry> it's like converting a PNG to JPG
[12:15] <hendry> any tips to avoid this lossy conversion, yet still be able to upload to Youtube ?
[12:15] <sacarasc> Capture your desktop using ffmpeg.
[12:15] <Tjoppen> why? just upload the original file
[12:15] <sacarasc> That's what recordmydesktop uses...
[12:17] <hendry> sacarasc: is there a guide to doing that somewhere?
[12:18] <hendry> Tjoppen: the orginal file that recordmydesktop produces: out.ogv, does not upload to Youtube.
[12:19] <hendry> sacarasc: for e.g. `ffmpeg -f x11grab -s wxga -r 25 -i :0.0 -sameq /tmp/out.mpg` doesn't record audio
[12:20] <Tjoppen> that's strange, because youtube uses ffmpge
[12:20] <Tjoppen> also: don't use avi
[12:20] <Tjoppen> ffmpeg -i out.ogv -vcodec libx264 -crf 20 -acodec copy out.mov   or something like that
[12:23] <hendry> Tjoppen: why not .avi ?
[12:39] <hendry> Tjoppen: btw sound is not working in the out.mov
[13:23] <RoyK> are there any support for opencl/cuda in ffmpeg for h.264 or web encoding?
[13:27] <JEEB> RoyK, there is a patch that's under review currently, which moves lookahead me to opencl in x264 IIRC
[13:27] <JEEB> but
[13:27] <JEEB> it's the usual case of "do I really want to lower the quality and get a 10% speed boost with a high-end GPU instead of just investing on a new CPU"
[13:29] <RoyK> it§s just that a colleague of mine told me he got a 4x speed gain by using the gpu on his machine with some proprietary software
[13:29] <RoyK> s/it§s/it's/
[13:29] <JEEB> first of all, did he properly test or just see the speed with random settings?
[13:30] <JEEB> because x264 can in general be very fast f.ex.
[13:31] <JEEB> also I'm not even sure how fast those GPU-based encoders are nowadays, but they mostly suck because GPUs need you to be able to throw a lot of stuff at the same time, aka the workload really has to be well multithreadable. Unless you strip features out of H.264 to its barebones you're not going to get that simply on the GPU
[13:32] <JEEB> when you have something using GPU for certain tasks, that /can/ be faster, but not always without a cost and the actual CPU algorithm might've not been fast to begin with
[13:33] <hendry> anyone know how i get the playing length of an AVI file?
[13:33] <JEEB> the integrated encoder chips in intel stuff f.ex. are really dedicated hardware for encoding and the results of those generally are quite a bit better than pure GPU encoders'
[13:34] <JEEB> but those too are mostly currently riding on the side of "we're fast and we're not using the 'CPU'", and the speed might be attainable with similar/better quality from x264 with a fast preset
[13:36] <JEEB> anyways, OpenCL/CUDA is not a silver bullet under any circuimstances. CUVID decoding can help on newer hardware stuff to take some stuff off of the CPU, but if you go fast enough on the encoder's side that might end up being your bottleneck since the hw decoder can only output as many frames per second
[16:16] <tomreyn> hi, i'm trying to analyze the audio files we have in MegaGlest (an open source RTS game, http://megaglest.org ), to get an idea whether playing them back is causing elevated CPU load (memory matters, too, but less so), and to find out ways to reduce the overhead they cause without decreasing the audible quality more than a little bit.
[16:17] <tomreyn> here's what i find, including high (?) bitrates of 1411 kbps: http://paste.ubuntu.com/1292282/
[16:17] <tomreyn> i'm looking into this because profiing the game turns out that a rather high percentage (up to 8%) of the processing power of the test system goes into audio processing.
[16:18] <tomreyn> now any suggestions on what to look into would be welcome, i'm pretty helpless with audio and audio processing.
[16:22] <ubitux> maybe ./ffmpeg -nostats -i input.wav -benchmark_all -f null - ?
[19:57] <Element9> while trying to configure ffmpeg for compiling, I got "libtheora not found". I tried with ffmpeg 1.0 package and the latest code from git. is this a known problem could it be that I am doing something wrong?
[19:58] <durandal_1707> Element9: paste relevant config.log somewhere
[20:02] <hendry> i'm trying to grab to a mov, but
[20:02] <hendry> ffmpeg -f x11grab -s 1366x768 -i :0.0 -vcodec libx264 -crf 20 -acodec libfaac -f alsa -i hw:0,0  out.mov # results in Unknown decoder 'libx264'
[20:02] <hendry> any ideas?
[20:03] <JEEB> before -i = settings for decoders/inputs
[20:03] <JEEB> so in other words exactly what it told you
[20:03] <JEEB> there is no such decoder as libx264 (and no libfaac decoder either)
[20:03] <JEEB> probably because of the fact that you have two -is
[20:04] <JEEB> also I would recommend fdk instead of faac if you want aac
[20:04] <JEEB> https://github.com/mstorsjo/fdk-aac
[20:06] <hendry> JEEB: `ffmpeg -f x11grab -s 1366x768 -i :0.0 -vcodec libx264 -crf 20 -acodec libfaac out.mov` works, so libx264,libfaac do exist
[20:06] <hendry> or must exist
[20:07] <hendry> it's just that cmdline has no sound
[20:07] <hendry> so it's useless to me :)
[20:08] <klaxa> well obviously since you don't specify an audio input
[20:08] <JEEB> hendry, did you like ignore all of what I said?
[20:09] <JEEB> re-read my lines
[20:09] <JEEB> also encoders are not decoders
[20:09] Action: JEEB sighs
[20:09] <klaxa> kinda spoonfeeding here, but try: ffmpeg -f x11grab -s 1366x768 -i :0.0 -vcodec libx264 -crf 20 -f alsa -i hw:0,0 -acodec libfaac out.mov
[20:10] <klaxa> and the fdk stuff... no idea about that though
[20:12] <JEEB> faac is actually pretty bad, but I guess building yet another binary is not an alternative to this guy :P
[20:12] <hendry> klaxa: that didn't work either
[20:13] <hendry> you make spoonfeeding sound like a crime
[20:13] <klaxa> opus looks good, i think it misses a lot of container implementations though... as a standalone audio-codec it has a great quality/bitrate ratio
[20:13] <hendry> this stuff is really rather tricky ffs
[20:13] <klaxa> did you google? did you look at the manpage?
[20:13] <hendry> klaxa: yup !
[20:13] <JEEB> more like, put those -is before you start setting up the encoder/output side settings
[20:13] <klaxa> what did you google? also, if you say it doesn't work, that doesn't help
[20:14] <JEEB> otherwise ffmpeg will parse those things before an -i as input/decoder settings
[20:14] <JEEB> as I said
[20:14] <klaxa> complete ffmpeg log would help
[20:14] <JEEB> <hendry> this stuff is really rather tricky ffs <- yes it is if you have never done it before, but you have gladly ignored what I told you.
[20:15] <klaxa> hendry: sorry if we sound insulting, but it's actually kinda tiring to help if people ignore you
[20:15] <hendry> https://gist.github.com/3924250 is the log
[20:15] <hendry> JEEB: IIUC you want me to group -i together
[20:16] <hendry> A) i tried that, I get hw:0,0: Protocol not found
[20:16] <klaxa> are you running pulse by any chance? that would make things easier (less efficient though)
[20:16] <klaxa> oh wait
[20:16] <hendry> B) this is not a problem when I use huffyuv
[20:16] <klaxa> you are missing -f alsa
[20:16] <klaxa> i think?
[20:16] <klaxa> -f also before -i hw:0,0
[20:16] <klaxa> *alsa
[20:16] <JEEB> and after the video -i I would guess
[20:16] <klaxa> you didn't run my command after all
[20:17] <klaxa> oh no you did
[20:17] <klaxa> however... >Unknown decoder 'libx264'
[20:17] <JEEB> hendry, I would be surprised if it wasn't a problem in case of ffvhuff or huffyuv :) Because the inputs should be the same
[20:17] Action: hendry wrote https://github.com/kaihendry/recordmydesktop2.0/blob/master/r2d2.sh#L24 today, trying to make ffmpeg more useful
[20:18] <JEEB> ugh sameq
[20:18] <JEEB> remove that
[20:18] <JEEB> first of all it doesn't make sense with ffvhuff or huffyuv
[20:18] <JEEB> because they are lossless
[20:18] <JEEB> second of all I hear sameq has been broken for years and didn't mean what you think it does
[20:18] <hendry> my experience of huffyuv is that they are lossy
[20:19] <hendry> since the picture / colours are off
[20:19] <JEEB> they are only lossy in the sense that there's a possible RGB->YCbCr conversion going on
[20:19] <hendry> e.g. http://r2d2.webconverger.org/2012-10-21/
[20:19] <JEEB> if you need rgb, you take on a codec that can do RGB
[20:19] <JEEB> I think ffvhuff might be able to do that, or maybe utvideo
[20:19] <JEEB> pick your choice off the lossless codecs
[20:19] <hendry> JEEB: well, I'm all ears. I don't want any loss ideally. I want it to look exactly like the screen capture
[20:20] <hendry> and i want it to work across browsers :-)
[20:20] Action: hendry tries ffvhuff
[20:20] <JEEB> utvideo at least does support RGB
[20:20] <JEEB> I coded that implementation ^^;
[20:20] <Element9> durandal_1707: sorry, got carried away, trying to fix it myself. :) here's the log: http://pastebin.com/vjKs0W0p
[20:20] <klaxa> JEEB: where would i get a list of video codecs with their respective color space?
[20:21] <JEEB> <hendry> and i want it to work across browsers :-) <- you're not telling me that you want the in-the-middle capture to be playable with browsers, right?
[20:21] <JEEB> klaxa, I'm not sure if it's listed anywhere easy :s
[20:21] <klaxa> :S
[20:21] <JEEB> also there could be defaults
[20:21] <klaxa> utvideo is lossless?
[20:21] <JEEB> f.ex. libx264 is defaulted to 4:2:0
[20:21] <JEEB> yes
[20:21] <JEEB> it's similar to huffyuv/ffvhuff
[20:21] <klaxa> ah
[20:22] <durandal_1707> Element9: have you installed libtheora?
[20:22] <JEEB> just has implementations on all major OS
[20:22] <JEEB> VFW/DS for Windows
[20:22] <JEEB> QT thingy for Macs
[20:22] <JEEB> and now libavcodec dec/enc as cross-platform
[20:22] <hendry> JEEB: I don't understand what you mean by "in-the-middle capture"
[20:22] <JEEB> hendry, you don't usually throw the lossless file you capture onto the internet as-is
[20:23] <JEEB> because if you do capture with one, it's usually quick with enc/dec and not exactly small
[20:23] <Element9> durandal_1707: yes
[20:23] <JEEB> you then usually convert those to whatever you need in the end
[20:24] <JEEB> and in most cases you need to convert to 4:2:0 YCbCr (colloquially called "YUV") for web video
[20:24] <JEEB> so if you don't want any change of color in the end result, that's a "tough luck" kind of thing, but the change should be minimal if you match RGB->YCbCr and YCbCr->RGB conversion matrices
[20:25] <Element9> durandal_1707: I have there: codec.h  theoradec.h  theoraenc.h  theora.h
[20:25] <Element9>   in /usr/include/theora
[20:25] <Element9> s/there/these
[20:25] <JEEB> of course 4:2:0 YCbCr means it will have chroma subsampling, but not like anything else is really supported in browsers (Except for Chrome or Flash)
[20:25] <JEEB> Flash uses Mainconcept's decoder and thus can do up to 4:2:2 YCbCr (in both 8bit and 10bit), and unless Chrome removes features from their H.264 decoder they should be able to deal up to 4:4:4, or possibly RGB
[20:26] <hendry> JEEB: just tried both utvideo & ffvhuff. I think ffvhuff.webm is better, but still the colours are off
[20:26] <Element9> I tried configuring without theora and then it says libvorbis is missing. hmmm...
[20:27] <JEEB> > ffvhuff.webm > WhatAmIReading
[20:27] <JEEB> anyways, quality-wise they should be the same but the utvideo encoder is not threaded so prolly slower
[20:27] <JEEB> I should really get to that one day :P
[20:27] <JEEB> hendry, check your log btw
[20:27] <JEEB> or gist it
[20:28] <JEEB> that will tell you what possible colorspace conversions were made
[20:28] <durandal_1707> Element9: you will probably need to add extra flags: --extra-cflags=-I/usr/include --extra-ldflags=-L/usr/lib
[20:28] <JEEB> at least I would think that with utvideo there should be none, as I've supported all of the usual ones (RGB24, RGBA, 4:2:2/4:2:0 YCbCr)
[20:29] <hendry> JEEB: https://gist.github.com/3924250
[20:30] <Element9> durandal_1707: lets try that
[20:31] <klaxa> >-t is not an input option, keeping it for the next output; consider fixing your command line.
[20:31] <klaxa> >Stream #0:0: Video: vp8, yuv420p, 1366x768, q=-1--1, 200 kb/s, 1k tbn, 29.97 tbc
[20:31] <klaxa> looks like it gets converted to 4:2:0 YCbCr
[20:32] <JEEB> yes
[20:32] <JEEB> Stream #0:0 -> #0:0 (rawvideo -> libvpx)
[20:32] <JEEB> and instead of ffvhuff you get vp8
[20:32] <JEEB> probably because it's webm
[20:32] <JEEB> but this is weird
[20:32] <klaxa> and vorbis instead of pcm_16se
[20:32] <JEEB> it shouldn't just override like that
[20:33] <hendry> http://r2d2.webconverger.org/2012-10-21/utvideo.webm.html versus http://r2d2.webconverger.org/2012-10-21/ffvhuff.webm.html
[20:33] <klaxa> *pcm_s16le
[20:33] <JEEB> it should instead kick you for using a container that couldn't take your selection
[20:34] <JEEB> hendry, you could report this as a bug I guess? Because instead of herping a derp at you it auto-sets things to be compliant for webm. Instead you probably would have wanted to get an error that would tell you that those things don't go into webm in case you /explicitly/ set thme
[20:34] <JEEB> *them
[20:34] <Element9> durandal_1707: now I get "Symbol mangling check failed."
[20:35] <durandal_1707> what OS is that?
[20:36] <hendry> JEEB: where should I file a bug?
[20:36] <Element9> durandal_1707:  xubuntu and cross compiling for windows
[20:36] <durandal_1707> Element9: not my experize
[20:37] <Element9> durandal_1707: oh, I think I saw a flag for cross compiling. I'll try to set that one
[20:37] <Element9> durandal_1707: I can always try to comment out that on check. maybe it goes well :)
[20:37] <Element9> durandal_1707: thanks for the help
[20:38] <JEEB> hendry, trac I guess
[20:40] <JEEB> hendry, also in general I would really say that you shouldn't be encoding straight into something that you give out as-is on the web, because you won't be able to compress it as well when it also has to be realtime
[20:40] <hendry> JEEB: tbh most of this stuff if over my head
[20:41] <hendry> JEEB: so if you could spoon feed me what I should be doing, I'd be extremely grateful
[20:41] <hendry> JEEB: I'll add it to https://github.com/kaihendry/recordmydesktop2.0/blob/master/r2d2.sh#L24
[20:43] <klaxa> hendry: record lossless to a temporary raw, then re-encode for publishing on the web
[20:43] <hendry> klaxa: do you have a step-by-step example please?
[20:44] <JEEB> hendry, I'm not going to be copying your long line, but look at this, more or less, ffmpeg <inputs> -vcodec ffvhuff -acodec pcm_s16le out.avi
[20:44] <JEEB> <inputs> meaning that mumbo-jumbo on your live input devices etc.
[20:44] <klaxa> hendry: i write something up for you
[20:46] <hendry> JEEB: my mumbo jumbo is pretty standard? x11grab & alsa for sound
[20:46] <hendry> klaxa: that would be much appreciated
[20:46] <JEEB> hendry, I'm just not well versed in that but it looked more or less correct
[20:46] <JEEB> but you should understand how my example would build around your inputs :P
[20:47] <klaxa> JEEB: i'll assume utvideo is very well suited for lossless screencasts?
[20:48] <hendry> added -report logs to http://r2d2.webconverger.org/2012-10-21/ if you are interested
[20:48] <hendry> I can do this in one step with my script like so: ./r2d2.sh -d 5 -u r2d2.webconverger.org:/srv/www/r2d2.webconverger.org -c ffvhuff ffvhuff2.webm
[20:48] <JEEB> as a format it's nice enough, supports RGB as well as various YCbCr formats, but it might not be fast enough, but I'm not sure if the native library encoder that Daemon404 wrote is any faster :)
[20:49] <klaxa> libx264 ultrafast lossless it is?
[20:49] <hendry> JEEB: I'm OK with slow encoding speed as long as results look like my desktop I'm trying to capture.
[20:49] <hendry> http://r2d2.webconverger.org/2012-10-21/utvideo2.webm.html is failing the "look like my desktop I'm trying to capture" test
[20:49] <hendry> and so is http://r2d2.webconverger.org/2012-10-21/ffvhuff2.webm.html
[20:50] <JEEB> ...
[20:50] <JEEB> did you actually read ANYTHING I wrote?
[20:50] <Mavrik> O.o
[20:50] <JEEB> and did you not read why I told you reporting that one thing might be a good idea AT ALL?
[20:50] <JEEB> if you set your output file name to be something dot webm
[20:50] <JEEB> it will auto-set stuff to be compatible with it
[20:50] <JEEB> that is not ut video and that is not ffvhuff
[20:51] <hendry> JEEB: yes, I do read but i might not understand
[20:51] <JEEB> real ffvhuff or ut video DO NOT DECODE IN YOUR BROWSERS
[20:51] <hendry> JEEB: I understand you want me to report a bug about something I understand very little
[20:51] <durandal_1707> webm is vp8 + vorbis
[20:51] <JEEB> there's a reason why I set the output file name as out.avi there
[20:51] <JEEB> meh
[20:52] <JEEB> durandal_1707, I think everyone by now knows that, but the funny thing is that instead of erroring out on incorrect video/audio codecs (that are set explicitly) it will just switch them over to compatible ones :P
[20:53] <durandal_1707> gonna fix it asap
[20:53] <hendry> ok, I understand `ffmpeg <inputs> -vcodec ffvhuff -acodec pcm_s16le out.avi`... and that's it? maybe i'll just shut up now and wait for klaxa
[20:54] <durandal_1707> hendry: what ffmpeg version you are using?
[20:55] <klaxa> JEEB: what audio codec would you suggest for lossy audio?
[20:55] <JEEB> hendry, yes?
[20:55] <hendry> durandal_1707: ffmpeg version 1.0 Copyright (c) 2000-2012 the FFmpeg developers built on Sep 29 2012 11:22:50 with gcc 4.7.1 (GCC) 20120721 (prerelease)
[20:55] <hendry> durandal_1707: http://r2d2.webconverger.org/2012-10-21/ffmpeg-20121021-024453.log may help
[20:56] <hendry> durandal_1707: i'm using Archlinux
[20:56] <durandal_1707> hendry: latest ffmpeg disallows muxing other codecs for webm
[20:58] <JEEB> klaxa, depends on what your intended use case is
[20:58] <klaxa> hendry: what is your intended use case?
[20:58] <hendry> durandal_1707: so I'm not using the latest IIUC
[20:58] <hendry> klaxa: recording my desktop
[20:58] <klaxa> will you record your soundcards audio or your voice or both?
[20:59] <hendry> klaxa: ideally to container that works with HTML5 video across browsers. I don't mind 2/3 copies if it means getting it working on IOS6 safari etc
[20:59] <klaxa> i see
[20:59] <hendry> klaxa: both I guess
[20:59] <durandal_1707> hendry: this is strange 1.0 should not allow it either
[20:59] <hendry> klaxa: "-acodec pcm_s16le -f alsa -i hw:0,0" has been fine for me
[21:00] <klaxa> well yeah that's raw pcm
[21:00] <klaxa> you don't want that to be sent over the internet
[21:00] <hendry> klaxa: ok, sure, whatever is the appropriate codec for voice really
[21:00] <klaxa> i think with webm you're pretty much bound to vorbis and vp8
[21:01] <hendry> klaxa: fine, but Safari IOS6 I assume is a different beast altogether
[21:02] <klaxa> anyone know the variables for libvx? i for one don't, i don't use libvx
[21:02] <klaxa> i.e. profile, preset, good crf or qp values
[21:03] <JEEB> libvpx?
[21:03] <JEEB> unfortunately it has no such easy things like the three first ones you mentioned :P
[21:03] <JEEB> it does have constant quant'ish thing
[21:03] <JEEB> but nothing like crf
[21:08] <klaxa> D:
[21:08] <klaxa> do the default values suffice for publishing on the web?
[21:08] <JEEB> it has some vpre files for certain stuff, but no idea how good those in reality are
[21:08] <klaxa> well.. hendry will find out for us :)
[21:08] <JEEB> you are most probably best off setting a constant quant or a bit rate with libvpx
[21:10] <klaxa> https://gist.github.com/7dcccbd86fdcce3c4ced
[21:10] <klaxa> hendry: ^
[21:11] <klaxa> ah fixed small thing, second step: raw.avi -> raw.mkv
[21:12] <hendry> klaxa: https://gist.github.com/3924250
[21:12] <hendry> klaxa: first step worked, second didn't
[21:12] <klaxa> <klaxa> ah fixed small thing, second step: raw.avi -> raw.mkv
[21:13] <klaxa> reload the paste
[21:13] <klaxa> wait what..
[21:13] <hendry> -strict -2 # didn't work either as the output suggests
[21:14] <klaxa> there is no audio in the .mkv
[21:14] <hendry> klaxa: when i `mplayer raw.mkv` i hear audio
[21:15] <klaxa> wat... no there is audio in the mkv
[21:15] <klaxa> but ffmpeg doesn't... someone... why doesn't vorbis work?
[21:15] <klaxa> that's beyond my knowledge... try specifying bitrate with -ab 192k or something after -acodec vorbis
[21:16] <durandal_1707> hendry: do not use internal experimental aac encoder (it have limited quality)
[21:16] <durandal_1707> neither internal vorbis encoder (which is just not for users)
[21:17] <durandal_1707> just do not use experimental things
[21:17] <hendry> durandal_1707: can you suggest edits to https://gist.github.com/7dcccbd86fdcce3c4ced please?
[21:18] <klaxa> yeah i'm actually a novice myself
[21:18] <klaxa> i feel like i should know a lot more to be able to give advice at all
[21:19] <hendry> i've got to go to bed since it's 3AM here, i'll be connected though and I'm desperately looking to improve https://github.com/kaihendry/recordmydesktop2.0
[21:21] <durandal_1707> hendry: use copy for last -acodec
[21:23] <hendry> durandal_1707: https://gist.github.com/3924250
[21:23] <hendry> i guess thre is a problem with klaxa's https://gist.github.com/7dcccbd86fdcce3c4ced step 1
[21:24] <durandal_1707> hendry: what are you doing?
[21:24] <klaxa> >[webm @ 0x1c2a6e0] Only VP8 video and Vorbis audio are supported for WebM.
[21:24] <klaxa> i think vorbis is mandatory for webm, no?
[21:25] <hendry> ffmpeg -f x11grab -s 1366x768 -r 30 -i :0.0 -f alsa -i hw:0,0 -acodec pcm_s16le -vcodec ffvhuff -acodec vorbis raw.mkv # doesn't work: Codec is experimental but experimental codecs are not enabled ....
[21:25] <klaxa> wait... don't use -acodec twice
[21:25] <hendry> durandal_1707: trying to record my desktop
[21:25] <klaxa> remove the -acodec vorbis
[21:25] <klaxa> that's for the reencode to webm
[21:25] <klaxa> the raw is supposed to be lossless anyways
[21:26] <hendry> klaxa: oops, no different if i remove 2nd -acodec vorbis
[21:26] <hendry> so it looks like we are stuck here or is there some intermediate stage ?
[21:27] <durandal_1707> klaxa: mkv can have anything for audio
[21:27] <durandal_1707> i mean hendry
[21:27] <klaxa> yeah the initial raw should have ffvhuff video and pcm_s16le audio
[21:27] <klaxa> then re-encode with vp8 and vorbis for webm
[21:28] <mark4o> should use -acodec libvorbis not -acodec vorbis
[21:28] <durandal_1707> hendry: you cant use webm for this
[21:28] <durandal_1707> pcm_s16le is overkill, it can take lot of space, use flac
[21:29] <amendes365> good morning/afternoon/evening everyone, im dont know why ffmpeg shows 'Unknown AudioCodec: aac' even when [ffserver -formats | grep aac] recognizes it
[21:30] <durandal_1707> amendes365: you want to stream aac?
[21:31] <klaxa> ah right... libvorbis, will flac be encoded fast enough? (it's audio so i guess so?)
[21:31] <amendes365> durandal_1707: im trying to stream H.264 video with AAC audio in MPEGTS
[21:32] <durandal_1707> klaxa: if disk space is small and disk is slow but CPU is reasonaly faste use flac
[21:32] <amendes365> following one of the tutorials in ffmpeg.org
[21:32] <hendry> ffmpeg -f x11grab -s 1366x768 -r 30 -i :0.0 -f alsa -i hw:0,0 -acodec libvorbis -vcodec ffvhuff raw.mkv && ffmpeg -i raw.mkv -acodec copy -vcodec libvpx encode.webm # but my machine is too slow to play this. worried the audio is out of sync.... need slep
[21:32] <durandal_1707> klaxa: if disk is big and very fast it is irrelevant
[21:32] <klaxa> hendry, reload the gist, go to sleep, come back tomorrow with a clear mind
[21:32] <klaxa> looks like you are a bit tired
[21:33] <durandal_1707> amendes365: you need aac encoder there are many of them supported by ffmpeg
[21:33] <hendry> klaxa: if we use flac then we get the Only VP8 video and Vorbis audio are supported for WebM. problem
[21:33] <hendry> web video is crazy hard :(
[21:33] <klaxa> you are supposed to use it the other way around
[21:33] <klaxa> flac for the raw, vorbis for the encode
[21:33] <klaxa> because flac is lossless
[21:33] <klaxa> and vorbis isn't
[21:34] <amendes365> durandal_1707: i know, ffserver -formats show raw ADTS AAC, Advanced Audio Coding, libfaac AAC but when starting ffserver shows Unknown audio codec: AAC
[21:34] <hendry> klaxa: so we need an intermediate step to convert the flac audio to vorbis for the final webm ?
[21:34] <durandal_1707> klaxa: but it is irrelevant in this case becuase it just copy it when muxing into webm
[21:34] <klaxa> hmm true
[21:34] <durandal_1707> hendry: try flac if libvorbis encoding is soo slow
[21:35] <durandal_1707> hendry: basically you recored as fast as possible to keep sync and reencode and mux after recording is finished
[21:36] <durandal_1707> amendes365: formats are not codecs, use ffserver -codecs
[21:36] <hendry> durandal_1707: understand that
[21:37] <hendry> durandal_1707: guess i need to figure out the intermediate step's args after some sleep
[21:39] <amendes365> durandal_1707: thanks. ffserver -codecs showing unrecognized option '-codecs'
[21:39] <amendes365> klaxa: thank you for the help yesterday, trying mpegts now
[21:39] <klaxa> i can run ffserver -codecs
[21:40] <klaxa> the only encodable line says:  DEA.L. aac                  AAC (Advanced Audio Coding)
[21:41] <amendes365> ill try reinstalling ffmpeg
[21:41] <durandal_1707> internal encoder is experimental
[21:41] <durandal_1707> use libfdk-aac
[21:41] <klaxa> how would you install and set up an external encoder?
[21:41] <klaxa> install encoder program and re-compile ffmpeg with the apropriate flags?
[21:41] <durandal_1707> yes
[21:43] <klaxa> so it probably won't work with pre-compiled binaries from <insert arbitrary package-manager/repository here> ?
[21:44] <amendes365> i've visited different forums where usually is recommended to use the medibuntu repositories for this type of codec
[21:44] <durandal_1707> klaxa: some external encoders are not redistributable(non-free)
[21:44] <durandal_1707> which means, to use best available encoder you need to compile it
[21:45] <amendes365> is there any other audio codec i can use instead of aac for streaming H.264 video in mpegts?
[21:46] <durandal_1707> mp2?
[21:46] <durandal_1707> or aac
[21:46] <durandal_1707> *ac3
[21:47] <amendes365> durandal_1707: will try mp2. this is my second day sleeping in the lab :/
[21:47] <durandal_1707> mpegts for audio acepts: mp2,mp3,ac3 and aac
[21:48] <durandal_1707> mp2 quality is not perfect
[21:48] <durandal_1707> amendes365: what bitrate you are targeting?
[21:49] <amendes365> durandal_1707: i want a very lightweight setup to see if i can get realtime audio/video streaming from a robot
[21:49] <amendes365> robot is running on debian
[21:49] <amendes365> as long as the audio is understandable is ok
[21:50] <durandal_1707> so robot is doing encoding?
[21:51] <amendes365> yes, but im testing a scenario between two laptops first
[21:51] <amendes365> and using a webcam
[21:52] <amendes365> i was using mjpeg which worked perfectly, but it has no audio
[21:53] <amendes365> and i was considering ffmpeg to have both video and audio, otherwise i would have to do them separately
[21:54] <durandal_1707> amendes365: so you can only use mpegts?
[21:55] <amendes365> i can use anything as long as it is lightweight and can get me close to real time streaming
[21:56] <durandal_1707> well mkv wiht mjpeg + any audio codec should work tooo
[21:56] <amendes365> durandal_1707: quality is not important as long as the stream is understandable
[21:56] <durandal_1707> if bitrate can be very high you can be completly lossless
[21:57] <amendes365> ok... will try that
[22:01] <klaxa> i don't think lossless video is important if a webcam is the source...
[22:02] <klaxa> durandal_1707: do you have experience with live streaming matroska? i had quite the trouble and didn't succeed in the end
[22:06] <durandal_1707> klaxa: what trouble?
[22:06] <klaxa> mostly that ffserver wasn't able to mux the videos correctly (h264 video, aac audio, ass subtitles)
[22:07] <klaxa> actually i don't remember what went wrong...
[22:08] <klaxa> i'll try again and come back... maybe i was just too stupid the last time i tried
[22:20] <amendes365> durandal_1707: got it working with the official theora/vorbis example, will make a conf for mkv + mjpeg, which will be the container and which one is the codec?
[22:20] <amendes365> audio will be mp2
[22:21] <amendes365> test
[22:35] <durandal_1707> amendes365: mkv is container
[22:36] <amendes365> excellent
[22:36] <amendes365> durandal1707: i made a test with output from my webcam to a .mkv video, mjpg is working
[22:36] <amendes365> im working on the audio right now : )
[00:00] --- Sun Oct 21 2012