[Ffmpeg-devel-irc] ffmpeg.log.20130902

Tue Sep 3 02:05:01 CEST 2013

[00:07] <Guest55224> I'm having some trouble compling 0.10.8 -- it's related to not finding libx264.a (which is clearly present).  The error is at :  http://pastebin.com/nZeSzda9
[00:07] <Guest55224> This is on CentOS 5.9 64-bit
[00:08] <klaxa> 0.10.8 is horribly outdated
[00:09] <Guest55224> I know that, but our internal devs need this version (for whatever reason).
[00:11] <Guest55224> The installation process in x264 is not standard - you have to select from several... lib-static lib-shared cli install-lib-dev install-lib-static install-lib-shared install-cli.  Install "all" doesn't do much.  I don't understand why they do this -- and I wonder if I'm missing something from the installation.   The pastebin shows that the library is clearly there.
[00:17] <braincracker> http://vimeo.com/8143217  impress the pharaoh with iphone ;>>
[04:01] <davesa> Hi, does ffmpeg by default enable vaapi for hardware acceleration? or i have to compile it with --enable-hwaccel=h263_vaapi --enable-hwaccel=h264_vaapi
[04:01] <davesa> --enable-hwaccel=mpeg2_vaapi --enable-hwaccel=mpeg4_vaapi? or this old?
[04:28] <davesa> please can anyone help me out with ffmpeg on ubuntu
[04:32] <grepper> Its better to just ask your question and hang around. If someone can answer they will.
[04:33] <grepper> maybe I missed it and you already did though
[04:36] <davesa> yes
[04:36] <davesa> i will resend
[04:36] <davesa> i am newbie to ffmpeg
[04:37] <davesa> and trying to do encoding anf decoding, but i was thinking of using hardware acceleration
[04:37] <davesa> is that possible?
[04:37] <davesa> i read online everywhere no complete information
[04:37] <davesa> and some says we have to build with --enable-vaapi
[04:38] <davesa> and some says no need with new ffmpeg as it use the libav libraries
[04:38] <grepper> I think so, using vdpau (for my nvidia)
[04:38] <davesa> hmmm
[04:38] <davesa> because i have i7 4770k
[04:38] <davesa> and it support intel quick sync
[04:39] <davesa> and has intel graphic card.
[04:39] <davesa> do you think i can enable ffmpeg to use intel quick sync new technology
[04:40] <grepper> dunno
[04:40] <davesa> intel media sdk for linux has been removed from there site, i was thinking of looking at it to see if anything can be done
[04:41] <grepper> hm, I was thinking of mplayer
[04:41] <davesa> but when you use vdpau , is it only for nvidia?
[04:42] <grepper> yeah, I think vdpau is nvidia. but I really don't think gpu encoding with speed things up based on some things I've randomly read. I'm not expert and I've never really tried. My core i5 is plenty fast enough :)
[04:42] <davesa> i got lost when i went and read online
[04:42] <davesa> heheehe
[04:42] <klaxa> GPU decoding is... well not really usable as it is
[04:42] <davesa> its really ot much info,
[04:43] <davesa> what about hardware acceleration?
[04:43] <klaxa> with OpenCL the GPU can be used for so-called "lookahead-threads" with x264
[04:43] <klaxa> but that's all i know
[04:43] <davesa> aha
[04:43] <klaxa> no idea what a lookahead-thread is actually
[04:44] <davesa> i will google that
[04:44] <davesa> :)
[04:44] <davesa> do you know how to enable fmpeg or x264 to use hardware acceleration?
[04:44] <klaxa> hardware acceleration for encoding? forget it
[04:45] <klaxa> hardware acceleration for decoding? it's possible, but it's also pointless
[04:45] <klaxa> at least if you are doing an encode that is sophisticated and takes a while
[04:46] <klaxa> in that case decoding of the frame is not really worth optimizing afaik
[04:47] <davesa> ok, i start reading about lookahead x264
[07:13] <vulture-> for a 50 fps video, why would the video decoder's context time_base num=1 den=100 ?
[07:14] <vulture-> seems like for most .avi's I've encoded myself it's correct (e.g. 1001 / 30000) but for mp4's and youtubes its more like twice what it should be
[09:15] <bigbox30> I need help:
[09:15] <bigbox30> between Lavf55.12.102 and Lavf55.15.100 MKVs muxed with FFMPEG stopped playing on my standalone.
[09:15] <bigbox30> any idea what changed and how i switch it back?
[09:16] <saste> bigbox30, are you able to git bisect in order to find the flawed commit?
[09:17] <bigbox30> not really no.
[09:17] <bigbox30> i mean possibly.
[09:17] <bigbox30> i don't really want to recompile it
[09:17] <bigbox30> i'm just a user.
[09:18] <saste> what's you standalone player?
[09:18] <saste> *your
[09:18] <bigbox30> LG bluray player.
[09:18] <bigbox30> it dosn't like compressed MKV headers
[09:18] <bigbox30> but ffmpeg has never been a problem
[09:19] <bigbox30> i basically batch remux all my mky using ffmpeg
[09:19] <bigbox30> then all of a sudden this weekend it stops playing them
[09:19] <saste> check the log if there is something which may have been affected MKV remuxing
[09:20] <bigbox30> where? this is the only change log i see.
[09:20] <saste> but the fastest way to tell is probably to git bisect, since you are the only one who can test the problem
[09:21] <bigbox30> well i wondered if maybe something like reserve_index_space had now been made default.
[09:25] <felixhandte> Anyone here? I'm having a problem mapping streams.
[09:26] <bigbox30> well i'm here but very much on the needing help side of things
[09:26] <bigbox30> not so much with the giving
[09:26] <felixhandte> Specifically, DaVinci Resolve puts out a
[09:26] <felixhandte> he
[09:26] <felixhandte> heh*
[09:30] <Mavrik> felixhandte, what seems to be the issue
[09:33] <felixhandte> so, I'm exporting from Resolve into a .mov
[09:33] <felixhandte> and
[09:35] <felixhandte> resolve's mov has 3 streams, the video (DNxHD 185), audio (pcm_s24le), and a weird third stream, marked tmcd
[09:35] <felixhandte> so
[09:35] <felixhandte> i'm trying to replace the audio in that clip with the separate audio recording
[09:36] <Mavrik> ah
[09:36] <felixhandte> with, e.g., ffmpeg -i RESOLVE_CLIP.mov -i AUDIO.wav -map 0:0 -map 1:0 -vcodec copy -acodec copy OUTPUT.mov
[09:36] <felixhandte> and
[09:36] <Mavrik> which part of it doesn't work? can you paste command line and output to a pastebin?
[09:37] <felixhandte> sure
[09:39] <felixhandte> ... one sec
[09:42] <felixhandte> Mavrik: http://pastebin.com/yA8yS6g3
[09:43] <Mavrik> ok, I see it
[09:43] <Mavrik> what's the actual issue?
[09:43] <felixhandte> well
[09:43] <felixhandte> though i'm not asking for it
[09:43] <felixhandte> you can see, in the output file
[09:43] <felixhandte> at the bottom
[09:43] <felixhandte> that i'm getting this third pseudo-stream
[09:43] <felixhandte> "stts"
[09:43] <felixhandte> which
[09:44] <felixhandte> normally i wouldn't care about
[09:44] <felixhandte> but Avid is pitching a fit if it's there
[09:44] <Mavrik> hmm, looking at the mov spec
[09:44] <Mavrik> it seems like an atom you'd need because you have raw wav in it
[09:44] <Mavrik> it's a time/sample mapping
[09:45] <felixhandte> ah
[09:45] <Mavrik> but it's wierd it doesn't show up in source
[09:45] <felixhandte> ok
[09:45] <felixhandte> now, if i then do ffmpeg -i OUTPUT.mov -vcodec copy -acodec copy OUTPUT_NEW.mov
[09:46] <felixhandte> that stream disappears
[09:46] <felixhandte> so
[09:46] <Mavrik> looking at the source
[09:46] <felixhandte> i can work-around that way
[09:46] <felixhandte> but
[09:46] <felixhandte> preferably i wouldn't have to
[09:46] <Mavrik> it seems ffmpeg always writes the stts atom
[09:46] <Mavrik> hmm
[09:47] <Mavrik> felixhandte, I'd say that's something that goes on ffmpeg bug tracker :)
[09:47] <felixhandte> heh
[09:47] <felixhandte> cool
[09:47] <Mavrik> it looks like a bug when writing sttc
[09:47] <Mavrik> probably wrong stream mapping or something
[09:47] <felixhandte> are you going to file it, or shall i?
[09:48] <felixhandte> (i'd be happy to, but i'd have to figure out how)
[09:48] <Mavrik> sorry, at the office now, don't have time to investigate it :/
[09:48] <felixhandte> ok
[09:48] <felixhandte> no problem
[09:48] <Mavrik> if you have time you can post on trac.ffmpeg.com
[09:49] <Mavrik> http://ffmpeg.org/bugreports.html
[09:49] <felixhandte> alright
[09:49] <felixhandte> i'll see if i can do that
[09:50] <felixhandte> i gotta get this stuff into avid first though :(
[09:50] <felixhandte> many thanks for your help!
[10:57] <yaron> Hi, l like to have a little advice. My goal is it to write a rtp stream to a local filesystem als hls "stream" just by "copy" the data. Under OSX it worked by using "ffmpeg -i rtp://@:5004 -qscale 2 -mbd 2 -scodec dvbsub -flags -global_header -hls_time 10 -hls_list_size 6 -hls_wrap 10 -start_number 1 mystream.m3u8" fine. But i struggle since several days to do the same thing under iOS.
[11:00] <yaron> As fare as I unterstand, there is no ios code which can give me the same interface like the OSX ffmpeg binary right ?
[12:15] <wodim> hello. i'm using -itsoffset -13.5 to synchronise a/v in a .mkv file. unfortunately this means the beginning of the movie has no video at all. how do i fill the beginning of the new movie with black frames?
[13:40] <bigbox30> anyone know about any changes to libavformat between Lavf55.12.102 and Lavf55.15.100 that would stop MKV working on my hardware player?
[13:42] <durandal_1707> no
[13:43] <bigbox30> :(
[13:43] <bigbox30> is there a changelog at all
[13:43] <durandal_1707> yes there is
[13:43] <bigbox30> ...
[17:16] <XwZ> does anyone have a problem with closing an rtsp link ? I close the link with avformat_close_input and the teardown message is sent but the camera don't release the slot
[17:16] <XwZ> do you have an idea ?
[17:17] <XwZ> I use avcodec_close and avformat_close_input to clean everything
[18:26] <Renich> hey guys
[18:27] <Renich> http://paste.fedoraproject.org/36539/37813921/
[18:27] <Renich> I want to record a screencast and join two audio sources. Basically, alsa_output.usb-Burr-Brown_from_TI_USB_Audio_CODEC-00-CODEC.analog-stereo.monitor and alsa_input.usb-Burr-Brown_from_TI_USB_Audio_CODEC-00-CODEC.analog-stereo
[18:27] <Renich> can anybody help me out with this?
[18:28] <durandal_1707> what are those files?
[18:28] <Renich> durandal_1707: pulseaudio sources
[18:29] <Renich> durandal_1707: http://paste.fedoraproject.org/36540/13781393/
[18:30] <durandal_1707> you want to concatenate them or?
[18:30] <Renich> durandal_1707: yes, I do
[18:30] <durandal_1707> you record them into same codec and container and later use concat demuxer
[18:31] <Renich> durandal_1707: I am making a screencast on pianoteq; a synth piano which rocks, and I need to record that (software generated) and my voice (through card)
[18:31] <Renich> durandal_1707: ok, how do I do that?
[18:32] <durandal_1707> how to do what exactly?
[18:33] <Renich> durandal_1707: well, record the audio into independent files and concat them after
[18:33] <Renich> durandal_1707: can ffmpeg save the two audio streams into separate files and the video into another one?
[18:34] <durandal_1707> yes
[18:34] <Renich> durandal_1707: ok, how do I do that exactly?
[18:34] <durandal_1707> it is explained in documentation
[18:34] <Renich> durandal_1707: ok
[19:55] <elfinit> Hello
[19:58] <elfinit> I am looking at https://github.com/dickontoo/omxtx/blob/master/omxtx.c. At line 386 the codec is set to CODEC_ID_H264. Now, I want to specify crf option of H264. I think i have to use X264Context, but I dont know how to access this structure from the general AVCodec structure.
[20:04] <durandal_1707> elfinit: you use av_opt_set
[20:08] <elfinit> durandal_1707: I assume the property name is just crf
[20:11] <durandal_1707> elfinit: i assume its mentioned in source code
[20:12] <durandal_1707> you can just use AVDictionary and set key = value as options
[20:13] <durandal_1707> or use av_opt_set_double() for crf
[20:35] <kode101> ayo
[20:35] <kode101> why is this reducing my video quality so much
[20:35] <kode101> 'ffmpeg','-i',tmpdir+'tmp-s.ts','-filter:v','setpts='+str(videostretch)+'*PTS',tmpdir+'slow.ts','-loglevel','quiet'
[20:37] <kode101> wait no nevermind
[20:37] <kode101> how can i retain quality when stripping audio from a video
[20:37] <kode101> oh doi
[20:38] <kode101> vcodec copy
[20:42] <elfinit> hmm, av_opt_set_double with name = "crf" applied to AVCodecContext does not have any effect
[20:43] <durandal_1707> elfinit: you apply it to encoder private context not AVCodecContext
[20:43] <durandal_1707> elfinit: see doc/examples/*
[20:44] <durandal_1707> anyway av_opt_set_dict is the most generic way to set codec options
[20:44] <kode101> why is my video blocky when i use -filter:v and setpts=0.5
[20:45] <durandal_1707> blocky?
[20:45] <kode101> yeah it looks all mpeggy?
[20:45] <kode101> i have a really clean ts thats 3 seconds long, i want that but 6 seconds long
[20:45] <durandal_1707> i still dunno what you are talking about
[20:45] <kode101> whats the best way to slow the video down at the highest possible quality
[20:46] <kode101> a video that has no audio
[20:46] <durandal_1707> setpts=0.5 certainly is not solution
[20:46] <kode101> haha
[20:46] <kode101> i see!
[20:46] <kode101> :)
[20:46] <kode101> how should I approach slowing the video down so?
[20:48] <zap0> ffmpeg -r 2 -i cock.avi balls.avi
[20:48] <kode101> i like your style zap0
[20:48] <zap0> try it.
[20:50] <durandal_1707> kode101: did you read setpts documentation?
[20:51] <kode101> is cock the input file there zap0
[20:52] <zap0> yes
[20:53] <kode101> what does -r refer to
[20:53] <kode101> 'rate'?
[20:53] <kode101> no durandal_1707  i did not i just been googling ffmpeg slow down video
[20:55] <durandal_1707> kode101: you really should, there are examples to slow down or speed up
[20:56] <zap0> do you think guessing things by choose a word that starts with teh same letter is really a good way to do things?
[20:57] <kode101> yeah if googling ffmpeg -r returns nothing and the guy who told me to use rate is right there
[20:57] <kode101> mr cock and balls
[21:00] <zap0> why are you using an advertising companies advert listings as a technical reference.   do you watch the shopping channel when you want know something?
[21:01] <kode101> yes like what to purchase
[21:01] <kode101> such as swords or jewels
[21:01] <kode101> anyway my main deal is, having looked at the setpts documentation now
[21:02] <zap0> mmmm... i wonder how many types of fish there are in the black sea... I know!  i'll turn on the shopping channel.
[21:02] <kode101> you have lacerated me by using your wit
[21:03] <zap0> i'm sure there is a remedy for that, on teh shopping channel.
[21:03] <durandal_1707> zap0: no you would contact wikipedia
[21:03] <kode101> durandal, what am i neglecting here:
[21:03] <zap0> wikipedia...  a blog setup by a pornographer.   what a great reference!
[21:04] <kode101> ffmpeg in.ts -filter:v setpts='0.5*PTS' out.ts
[21:04] <kode101> apart from the -i at the start ;)
[21:06] <kode101> can i fit a -vcodec copy in there?
[21:06] <zap0> there are more people failing with setpts stuff than successeeding.   there must be a reason.
[21:06] <durandal_1707> kode101: no
[21:07] <durandal_1707> kode101: 0.5 speeds up
[21:08] <kode101> oh yeah apart from that aswell, the point is that the result comes out very artefacted
[21:08] <kode101> while the input file, though too fast, is much cleaner
[21:08] <kode101> so as i understand i am encoding it in a lossy manner but don't know how that relates to setpts
[21:09] <kode101> since it's one line
[21:09] <kode101> ffmpeg -i in.ts -filter:v 'setpts=2*pts' out.ts
[21:09] <kode101> it slows it down just fine, just the result is very artefacted
[21:10] <durandal_1707> because default mpegvideo paramters are retarded
[21:10] <kode101> haha boosh there it is
[21:11] <durandal_1707> anyway this is doing transcoding...
[21:11] <kode101> ok ok fresh i'm down, so i'm not specifying a decent enough transcode setting for the setpts command?
[21:12] <durandal_1707> no, what are you trying to archieve(slow down) is transcoding
[21:13] <durandal_1707> there is no way to do same with -c:v copy
[21:13] <durandal_1707> so you are just doing this for fun, or what?
[21:14] <kode101> yeah a project
[21:14] <durandal_1707> well try adding '-q 0' after filter params
[21:15] <zap0> -r slows things down.
[21:15] <kode101> zap0 slows things down
[21:15] <kode101> *audience goes wild
[21:15] <zap0> -q [low number]   make better quality
[21:15] <durandal_1707> -r have documentation
[21:16] <durandal_1707> -r rate             set frame rate (Hz value, fraction or abbreviation)
[21:16] <zap0> or you could keep guessing at setpts
[21:16] <durandal_1707> there is nothing guessing in setpts
[21:17] <durandal_1707> you just need ot understand  presentation timestamp and stream time base concept
[21:19] <kode101> ok cool
[21:19] <kode101> can i express the -r value as a speed relative to the input file
[21:19] <kode101> -r [half speed]
[21:19] <kode101> or do i need to calculate that first
[21:19] <kode101> via querying its framerate
[21:26] <zap0> query
[21:26] <zap0> -r is a set value (as far as i know)
[21:27] <zap0> kode101,  http://ffmpeg.org/ffmpeg.html#Generic-options   section 5.5 Video Options.      2nd item
[21:29] <kode101> hmm
[21:30] <kode101> durandal_1707 how would i include -q 0 there
[21:30] <kode101> -filter:v -q 0?
[21:31] <kode101> ffmpeg -i in.ts -filter:v -q 0 'setpts=2*pts' out.ts
[21:31] <zap0> kode101,  -q 0  is a single param   you've stuck it inside  the   -filter:v '...'    which is a single param
[21:32] <zap0> move -q to  before  -filter
[21:32] <kode101> and the 0?
[21:32] <DuelShark> hi all - glad to see this channel exists! don't suppose anyone has a tip on making ffmpeg as fast as possible. Using preset ultrafast currently and 8 threads, any help appreciated
[21:32] <zap0> like this       ffmpeg -i in.ts -q 0  -filter:v  'setpts=2*pts' out.ts
[21:32] <zap0> kode, yes, and the 0
[21:32] <kode101> safe
[21:32] <kode101> trying it now
[21:33] <leandrosansilva> Hello to all. I'm having some problems decoding h264 streams. In fact in my application (I'm using ffmpeg-1.2.3 on a C++ application on Ubuntu Linux 64-bit)  I decode, simultanously 4 different streams. I'm often having errors to decode these streams. They come from high resolution IP Cameras:
[21:33] <leandrosansilva> http://pastebin.com/NrgGxSg3
[21:33] <zap0> -filter:v  'setpts=2*pts'     <-- all that is a single parameter,  don't stick things inside it ;)
[21:34] <kode101> ohhhh
[21:34] <leandrosansilva> (the transport is rtsp over tcp)
[21:34] <kode101> thanks zap0
[21:35] <kode101> is there a quality& below 0
[21:35] <kode101> *checks around to make sure no police are nearby
[21:35] <zap0> kode101  there is likely no quality below 2 for most encoders.
[21:35] <kode101> the premium quality...
[21:35] <kode101> ultimate lossless...
[21:35] <zap0> so 0,1,2  are (for most encoders)  the same.
[21:39] <kode101> safe
[21:39] <kode101> it looks great now
[21:39] <kode101> thanks yall
[21:39] <zap0> :)
[21:44] <kode101> so if i have a .ts
[21:44] <kode101> whats the highest quality mp4 encoding i can give it
[21:45] <durandal_1707> lossless with x264 encoder
[21:45] <kode101> what's that look like
[21:45] <kode101> command wise
[21:48] <kode101> ffmpeg -i input -c:v libx264 -preset veryslow -qp 0 output.mkv
[21:48] <klaxa> if you want to wait forever, use veryslow
[21:48] <klaxa> if you want to wait forever and 3 days, use placebo
[21:48] <klaxa> but yeah that'd probably be lossless
[21:49] <klaxa> but it will probably be even larger than the .ts
[21:50] <durandal_1707> preset is meaningless with lossless
[21:50] <klaxa> not really afaik
[21:50] <durandal_1707> and lossless should be faster than placebo
[21:50] <klaxa> but i might be wrong
[21:51] <kode101> ah trying ultrafast
[22:34] <Adrellias> hey guys quick question
[22:35] <Adrellias> when i try and use x264 with ffmpeg with lossless_ultrafast
[22:35] <Adrellias> i get a error where it cannot find the preset
[22:35] <Adrellias> i also do not see the preset in /usr/share/ffmpeg
[22:36] <klaxa> just use -preset ultrafast -qp 0
[22:38] <Adrellias> klaxa: ultrafast is also not in /usr/share/ffmpeg
[22:38] <Adrellias> tried both
[22:38] <klaxa> what version are you running?
[22:38] <Adrellias> could you paste bin your ultrafast preset for me ?
[22:39] <Adrellias> let me check
[22:39] <klaxa> the use of -vpre is deprecated
[22:39] <klaxa> ultrafast is a preset of x264
[22:39] <klaxa> read the topic please :/
[22:39] <Adrellias> media-video/ffmpeg-1.0.7
[22:39] <Adrellias> x264 use flag is enabled
[22:39] <klaxa> what is your command line?
[22:39] <klaxa> can you pastebin what you execute alongside with the complete output of ffmpeg?
[22:40] <Adrellias> yep gime a second
[22:40] <jned> When I try: ffmpeg -f v4l2 -s 320x240 -r 25 -i /dev/video0 -f alsa -ac 1 -i hw:0 http://localhost:8090/feed1.ffm I get TCP connection to localhost:8090 failed: Connection refused
[22:41] <klaxa> did you start ffserver?
[22:42] <Adrellias> klaxa: http://pastebin.com/raw.php?i=Fsf2yhv2
[22:42] <klaxa> <klaxa> just use -preset ultrafast -qp 0
[22:42] <klaxa> so the whole command line would be:
[22:42] <Adrellias> klaxa: bleh copied the wrong line
[22:42] <klaxa> ffmpeg -f alsa -ac 2 -ab 192k -i pulse -f x11grab -s 1366x768 -r 30 -i :0.0 -acodec pcm_s16le -vcodec libx264 -preset ultrafast -qp 0 -threads 0 test.mkv
[22:42] <jned> klaxa, thx man, now its running, so how do I view it? When I goto localhost:8090/feed1.ffm it says "This fee is already being received"
[22:43] <klaxa> that is defined within ffserver
[22:43] <klaxa> the config file of ffserver
[22:43] <Adrellias> klaxa: http://pastebin.com/raw.php?i=hv7JbDja
[22:43] <klaxa> don't use -vpre
[22:43] <Adrellias> klaxa: so just preset ?
[22:44] <klaxa> yes
[22:44] <Adrellias> ahhh
[22:44] <Adrellias> *fp*
[22:44] <Adrellias> sorry
[22:44] <Adrellias> Thank you and sorry for not rtfs
[22:44] <klaxa> heh :P
[22:48] <jned> I don't get how to view the stream once its playing
[22:49] <jned> ffplay http://localhost:8090/feed1.ffm give me error:  HTTP error 404 Not Found http://localhost:8090/feed1.ffm: Input/output error
[22:49] <klaxa> can you pastebin your ffserver.conf?
[22:50] <jned> klaxa, I didn't edit it, should be default
[22:50] <jned> 1 sec
[22:52] <jned> klaxa http://pastebin.com/0wqWaVtz
[22:53] <klaxa> you should be able to play http://localhost:8090/test1.mpg
[22:54] <jned> klaxa, thx
[22:55] <jned> so you don't have to write any html with ffmpeg, it streams to internet on its own
[00:00] --- Tue Sep  3 2013