[Ffmpeg-devel-irc] ffmpeg.log.20120510

[burek]

[00:00] <tibo> hi, I'm using FFmpeg for Android compiled with Neon support. I play h.264 video, and I'm wondering if the function sws_scale is optimized for Neon ?
[00:00] <maujhsn> Anybody interested in participating in a live audio streaming test that want to run using ffmpeg? You will need a vlc media player to participate.
[00:01] <burek> tibo, you might take a look at the source code
[00:01] <burek> or ask in #ffmpeg-devel
[00:02] <burek> maujhsn, why can't you test it yourself?
[00:04] <maujhsn> burek  I'd like to but only have one computer in front of me at this time!
[00:06] <maujhsn> burek  Hypothetically the ffmpeg streaming command works! Check this out: http://pastebin.com/nTjyWeFA
[00:07] <burek> what exactly do you need to test
[00:27] <maujhsn> burek You left me hanging!
[00:28] <burek> ?
[00:28] <burek> I asked you what exactly do you need to test
[00:29] <cbsrobot> burek: I need to test a beer - you in ?
[00:31] <burek> always! :D
[00:31] Action: cbsrobot is opening a beer for burek
[00:31] <cbsrobot> splop
[00:31] <cbsrobot> here you go
[00:32] <cbsrobot> cheers
[00:32] <burek> thank you :)
[00:32] Action: burek drinks entire beer :)
[00:32] <burek> yummy :))) is there more to test :)
[00:32] <cbsrobot> sure
[00:32] <cbsrobot> let me go to my beer cellar
[00:32] <burek> I so like debugging :D
[00:33] Action: cbsrobot comes back from his cellar with some more beer
[00:34] <cbsrobot> burek - you prefer chips or salty nuts ?
[00:34] <cbsrobot> for beer testing ?
[00:34] <burek> peanuts :excited:
[00:35] <cbsrobot> here you go
[00:36] <burek> thank you :)
[00:36] <burek> testing is cool :)
[00:36] <burek> I wanna be a professional tester :)
[00:36] <cbsrobot> burek: thanks a lot - I could never have done it without you !!!
[00:37] <maujhsn> Anybody interested in participating in a live audio streaming test that want to run using ffmpeg? You will need a vlc media player to participate.
[00:37] <burek> you're always welcome cbsrobot :)))
[00:38] <Flachzang> hi
[00:39] <burek> identity crisys huh? :)
[00:39] <mrmuhhh> lol no but all the nicks are registered
[00:42] <maujhsn> This chat S-U-C-K-S!
[00:42] <mrmuhhh> iam currently working on an encoder that write frames from an opengl app (frame = rgb array) into a video. therefore i try to adapt the encoding example from http://cekirdek.pardus.org.tr/~ismail/ffmpeg-docs/api-example_8c-source.html
[00:42] <mrmuhhh> is it right that only rawvideo supports RGB24 as pixel format ?
[00:44] <burek> maujhsn, relax.. nobody is forcing you to be here..
[00:45] <burek> huh mrmuhhh, I was hoping you are going to ask something about cmd line ffmpeg :)
[00:45] <burek> you might ask in #ffmpeg-devel but be very patient
[00:45] <burek> because everyone is always working on something
[00:45] <maujhsn> burek E-A-T-M-E!
[00:46] <burek> maujhsn, ok
[00:46] <cbsrobot> lol
[00:46] <mrmuhhh> hrhr ok burek :D then it is an project for tomorrow :)
[00:46] <mrmuhhh> -n
[00:46] <mrmuhhh> burek ...
[00:47] <burek> :))))
[00:47] <mrmuhhh> when i want to convert this rawvideo (rgb24) to h264 or whatever ...
[00:47] <mrmuhhh> do i have to provide the hole codec information ?
[00:47] <mrmuhhh> *whole
[00:48] <burek> I really wouldn't know.. I never used ffmpeg's libs to write my own app
[00:48] <mrmuhhh> ah i mean converting by using the cmd app
[00:48] <burek> ffmpeg -i input.raw -vcodec libx264 out.h264 ?
[00:49] <mrmuhhh> and how can i define the codec for the input video ?
[00:50] <burek> ffmpeg -f <format/muxer> -vcodec <codec_here> -i input.raw -vcodec libx264 out.h264 ?
[00:50] <mrmuhhh> ah great thx :)
[00:50] <burek> :beer: :)
[00:50] <burek> (tested)
[00:50] <mrmuhhh> yeah beer
[05:55] <tg2> is there a frame equivalent of -ss ?
[06:26] <tg2> ^
[06:41] <grepper> tg2: dunno, but as I understand it, it will be more accurate if you use -ss after the -i FILE rather than before (decodes video that way).  In case you weren't doing that ...
[07:23] <tg2> no i eant an -ss that starts on a frame-basis instead of time
[07:24] <tg2> like -frames <framecount> for converting x frames
[07:24] <tg2> but -sframe
[07:24] <tg2> start on <x> frame
[07:25] <tg2> ex ffmpeg -i infile -sframe 4000 -frames 1000
[07:25] <tg2> would convert frames 4000-> 4999
[07:33] <grepper> yes, I know
[07:33] <grepper> closest I know of would be to use a float value with -ss and use -ss after the -i FILE
[07:34] <grepper> things could have changed though, perhaps there is another way now ...
[08:04] <tg2> hm
[08:08] <spal> This outputs an video + audio file. ffmpeg -t 60 -i out.ogv -acodec copy out.ogg
[08:08] <spal> How can I ask ffmpeg to output the audio only.
[08:08] <spal> ?
[08:08] <ubitux> use -map, or -vn
[08:11] <ubitux> if you have your audio stream in position 1, add -map 0:1 (after your input file)
[08:11] <ubitux> btw, you should put the -t after your input file too
[08:52] <dodobas> yello
[08:54] <dodobas> i'm having issues transcoding  1920x1080 at 50fps to 1280x720 at 25fps
[08:54] <dodobas> acutal video is from a camera .MTS
[08:55] <dodobas> issues appear as artifacts
[08:55] <dodobas> like missing data, black squares
[08:56] <dodobas> but if i set -r 30, then process reports duplicate frames whils encoding
[08:56] <dodobas> resulting in a video without artifacts
[08:58] <mystica555> wtrange
[08:58] <mystica555> *strange
[08:58] <mystica555> can you pastebin the command you are using?
[08:59] <mystica555> and im guessing the 50fps is 50 interlaced fields?
[08:59] <dodobas> just a sec
[09:01] <dodobas> mystica555: http://pastebin.com/nyYvc1YF
[09:02] <mystica555> first, id suggest not putting mp3 audio in an mp4 container; aac or use mkv or something.. but for the video.. hmmm i dont think -vpre will do quite what you want either anymore as newer versions are using profile/preset/tune
[09:02] <mystica555> id also suggest maybe yadif instead of -deinterlace
[09:03] <mystica555> let me try some 1080i60 material to see what happens if i halve the rate to 30 using both -deinterlace and the yadif filter.. i know yadif works quite well for most things :)
[09:04] <dodobas> mystica555: tried yadif... better results, artifacts still appears
[09:04] <mystica555> hmm
[09:04] <mystica555> got a screenshot or perhaps a sample clip?
[09:07] <dodobas> mystica555: im using mplayer to view the videos... is player maybe a factor
[09:07] <mystica555> shouldn't be; it uses ffmpeg to decode internally
[09:07] <mystica555> also i use mplayer and mplayer2 here ive not found much better to play with than those
[09:07] <dodobas> ill upolad that video... just a sec
[09:08] <mystica555> ok
[09:09] <dodobas> mystica555: http://dl.dropbox.com/u/7897514/out_default_15_9.mp4
[09:15] <mystica555> hmm the only artifacts i saw were what appeared to be bad macroblock vectors at like 7 seconds in, and a few other plaecs
[09:16] <dodobas> ok, so that than :)
[09:17] <dodobas> how to fix that, and why do they appear at 25fps, but not at 50
[09:18] <mystica555> not sure :\
[09:21] <mystica555> what version of libx264 are you using?
[09:21] <mystica555> and, have you tried replacing the -vpre hq bit with perhaps -preset slow
[09:21] <mystica555> or something similar
[09:23] <mystica555> hrm my one source of 1080i is making ffmpeg segfault
[09:24] <mystica555> seems to not like the 2.0 > 5.1 sound switch coming out of a commercial
[09:30] <dodobas> mystica555: its ftp://ftp.videolan.org/pub/videolan/x264/snapshots/x264-snapshot-20120204-2245-stable.tar.bz2
[09:30] <mystica555> ok so thats pretty new
[09:30] <mystica555> try the -preset option(s) vs -vpre anything
[09:31] <dodobas> preset slow... same thing... possibly a bit less bad macroblock vectors
[09:31] <mystica555> hm
[09:31] <mystica555> to be sure, the mts file looks fine?
[09:32] <dodobas> yes... it's iterlaced... but yes
[09:33] <mystica555> howabout if you run mplayer with -vf yadif ?
[09:33] <mystica555> and frame step near the 7s point
[09:33] <mystica555> (with the . key)
[09:33] <mystica555> ive *never* seen that sort of artifacting with any of my encodes unless the source was bad :\
[09:35] <dodobas> mystica555: original is 140mb... i can upload it
[09:35] <mystica555> ok
[09:35] <dodobas> mplayer with yadif no artefacts
[09:36] <mystica555> id be curious to test some things on it here
[09:40] <dodobas> mystica555: http://dl.dropbox.com/u/7897514/00140.MTS
[09:41] <mystica555> thanks, downloading now
[09:51] <dodobas> mystica555: tnx for everything
[09:52] <mystica555> no prob
[09:54] <dodobas> it's not a problem if i need to build/run devel versions
[09:58] <mystica555> gah.. ffmpeg is now segfaulting on your source too.. wtf have i done to it
[09:59] <mystica555> hm.. well i fixed it..
[09:59] <mystica555> what was it now..
[10:05] <mystica555> removal of -deinterlace it would seem
[10:05] <mystica555> ffmpeg -threads 0 -i 00140.MTS -filter:v scale=1280:720 -vcodec libx264 -preset slow -crf 20.0 -acodec libmp3lame -ab 96k -ar 48000 -ac 2 -sn  -r 25 out_video.mkv
[10:05] <mystica555> seems to produce proper output video for me
[10:05] <mystica555> granted, i am using current (as of about 10 mins ago) versions of both ffmpeg and libx264
[10:06] <mystica555> err wait
[10:06] <mystica555> wrong commandline copuied
[10:06] <dodobas> k
[10:06] <mystica555> ffmpeg -threads 0 -i 00140.MTS -filter:v yadif,scale=1280:720 -vcodec libx264 -preset slow -crf 20.0 -acodec libmp3lame -ab 96k -ar 48000 -ac 2 -sn  -r 25 out_video.mkv
[10:07] <mystica555> was making sure -filter:v yadif actually worked; a couple weeks ago it seemed that it was indeed not working but would work with -vf yadif
[10:08] <mystica555> also checked, and the output format of mp4 also works (not just mkv)
[10:08] <dodobas> so you are video scale filter i applied differently then -s ?
[10:09] <dodobas> blah...
[10:09] <dodobas> is video scale filter applid differently then -s ?
[10:10] <dodobas> ok, that works :)
[10:10] <mystica555> just checked, it doesn't seem to make a difference either way; i just have a habit of using the video filter at a specific size, and after any deinterlacing as to produce results i know will be proper; ive never even noticed the -s command before today actually
[10:20] <dodobas> mystica555: that makes sense
[10:20] <dodobas> tnx for help
[10:20] <mystica555> no prob; i just wish i knew what ended up working different enough to produce no artifacts :)
[10:20] <mystica555> at least for me, -deinterlace caused segfaults
[10:20] <mystica555> so perhaps if it was working for you, it was buggy
[10:20] <mystica555> otherwise, *shrug*
[10:21] <mystica555> o boy
[10:21] <mystica555> erp wrong window
[10:26] <juanbobo> is it normal for max_analyze_duration to be "reached"?
[10:26] <mrmuhhh> hi! iam trying to encode some frames from a opengl rendering app into a h264 video .... the frames are in rgb, but h264 seems to require the image in yuv; is there a sample or something else how this could be converted (c api)?
[10:44] <manizzle> is there a way to take two synchronized input streams and convert them to one stream with each stream side by side?
[10:45] <juanbobo> audio and video manizzle?
[10:46] <juanbobo> manizzle: if this is what you're looking for... http://cfc.kizzx2.com/index.php/muxing-audio-and-video-with-ffmpeg/
[10:46] <juanbobo> muxing
[10:47] <manizzle> awesome cool. thanks
[10:47] <Psy-Fi> morning, can anyone help with ffmpeg configuration? I am getting undefined references on dirac functions
[10:47] <juanbobo> manizzle: no prob
[10:51] <Psy-Fi> nevermind it was my own mistake :)
[11:49] <RushPL> hi guys I need to extract audio as a wav file in my program(C API) but when I decompress an audio stream "Stream #0:1(eng): Audio: aac (mp4a / 0x6134706D), 48000 Hz, mono, s16, 39 kb/s" I only get very few packets of data when there is silence. How could I stretch this to receive all samples that should go to the wav file/sound card?
[11:58] <burek> RushPL, you can take a look at the source code of ffmpeg and see how does it do programmaticaly
[11:59] <RushPL> that's like looking for a needle in a haystack :) I am looking for straight answer as .. you can use some filter API defined in xxx.h that will produce a straightforward stream of audio samples :-)
[12:00] <RushPL> now I am reading ffplay.c but still only a few clues
[12:01] <burek> well, then wait here and be patient
[12:01] <RushPL> yes
[12:09] <RushPL> what's the relation of av_resample_init / and libswresample ?
[12:09] <RushPL> I have been using av_resample.. but ffplay uses libswresample
[12:15] <burek> mystica555, did you use latest git ffmpeg when you got seg fault?
[12:15] <mystica555> burek: yes; i even updated it to see if that fixed it
[12:15] <mystica555> and it would seem taking out -deinterlace (deprecated anyway) and using -filter:v yadif worked instead
[12:16] <burek> I see, could you please use pastebin, just to show the command that caused segfault
[12:16] <burek> or if you have time, to report that as a bug to ffmpeg trac?
[12:16] <burek> just so that we can fix it :)
[12:19] <mystica555> http://pastie.org/3888799
[12:21] <burek> mystica555, thanks :) is there any chance you could also upload a sample 00140.MTS that you used?
[12:27] <mystica555> it ends up doing it with any source.
[12:28] <mystica555> perhaps any large source
[12:28] Action: mystica555 is trying arbitrary files now
[12:28] <burek> thx :)
[12:29] <juanmabc> RushPL: i think swresample and the swscale are the new way of performing modifications, pretty new
[12:30] <mystica555> seems to be >1280x720 as the kicker.
[12:30] <mystica555> all 1080i and random sized screencaptures larger cause the segfault
[12:30] <mystica555> 1280x720 and smaller no
[12:31] <burek> can you make 5 seconds sample of 1080i that causes segfault?
[12:32] <mystica555> easy way to do this?
[12:32] <burek> ffmpeg -i input.mts -vcodec copy -acodec copy -scodec copy -t 5 output.mts
[12:32] <burek> or just ffmpeg -i input.mts -c copy -t 5 output.mts
[12:33] <mystica555> -t 5
[12:33] <mystica555> ah
[12:33] <mystica555> i just did a dd on the initial mts file.. heh
[12:33] <mystica555> 5 megs worth
[12:33] <mystica555> still does it
[12:34] <mystica555> http://dl.dropbox.com/u/7897514/00140.MTS was the original file fwiw; 130 meg or so
[12:34] <mystica555> if its still there
[12:35] <burek> can you upload that 5 megs, that still causes segfault please
[12:36] <mystica555> Omploaded '5mtest.mts.gz' to http://ompldr.org/vZG92eg
[12:36] <burek> thanks a lot :)
[12:36] <mystica555> np
[12:38] <burek> ok, I got segfault too
[12:39] <burek> although using -acodec aac instead of libfaac
[12:39] <burek> thanks again
[12:39] <mystica555> it doesnt seem to be audio codec related either
[12:39] <mystica555> i did copy, mp3, mp4
[12:39] <mystica555> erm aac
[12:39] <mystica555> the only change that removed segfault was taking out -deinterlace
[12:39] <burek> I guess it's related to mp4 muxer
[12:39] <mystica555> and as it says, its deprecated, so maybe it should go away
[12:39] <mystica555> did it in mkv too
[12:39] <burek> but it's just a wild guess
[12:39] <mrmuhhh> is someone here familiar with the C API ?
[12:40] <burek> ok, I'll create a bug report now
[12:40] <mystica555> ffmpeg -threads 0 -i 5mtest.mts  -deinterlace -s hd720 -vcodec libx264 -preset slow -crf 20.0 -acodec libfaac -ab 96k -ar 48000 -ac 2 -sn -r 25 out_video1.mkv
[12:40] <mystica555> also causes it
[12:40] <mystica555> (-sn as mkv yells as the subtitles arent in ass format)
[12:41] <RushPL> mrmuhhh: I have the same need :)
[12:42] <mrmuhhh> hrhr :D
[12:43] <RushPL> but maybe you know .. I am decoding audio with avcodec_decode_audio4 and when there is silence I get very little data out of it.. is there a generalized way to turn it into a data stream that should go to a WAV file or should I do it manually or what? :>
[12:44] <mrmuhhh> i´m really sorry started yesterday with the ffmpeg lib :/ and iam trying to encode some frames to a h264 video
[12:51] <burek> RushPL, I told you, reading ffmpeg source code is the fastest way to get the answer :) and also using google :)
[12:52] <burek> this channel is mostly related to help with command line usage
[12:52] <burek> you could also ask in #ffmpeg-devel but I doubt the developers will have time to dedicate to newbie's questions like that, which could be solved using google and reading the source code
[12:52] <burek> no offense, really
[13:17] <RushPL> burek: sure, eventually I will figure that out probably ..
[13:17] <RushPL> but you can't blame me for seeking quick answers
[13:57] <jettero> burek: saw a highlight, from the other day, re: -filter:a pan=1:c0=$i ... yeah, I sloved it, but it was more convoluted than you'd think
[14:01] <jettero> burek: http://ffmpeg-users.933282.n4.nabble.com/really-simple-pan-question-tp4617623p4617737.html
[14:10] <burek> jettero, oh yes
[14:10] <burek> I forgot to tell you that -af is not supported
[14:10] <jettero> :p
[14:10] <burek> only through -lavfi
[14:10] <burek> sorry
[14:10] <jettero> that would have saved me a couple hours...
[14:10] <jettero> no worries
[14:10] <burek> I really didn't think about it at all :)
[14:10] <jettero> whenever I get really into an ffmpeg project it gets ... confusing in spots
[14:10] <burek> I thought it was a bug in the filter
[14:11] <jettero> I'm annoyed that ffmpeg doesn't produce an error when you use -af, it just ignores it
[14:11] <ubitux> -af will be available soon
[14:11] <jettero> but besides that...
[14:11] <burek> I agree, an error/warning would be more appropriate
[14:11] <jettero> warning options accepted, but ignored, use -f lavfi
[14:11] <jettero> is ffmpeg on github or some svn pile?
[14:11] <jettero> I never really looked
[14:12] <jettero> oooo, it's in git
[14:12] <JEEB> its own git repository, yes
[14:12] <jettero> I might go source diving and look for a place to put a warning
[14:12] <jettero> doubt it, but I might
[14:13] <JEEB> if you do that, poke #ffmpeg-devel to have your patch reviewed and commented upon
[14:13] <burek> jettero, feel free to :)
[14:13] <jettero> where do people post their patches? devel list?
[14:14] <burek> yes, but you can ask michaelni or #ffmpeg-devel
[14:14] <jettero> I like best about gitorious and github that it's so easy to do a pull request, but I supppose I could just link to my patch and pull request via #ffmpeg-devel
[14:14] <burek> it's all the same
[14:14] <jettero> yeah, doesn't matter
[14:14] <jettero> kinda the point.  <3 git
[14:14] <jettero> anyway, doubt i"ll work on it.  I"m all swamped at work
[14:14] <jettero> if someone marginalizes my work or something, I'm on it
[14:15] <burek> ok :)
[14:53] <symtab> hi
[14:54] <symtab> any ideas why nothing happens: http://pastebin.com/v7jdx0KL (the conversion doesnt start, but there is no error)
[16:04] <mystica555> heh; wonder why he had '-version 3' in the commandline..maybe he meant -level
[16:04] <mystica555> or who knows
[17:42] <frogs> http://thread.gmane.org/gmane.comp.video.ffmpeg.cvs/50476
[17:43] <frogs> this breaks compilation for me     libavformat/sctp.c:230:28: error: use of undeclared identifier 'SOL_SCTP'
[23:17] <UngureanuVlad> hello, i have a question how can i get a single frame from a rtsp stream into a jpeg image with ffmpeg ?
[23:37] <burek> ffmpeg -i input.avi -vframes 1 out.jpg
[23:37] <burek> if you need a frame at 00:02:04 then
[23:37] <burek> ffmpeg -i input.avi -s 00:02:04 -vframes 1 out.jpg
[23:38] <UngureanuVlad> burek: http://pastebin.com/AyZ9CpCr do you have any ideas why i get those errors ? are from my rtsp stream or the problem is on the ffmpeg side ?
[23:38] <UngureanuVlad> burek: the 1st version was good because it is a continuous video stream
[23:39] <burek> UngureanuVlad, that's normal because when you connect to the stream, your player is missing key frame
[23:39] <burek> and receiving only delta frames
[23:40] <burek> so after some moments, it receives the first key frame and then the stream normalizes
[23:40] <burek> and no more errors :)
[23:41] <UngureanuVlad> burek: another question http://pastebin.com/WQJfS6Fv when i ffplay the stream at get some errors like "mb_type 40 in P slice too large at 62 46" have any idea ?
[23:42] <UngureanuVlad> burek: my image gets scrambled then when it shows the errors
[23:42] <burek> just at the beginning or somewhere in the middle of the streaming?
[23:42] <UngureanuVlad> in the middle
[23:43] <UngureanuVlad> when the objects filmed are moving
[23:43] <UngureanuVlad> if the image is 70% still the play is good
[23:43] <UngureanuVlad> but when most of the objects are moving the errors appear
[23:43] <burek> your h264 stream is damaged
[23:44] <burek> what is generating rtp?
[23:44] <UngureanuVlad> an IP camera
[23:44] <burek> RTP: missed 77 packets
[23:44] <burek> your network might be disrupted too
[23:45] <UngureanuVlad> the camera is connected to my home router and i`m working on the same network
[23:45] <burek> check the cables
[23:45] <UngureanuVlad> the ping is perfect no loss of packets
[23:45] <burek> also, try playing the stream in vlc
[23:46] <burek> ping has got nothing to do with this
[23:46] <burek> in vlc go to tools - messages
[23:46] <burek> set the verbosity to 2
[23:46] <burek> and see if there are errors
[23:46] <burek> if yes, disconnect your computer from the home router  and connect ip camera directly to your computer
[23:46] <burek> and see if there are errors.. if yes, the network is a problem
[23:48] <burek> if no*
[23:48] <UngureanuVlad> i`m on osx and i cannot find tools in vlc here:))
[23:48] <burek> main menu
[23:48] <burek> tools - messages
[23:49] <burek> need 2 go :( bbl
[23:50] <UngureanuVlad> bye, thx anyway
[00:00] --- Fri May 11 2012