[Ffmpeg-devel-irc] ffmpeg.log.20120807

burek burek021 at gmail.com
Wed Aug 8 02:05:01 CEST 2012


[00:01] <llogan> i don't see an audio input
[00:02] <JEEB> well, looking at the command line on the whole it's made for both audio and video
[00:02] <JEEB> so I just noted
[00:03] <ylluminate> perfect, thanks JEEB
[00:04] <ylluminate> yeah, there was no audio in this. it's a youtube upload
[00:04] <seba-> hm
[00:04] <ylluminate> i guess the dev who initially built this put in an audio component
[00:05] <ylluminate> i am curious, is there a particular incantation that is optimized for hd youtube uploads of which you guys are aware?
[00:06] <seba-> i'm having a problem with capturing video+audio from a webcam using linux, it works normally usually, but sometimes i get a stream of DTS and PTS errors
[00:06] <seba-> any ideas?
[00:07] <seba-> with errors it works
[00:07] <seba-> but the audio/video gets slowly desynced
[00:07] <seba-> when the errors don't pop out it works ok
[00:07] <llogan> ylluminate: typically you give it the best quality that is practical for you to upload since youtube will re-encode anyway
[00:07] <JEEB> ^
[00:08] <JEEB> also you usually don't go over 30fps IIRC, since youtube will just convert it to that or lower frame rate if you do
[00:09] <JEEB> if you have the bandwidth you can just do -crf 0 and have something lossless as audio
[00:09] <ylluminate> right, well i've seen some issues previously when encoding in other apps such as compressor where youtube fails miserably to reencode properly.  it has gone away in recent times with some good presets that have been now provided, however i was thinking that there may be something to take into consideration here with ffmpeg
[00:09] <llogan> or if you need something less big than lossless: ffmpeg -i input -c:v libx264 -preset <your desired preset> -crf 18 ... output
[00:10] <JEEB> or any crf from the range of 10 to 20 depending on your eyes and the end result's size
[00:12] <JEEB> up to 720p at times if your source doesn't get enough bitrate, you can try upscaling, but that usually only helps up to 720p (the bitrate rise to 6mbps from 2mbps isn't as big of a factor IIRC)
[00:12] <JEEB> also, if you have something very detailed and fast, and if the output looks bad -- you might just have to simply blur the clip
[00:13] <ylluminate> so the lower the crf, theoretically the higher the quality
[00:13] <JEEB> that is the idea
[00:13] <JEEB> it is probably the closest rate control mode that could be called "constant quality"
[00:16] <ylluminate> interesting
[00:16] <ylluminate> okay thanks for that
[00:16] <JEEB> presets control the algorithms and features used for overall compression, and crf controls the quantization
[00:16] <JEEB> the idea is to find the slowest preset that still is fast enough for you
[00:16] <JEEB> and the highest crf that still serves you
[00:16] <ylluminate> in this case a crf from 24 to 10 is not making any visual difference as they are coming from a set of jpeg images from scientific imaging
[00:17] <JEEB> of course, in this case there might some reason to use crf 0, too -- which would be lossless, and which youtube finally started to support as they updated their internal libavcodec
[00:17] <JEEB> until then they were seemingly using something from 2007 or so
[00:18] <JEEB> (they updated somewhere last year or so)
[00:18] <ylluminate> O.o
[00:19] <JEEB> nothing surprising really, using these tools and x264 as the encoder is the most optimal solution methinks
[00:19] <JEEB> (youtube did have its own H.264 encoder for a little while, but they switched to x264 quite quickly)
[00:28] <ylluminate> where can i find an explanation of the difference between baseline, main, high, etc?
[00:29] <JEEB> those are feature levels, basically
[00:29] <JEEB> baseline is for devices that have very few processing power usually
[00:30] <JEEB> http://en.wikipedia.org/wiki/H.264#Profiles
[00:30] <JEEB> not perfect, but should give an image
[00:31] <JEEB> oh, baseline doesn't have CABAC -- never noted that
[00:31] <seba-> so no ideas
[00:31] <seba-> for DTS / PTS errors?
[00:32] <JEEB> seba-, something leads to unwished timestamps I would guess? I have no experience of video capture on linux so I will not be of much help
[00:34] <JEEB> ylluminate, the default in libx264 is the high preset, and depending on your settings it gets lowered to the lowest one your used features let
[00:34] <JEEB> if you set it, it usually means you have to limit compression features for something
[00:34] <JEEB> that said, youtube doesn't need that
[00:35] <ylluminate> i was going to say, prior to your last remark: so really setting a profile wouldn't be an issues presently with youtube since it reencodes and groks it now anyway
[00:36] <JEEB> yup
[00:36] <seba-> JEEB, yes, i don't really know why sometimes it works, sometimes it doesn't
[00:36] <seba-> if it works at start it can works with zero or few errors over hours
[00:37] <seba-> but if it starts with errors, it's just a big flow of errors then
[00:37] <JEEB> ylluminate, youtube's mp4 support could still be derp, you could try using matroska (mkv) instead for example
[00:38] <ylluminate> what would matroska gain?  i wasn't aware youtube supported it
[00:39] <JEEB> simpler timestamps
[00:39] <JEEB> mp4 is a format that older parsers often can mis-parse
[00:39] <ylluminate> ah
[00:39] <JEEB> just guessing what you're getting as groks
[00:40] <JEEB> (youtube used to even have a warning for mp4 usage in case the file contained a feature that might be misparsed)
[00:40] <JEEB> and youtube pretty much supports what libavformat and friends supports I would guess
[00:40] <ylluminate> yeah, i recall something like that from some time ago
[00:47] <ylluminate> curious, is there a visual diff tool for videos that will create a differential video of two videos?
[00:47] <JEEB> you could use avisynth and various filters it has on masks etc.
[00:48] <ylluminate> hmm, interesting
[00:49] <JEEB> coupled with something that can preview the scripts (avspmod, virtualdub), you can then use it for various things
[00:55] <ylluminate> really interesting
[00:56] <saste> i wonder if a diff filter would make sense...
[00:57] <ylluminate> sure would be handy saste
[00:57] <ylluminate> i could see that used in a number of scenarios from simply seeing more easily the differences one encode to the next to many other applications
[00:57] <saste> then with some overlay+pad magic you should be able to compare img1,img2 and diff on the same image
[00:58] <ylluminate> yes
[00:59] <ylluminate> it would be interesting to have a few different styles of diffing from an outline to color inversion and clipping: http://jeffkreeftmeijer.com/2011/comparing-images-and-creating-image-diffs/
[01:01] <saste> ylluminate: => feature request
[01:02] <saste> or even better: => patch
[01:02] <saste> or send money and i'll implement it ;-)
[01:02] <ylluminate> lol
[01:03] <ylluminate> would that i had the time and money at the moment to do it!  it is very cool and useful though
[01:04] <ylluminate> shame on JEEB as he's got me going off on tangents now regardless telling me stuff ;)
[01:04] <saste> ylluminate: time and money are mutually exclusive
[01:05] <ylluminate> lol, indeed
[01:05] <ylluminate> i am curious JEEB, what kind of interesting applications of leveraging these together pop off the top of your head as you mentioned earlier?
[01:06] <JEEB> I've mostly looked at differences of clips and tried masked filtering
[01:06] <JEEB> for example, I had artifacts (haloing) mostly affect the lineart in an animated series
[01:06] <JEEB> http://img16.imageshack.us/img16/4586/13198mask.png
[01:07] <JEEB> made the active (white) mask a bit larger around itself, and applied the filter only there to fix it
[01:08] <JEEB> http://forum.doom9.org/showthread.php?t=98985 <- avisynth filter used for that stuff
[01:24] <brontosaurusrex> any clues what are the correct -profile:a names for libfdk-aacenc?
[01:24] <brontosaurusrex> like he-aac, he-aac2
[01:25] <brontosaurusrex> staring at https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/libfdk-aacenc.c
[01:25] <brontosaurusrex> doesn't make me any smarter
[01:26] <ylluminate> very interesting JEEB, i would be interested in seeing your shell incantations to see how you worked through that
[01:33] <alyawn> brontosaurusrex, look at line 370 of that file
[01:33] <brontosaurusrex> ok
[01:34] <brontosaurusrex> alyawn: yeah i tested all that, not working
[01:35] <brontosaurusrex> unless perhaps vbr mode cannot use those profiles?
[01:36] <alyawn> getting any errors, or is it simply failing silently
[01:37] <brontosaurusrex> like this http://pastebin.com/zrDLpSkg
[01:39] <ylluminate> saste: as you asked: https://ffmpeg.org/trac/ffmpeg/ticket/1615
[01:41] <llogan> saste: when "documentation" is selected in trac it auto-assigns to you.
[01:41] <saste> llogan: yes, and that's annoying
[01:41] <saste> what's that a curse?
[01:41] <saste> also note that "stefano" is not an existing account name
[01:42] <saste> mine is "saste"
[01:43] <saste> ylluminate: ok
[01:47] <llogan> saste: michael is the only one that can change that trac behavior, AFAIK.
[01:48] <saste> michaelni ^
[01:50] <michaelni> hmm
[01:50] Action: michaelni checks trac
[01:52] <michaelni> saste removed from default for docs
[01:52] <saste> michaelni: thanks
[01:52] <michaelni> Sashmo, btw "stefano" is allso default for ffprobe, what should i do there ?
[01:53] <saste> michaelni: i'd prefer not to "own" tickets
[01:53] <michaelni> ok
[01:53] <saste> i regularly check tickets in my area (ffprobe, libavfilter, docs)
[01:53] Action: michaelni curses xchats auto completion ...
[01:53] <saste> but i don't like the idea of "owning" them
[01:54] <saste> also, my account name in trac is "saste", i don't even know if we have a "stefano" or not, what's sure he's not me
[01:54] <michaelni> ok, new ffprobe tickets wont be auto assigned amynore
[01:54] <michaelni> i dont think theres a stefano :)
[01:55] <llogan> it's listed in some tickets
[01:55] <llogan> https://ffmpeg.org/trac/ffmpeg/ticket/1492
[01:55] <saste> https://ffmpeg.org/trac/ffmpeg/query?status=!closed&owner=stefano
[01:55] <saste> still he doesn't exist or he's not me!
[01:56] <llogan> is he a fork?
[01:56] <saste> yes he's the good side of me
[01:57] <llogan> i like trac so far. much better than bugzilla. that one is just annoying.
[01:58] <iive> michaelni: xchat have an option to autocomplete first the nicks that spoke most recently.
[02:00] <michaelni> iive, where do i find that option
[02:00] <saste> llogan: yes i'm fond of trac as well, don't get into the way, better than roundup as well
[02:02] <iive> preferences->input_box->nick completition sorted: [last-spoke-order] (the last option on the page)
[02:04] <michaelni> iive, thanks alot !
[02:04] <iive> it is not full proof, but it helps a lot.
[02:05] <iive> e.g. if somebody quits or somebody with similar nick spokes suddenly, it could still mess up.
[02:11] <ylluminate> perhaps you guys would want to take a look at chiliproject.  it's been really nice from my experiences so far.
[02:11] <ylluminate> anyway tanks for the chat earlier guys. l8r
[02:19] <brontosaurusrex> well, read me here: http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=95989&view=findpost&p=804484
[02:23] <brontosaurusrex> its interesting how slow this thing is
[02:31] <brontosaurusrex> or might be some compiling mistake
[02:59] <brontosaurusrex> well, night and thanks
[03:27] <juanmabc> what can you do with filters? could you kinda simulate an audacity-cli?
[03:42] <alyawn> is -muxrate with mpegts thought to be working? I keep getting dts < pcr, TS is invalid...
[04:07] <seba-> hm
[04:39] <necktie> hi. does anyone know if ffmpeg can remove cinavia protection?
[06:15] <alyawn> so does "dts < pcr, TS is invalid " necessarily mean that I've specified an incorrect muxrate for an mpegts? currently I have muxrate == bitrate == vbv-maxrate
[10:21] <twid> I converted file by remuxing to avi.. BUt i cant play it in Widows media player.. it does play in VLC... So is there any tool that allow me to detect/debug error in Video File?
[10:24] <twid_> I converted file by remuxing to avi.. BUt i cant play it in Widows media player.. it does play in VLC... So is there any tool that allow me to detect/debug error in Video File?
[10:36] <zap0> who says the video file has an error?
[10:45] <twid_> zap0 : R u asking to me.... I am thinking Windows media player might not able to read video Header..
[10:46] <zap0> then its WMP that is broken.
[10:55] <twid_> I remuxed Video file to avi... but now it doesnt play in Windows media plaer, But it does play in VLC... So i think there some problem with video header..So is there any tool available to view technical info about File? Thx...
[11:08] <MMlosh> twid_, remuxed from what?  how?  what are the codecs used in that file?   (I won't help you, but those are important info clues you are missing in your request)
[11:09] <twid_> Sorry... its code (MPEG-4)FMS4... Container is avi.... It does contain only one stream that is Video Stream no other stream in that....
[11:10] <twid_> Codec MPEG-4(FMS4)
[11:28] <An_Ony_Moose> can anyone suggest a lossless codec for use as an intermediate format video processing?
[11:28] <An_Ony_Moose> s/format/format in/
[11:29] <ubitux> animes ppl seem to use utvideo now
[11:30] <ubitux> ffv1 should do the trick as well
[11:47] <zap0> An_Ony_Moose, PNG.
[12:16] <newenc> I'm trying to encode CBR mpeg2 files but result is always VBR. i have used -b -minrate -maxrate -bufsize -bt  but still VBR.
[12:16] <newenc> how can i force car?
[12:16] <newenc> *CBR
[12:18] <An_Ony_Moose> zap0: sorry, for a single video file
[12:19] <zap0> An_Ony_Moose, wat?
[12:20] <An_Ony_Moose> and I found one, never mind :P I used  huffyuv
[12:21] <An_Ony_Moose> is huffyuv raw YUV compressed with huffman's algorithm?
[12:22] <JEEB> more or less, although huffyuv is an older one that only supports certain colorspaces. There's also ffvhuff that supports more if I recall correctly
[12:32] <zap0> conversion too-from  YUV may be lossy.
[15:03] <twid_> How do i copy only single Stream(video Stream ) from file to another file?
[15:05] <ubitux> with -map
[15:05] <ubitux> -map 0:1 to map the stream 1 from input 0
[15:12] <twid_> ubitux thx..
[15:30] <Lju2> hello, need some help with burning subtitles on windows build.. anyone?
[15:33] <ubitux> https://ffmpeg.org/trac/ffmpeg/wiki/How%20to%20burn%20subtitles%20into%20the%20video
[15:33] <Lju2> thnx, thats where i started looking
[15:33] <Lju2> but it does not work
[15:34] <Lju2> and i dont know why, would an output help?
[15:34] <ubitux> yes, please use pastebin to paste the full output
[15:35] <ubitux> *and* your command line.
[15:35] <Lju2> http://pastebin.com/CH6G7DAm ive also attached used .ass file
[15:36] <Lju2> theres a couple of runs in pastebin-ed file..
[15:36] <Lju2> last one is what is suggested on the page you provided
[15:36] <ubitux> ffmpeg -i sample.avi -vf ass=sample.ass,scale=608:256 out.avi
[15:36] <ubitux> does that work?
[15:37] <ubitux> (you can put the scale before the ass burning, depending on what you want)
[15:37] <Lju2> ok, just a sec..
[15:38] <Lju2> noup, same thing
[15:43] <ubitux> the ass format doesn't look correct
[15:44] <ubitux> mmh or maybe it is
[15:45] <ubitux> :qa
[15:45] <ubitux> oups
[15:45] <Lju2> i have even made an .ass from .srt through ffmpeg..
[15:46] <ubitux> ./ffplay -f lavfi -i testsrc,ass=test.ass
[15:46] <ubitux> i've tried that
[15:46] <ubitux> and the subtitles appear
[15:46] <ubitux> after 5 seconds
[15:46] <ubitux> (test.ass being your file)
[15:46] <ubitux> (it is valid forget what i said)
[15:49] <Lju2> interesting, i dont get anything..
[15:49] <Lju2> thats not good :D
[15:51] <Lju2> on what OS are you? i'm on w7-32bit.. that could be a problem..
[15:51] <ubitux> linux
[15:51] <ubitux> maybe a problem with fontconfig
[15:52] <ubitux> "
[15:52] <ubitux> [Parsed_ass_0 @ 005487a0] Fontconfig disabled, only default font will be used."
[15:52] <ubitux> not sure if that would be the root of the problem
[15:52] <Lju2> how can I get rid of that message?
[15:55] <ubitux> i have no idea how you're supposed to use fontconfig on windows
[15:55] <ubitux> i didn't use a windows since years
[15:55] <Lju2> ok, you've helped me enough.. i think i'm gonna try on linux :)
[15:56] <Lju2> after all, i'm gonna run it on linux
[15:56] <JEEB> fontconfig should work on windows, but you'd have to compile ffmpeg with it I'd guess?
[15:56] <Lju2> thats what i wanted to avoid..
[16:02] <jessy> hey there, i'm trying to compile ffmep with x11grab feature but in make step i'm getting bunch of errors, in debian
[16:02] <jessy> ffmpeg* :p
[16:02] <Mavrik> yay!
[16:02] <jessy> if i don't use --enable-x11grab everything is fine
[16:03] <JEEB> you might want to pastebin the errors you're getting
[16:03] <ubitux> jessy: you likely need some -dev package on debian
[16:04] <JEEB> aye
[16:04] <jessy> actual error is ; In file included from libavdevice/x11grab.c:48: /usr/include/X11/Xlibint.h: In function _XIOError:
[16:04] <Mavrik> my magic crystal ball tells me it's "libx11-dev" :)
[16:04] <jessy> any idea ?
[16:04] <Mavrik> jessy: paste full output please.
[16:04] <Mavrik> in a pastebin.
[16:04] <JEEB> ^
[16:05] <ubitux> libx11-dev looks appropriate for that header
[16:05] <ubitux> though, the configure should have break
[16:05] <jessy> http://pastebin.com/bb6MeHXJ
[16:06] <jessy> configure is just fine but i can't compile with x11grab feature
[16:06] <ubitux> apt-get install libx11-dev
[16:06] <ubitux> and try again
[16:07] <jessy> ubitux: i already have libx11-dev
[16:08] <ubitux> didn't you change the code?
[16:08] <jessy> ubitux: what you mean , i didn't change anything
[16:20] <jessy> wew i stuck, where can i find static compiled ffmpeg with x11grab support ?
[16:30] <Mavrik> hmm, why the kernel difference?
[16:31] <jessy> fflogger: thanks, checking, but i guess i already tried http://dl.dropbox.com/u/24633983/ffmpeg/index.html this one and it doesn't have x11grab feature
[16:33] <jessy> fflogger: yeah as i guess Unknown input format: 'x11grab' these builds doesn't support x11grab
[16:34] <microchip_> fflogger is a bot ;)
[16:34] <jessy> oh :)
[16:34] <jessy> lol
[16:35] <sacarasc> fflogger is the bot which flogs microchip_ every night.
[16:36] <microchip_> sacarasc: you jelly? :p
[16:37] <sacarasc> Yes, except I would have said it in non-idiot.
[17:08] <jessy> i installed linux*-dev* with apt i'm still getting same error
[17:11] <sacarasc> For X11, you'll need the xorg dev files...
[17:11] <sacarasc> Or just quit.e :D
[21:16] <necktie> hi. does anyone know if ffmpeg can remove cinavia protection?
[21:16] <JEEB> no
[21:34] <Mista_D> why multithreaded encoding produces slighthly different files (+/- 0.2%), while single threded are byte identical. Using same file/same encoding params.
[21:35] <Mista_D> Using libx264.
[21:36] <JEEB> Mista_D, are you using vbv? (maxrate/bufsize and friends)
[21:36] <Mista_D> JEEB: yes.
[21:36] <JEEB> vbv is intentionally nondeterministic with vbv
[21:36] <JEEB> with threads
[23:41] <yano> i'm trying to convert a .avi file to a .flv with a frame rate of 30 fps, and 1250k being the bitrate using CBR mode. I'm confused as to why the following command errors out:
[23:41] <yano> ffmpeg -i PV01.avi -s 640x480 -r 30 -bufsize 1250k -minrate 1250k -maxrate 1250k -f flv PV01-ffmpeg.flv
[23:42] <yano> and I get the error message: Error while opening encoder for output stream #0.0 - maybe incorrect parameters such as bit_rate, rate, width or height
[23:48] <yano> i can get around the error if i specify -vcodec libx264 but then ffmpeg doesn't acknowledge the value i set with -r
[23:48] <yano> or the bitrates i set
[00:00] --- Wed Aug  8 2012


More information about the Ffmpeg-devel-irc mailing list