[Ffmpeg-devel-irc] ffmpeg.log.20121009

[burek]

[01:10] <zule> can anyone tell me why when creating a mp4 from a set of jpg's everything works fine until I specify a frame rate of 5 (-r 5) at which point is only encodes a small percentage of the available frames...
[01:52] <aleray> hi, is it possible to get metadata from a byte range?
[01:57] <aleray> to get the metadata of an online og resource i meant
[02:07] <burek> ?
[05:54] <Sashmo> can anyone help me out? trying to compile on mac http://pastebin.com/z7myyHCs
[06:10] <grepper> sashmo: perhaps wrong version of texi2html ?
[06:12] <Sashmo> Im trying again
[06:16] <grepper> sashmo:  guess you could always --disable-doc as a last resort
[06:16] <Sashmo> good idea
[06:17] <Sashmo> what is texi2html?
[06:18] <grepper> documentation system
[06:19] <grepper> "info ffmpeg" for example
[06:20] <Sashmo> or man ffmpeg?
[06:23] <grepper> well, I guess "2html" is a clue.  I was speaking  in general about Texinfo
[07:00] <Sashmo> can anyone point me to some tutorial for using ffmpeg and a webcam on mac?
[07:25] <grepper> dunno, on linux you would use /dev/video0 or similar
[08:31] <newenc> how to i create multi-channel wav from separate tracks?
[08:41] <klaxa> newenc: i'm not sure if ffmpeg supports it, but sox does, maybe try that out
[08:44] <newenc> I've been playing with sox for the past hour but I'm not getting very far
[08:46] <zap0> you might progress to shoes one day..
[08:50] <klaxa> if nothing works and you have a gui, you could try out audacity
[09:24] <newenc> @zap0 touche
[09:25] <newenc> well heres the deal
[09:26] <newenc> i have a bunch of prores files with 7 audio tracks. 1-6 are mono 5.1 and 7-8 are a stereo pair.
[09:27] <newenc> i want to take all 8ch and mix them into one 8ch track
[09:28] <newenc> and so far this kind of seems to work http://pastebin.com/HE5Jega7
[09:30] <newenc> so far so good.
[09:34] <newenc> the audio is all there in only one track.
[10:05] <brontosaurusrex> is there a deinterlacer  that works directly with 4:2:2 material in ffmpeg?
[10:06] <brontosaurusrex> yadif is craping out
[10:06] <brontosaurusrex> aliasing
[10:21] <divVerent> Hi... I need some small help using ffprobe
[10:21] <divVerent> how to get XML output?
[10:21] <divVerent> https://gist.github.com/3857325 - I tried ffprobe -print_format xml and failed horribly
[10:22] <ubitux> divVerent: -show_format, -show_streams, etc
[10:22] <ubitux> -v 0 is your friend too
[10:22] <divVerent> ah, thanks
[10:23] <divVerent> manpage a bit non-obvious, but good help here ;)
[10:24] <divVerent> now wondering which of the formats is best to use with shell or perl scripts... ;)
[10:24] <ubitux> flat
[10:24] <ubitux> with sep=_ + eval
[10:24] <divVerent> well, eval probably is bad due to evil characters possibly somewhere
[10:25] <ubitux> it's escaped
[10:25] <ubitux> hopefully correctly
[10:25] <divVerent> codec_long_name=H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10
[10:25] <divVerent> that is sure not correctly escaped ;)
[10:25] <ubitux> ffprobe -of flat=s=_ ?
[10:25] <divVerent> [rpolzer at nb-04 ~]$ codec_long_name=H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10
[10:25] <divVerent> bash: /: Is a directory
[10:25] <divVerent> oh, it doesn't have the flat option
[10:25] <divVerent> I thought you meant default
[10:26] <divVerent> okay, looks like I need to compile a current build then ;)
[10:26] <ubitux> 1.0 should have it
[10:26] <divVerent> okay, know what to do then
[10:27] <divVerent> system upgrade, and if it's still a 0.11 one, click the nice trolly "Mark outdated" button ;)
[10:27] <ubitux> :)
[10:28] <divVerent> yes, they have newer ffmpeg
[10:28] <divVerent> ah, flat format looks indeed somewhat shell compatible
[10:28] <divVerent> in case I find issues, is the escaping "supposed" to be shell compatible?
[10:28] <divVerent> or is that just a coincidence
[10:28] <ubitux> it's supposed to be shell compatible
[10:28] <divVerent> but it still uses dots by default as sep ;)
[10:28] <divVerent> but nice, that's the smallest issue one can have
[10:29] <divVerent> haha, I actually see a small bug already ;)
[10:29] <divVerent> streams.stream.8.index=8
[10:29] <divVerent> ...
[10:29] <divVerent> streams.stream.tags.filename="mriam.ttf"
[10:29] <divVerent> streams.stream.tags.mimetype="application/x-truetype-font"
[10:29] <divVerent> the .8. is missing
[10:29] <ubitux> 10:25:46 <@ubitux> ffprobe -of flat=s=_ ?
[10:29] <ubitux> use flat=s=_
[10:30] <divVerent> streams_stream_8_nb_read_packets="N/A"
[10:30] <divVerent> streams_stream_tags_filename="mriam.ttf"
[10:30] <divVerent> same issue
[10:30] <ubitux> i don't have that problem with git/head
[10:30] <ubitux> maybe it was fixed by saste
[10:30] <divVerent> ah, ok then
[10:30] <divVerent> right :)
[10:30] <ubitux> streams_stream_2_disposition_attached_pic=0
[10:30] <ubitux> streams_stream_2_tags_creation_time="2008-05-27 18:40:35"
[10:30] <ubitux> streams_stream_2_tags_language="eng"
[10:30] <ubitux> streams_stream_2_tags_handler_name="Apple Alias Data Handler"
[10:30] <divVerent> I of course could work around this from my code... but not sure if worth it
[10:30] <divVerent> if I can just compile from git
[10:32] <divVerent> while I am waiting for the compile... is there a way to find the number of streams from the output?
[10:32] <saste> divVerent, -show_format tells the number of streams
[10:32] <divVerent> ah, ok
[10:32] <divVerent> just asking because there MAY be already a variable streams_stream_N_whatever in the shell ;)
[10:33] <divVerent> yes, it's me and my weird corner cases that never happen anyway again ;)
[10:33] <ubitux> divVerent: if you have issues with the escaping please raise them
[10:33] <ubitux> it might even depends on the shell
[10:33] <ubitux> here are the two functions FYI: http://git.videolan.org/?p=ffmpeg.git;a=blob;f=ffprobe.c;h=2187ce2652a9cae5d7bc33d0d546aaac7d5abd30;hb=HEAD#l870
[10:35] <divVerent> looks MOSTLY good to me... except for some smaller suggestions
[10:35] <divVerent> it is probably better to use single quotes, as then only \ and ' need escaping at all
[10:36] <divVerent> \ to '\\' and ' to '\''
[10:36] <divVerent> also, this \n and \r escape doesn't work, but you probably know that
[10:36] <divVerent> shells don't interpret these in double quoted strings
[10:36] <divVerent> but none of these cause actual breakage, and changing newlines and carriage returns to \n and \r anyway may actually be for the better
[10:37] <divVerent> e.g. so one can grep the output for a prefix easily before using it
[10:37] <ubitux> mmh
[10:38] <ubitux> iirc ' escaping is a bit special with some shells :(
[10:38] <ubitux> divVerent: if you have a failing string in mind...
[10:38] <ubitux> for the \n i wanted to avoid multilines
[10:40] <divVerent> right
[10:41] <divVerent> no, I see nothing failing
[10:41] <divVerent> I don't care for csh, though ;)
[10:41] <divVerent> but for posix shells, this is all right except for the \n/\r, and these two are harmless
[10:46] <divVerent> how to use -dump_attachment?
[10:47] <divVerent> I tried -dump_attachment:0:3 filename.ttf -i filename.mkv
[10:47] <divVerent> this does extract it, but complains at the end with "At least one output file must be specified"
[10:48] <divVerent> i.e. the extraction works fine, but I want exit status 0 ;)
[10:49] <ubitux> mmh
[10:49] <ubitux> looks like a bug
[10:49] <divVerent> hm... from the source of 0.11.2 I have lying around here, nb_output_files really has to be > 0
[10:51] <divVerent> -f null -frames 0 /dev/null kind of works around ;)
[10:52] <ubitux> haha :)
[10:52] <ubitux> divVerent: care to open an issue?
[10:52] <divVerent> sure, if this is a bug, that is
[10:53] <ubitux> i believe so, doesn't look correct
[10:53] <divVerent> I see why it does this, and I see no good way to fix it... but e.g. sox has a workaround for this kind of thing
[10:53] <divVerent> it has a -n option which means "no output"
[10:53] <divVerent> if you e.g. just want to run analysis on a file
[10:54] <divVerent> my goal BTW is to try to implement automatic language based stream selection, extracting fonts, coercing fontconfig, then trying to use -vf ass
[10:54] <divVerent> from a frontend script
[10:54] <divVerent> I am unsure about the coercing fontconfig part though ;)
[10:55] <divVerent> I now have the stream ID -> language mapping working, and the font extraction too
[10:55] <divVerent> hehe, srt2ass already has been written too :P
[10:56] <ubitux> ffmpeg -i in.srt out.ass ?
[10:56] <divVerent> haha, really?
[10:56] <ubitux> sure.
[10:56] <divVerent> that's even neater...
[10:56] <divVerent> especially given I need to extract the subtitle script anyway
[10:56] <divVerent> so I can just select the subtitle stream, and let it write to the ass file
[10:56] <ubitux> ffmpeg -i in.mkv out.ass ? :p
[10:56] <divVerent> exactly
[10:56] <divVerent> apart from selecting the stream still ;)
[10:57] <ubitux> :)
[10:57] <divVerent> which is easy, as I already have the data in a variable
[10:57] <divVerent> what will be harder is making the selection a  bit more clever
[10:57] <divVerent> especially, cleverer than mplayer's
[10:57] <divVerent> which e.g. tends to lead to viewing english with english subs ;)
[10:57] <ubitux> mplayer's selection is kind of nice afaict
[10:58] <divVerent> an annoyance not really fixable without a scripting facility
[10:58] <divVerent> it is
[10:58] <divVerent> and fixing this is not really easy there
[10:58] <divVerent> what I want is something like "try jpn/eng; if it failed, try eng/-"
[10:59] <divVerent> but doing such things just doesn't fit into mplayer's option syntax, or in fact, any option syntax I have ever seen+
[10:59] <divVerent> this is more like a job for some kind of control file
[10:59] <ubitux> you should look at honoring disposition/default as well btw :)
[10:59] <divVerent> how is that set exactly?
[11:00] <ubitux> libavformat/matroskadec.c:            st->disposition |= AV_DISPOSITION_DEFAULT;
[11:00] <divVerent> no, I mean
[11:00] <divVerent> what is it used for
[11:00] <divVerent> is this for the english subs that are only meant to replace e.g. signs
[11:00] <divVerent> but not dialogue?
[11:00] <ubitux> if the stream is the default or no so...
[11:00] <divVerent> I wanna decide the default, not the author of the file ;)
[11:01] <ubitux> if you have two audio streams and the second has default bit, then it should be selected by default :p
[11:01] <divVerent> no ;)
[11:01] <divVerent> the one which I understand should
[11:01] <divVerent> but I see, if multiple matches remain, these should be preferred
[11:01] <divVerent> so this is useful e.g. for different audio encodings of the same language
[11:01] <divVerent> to pick the "better" one
[11:01] <ubitux> i was talking about the case where you have no preferences obviously
[11:01] <ubitux> (like when doing the default picking stream)
[11:02] <divVerent> sure
[11:02] <divVerent> I will keep that in mind when writing my script
[11:02] <ubitux> and yeah it could influence the selection with preferences
[11:02] <ubitux> in case of equality, like you said
[11:03] <divVerent> there is more disposition flags though...
[11:03] <ubitux> :)
[11:03] <divVerent> is there a spec? ;)
[11:03] <ubitux> avformat.h? :)
[11:04] <divVerent> nope.avi
[11:04] <divVerent> that just tells me their bitmask values
[11:04] <divVerent> for some of them, that is
[11:04] <divVerent> I am mainly interested about what "disposition/dub" is
[11:05] <divVerent> and "original"
[11:06] <ubitux> i can't tell :)
[11:56] <KING_LEE> hello again
[11:57] <KING_LEE> i had a problem with MTS files yesterday
[11:57] <KING_LEE> i solved that now but i now noticed that the recoded clips i get are all shortened by a second
[12:01] <KING_LEE> http://pastie.org/5023383
[12:01] <KING_LEE> this i my script
[12:01] <KING_LEE> http://pastie.org/5023386
[12:03] <KING_LEE> i gave cinelerra and ffmpeg a fresh install and it worked then, i have this script from the cinelerra mailing list and added the deinterlaceing
[12:11] <burek> KING_LEE, http://stackoverflow.com/questions/965053/extract-filename-and-extension-in-bash
[12:13] <burek> also "-deinterlace is deprecated, use -filter:v yadif instead"
[12:13] <burek> http://ffmpeg.org/ffmpeg.html#yadif
[12:16] <KING_LEE> i testes i few clips and i dont really see a difference but the duration is always shortened by 1 sec
[12:17] <burek> you might run your ffmpeg with -debug_ts
[12:17] <burek> to see if all the frames are being processed
[12:17] <KING_LEE> and that shortens the clips?
[12:19] <KING_LEE> oh, its even more in that longer clip
[12:20] <KING_LEE> burek: my battery runs out, can i get back to you tonight?
[12:20] <burek> ping me, and if im online ill answer
[12:20] <KING_LEE> thanks man
[12:21] <xxthink> can ffmpeg generate mkv format files?
[12:22] <burek> xxthink it should
[12:22] <burek> ffmpeg -formats
[12:22] <burek>   E matroska        Matroska file format
[12:26] <xxthink> ok
[12:26] <xxthink> I use -f mkv,
[12:26] <xxthink> it should be -f matroska
[12:26] <burek> -f matroska
[12:26] <xxthink> yes
[12:47] <kakohari> Hi all
[12:47] <kakohari> bcoudurier: did you receive my email?
[12:53] <divVerent> which of the aac encoding libraries is recommended?
[12:54] <divVerent> if I see it right, possibly good chances have libfdk-aacenc and libvo-aacenc, but which is better? ;)
[12:55] <JEEBsv> fdk
[12:55] <divVerent> ok, thanks
[12:55] <JEEBsv> vo-aacenc is barely ffaac level IIRC :<
[12:55] <divVerent> ah, good to know
[12:55] <JEEBsv> and fdk is the "fast" fraunhofer encoder
[12:55] <JEEBsv> <3
[12:55] <divVerent> and does fdk-aac have to be current git or is 0.1.0-1 good too?
[12:56] <JEEBsv> it should be in that release too, methinks (if that's the "1.0" release)
[12:58] <divVerent> and how does libfdk-aac compare to neroAacEnc?
[12:58] <divVerent> (asking while compiling it)
[12:58] <JEEBsv> should be similar or better
[12:59] <burek> divVerent recommend for what exactly
[12:59] <divVerent> burek: listening to sound of videos with crappy headphones ;)
[12:59] <burek> ed*
[12:59] <JEEBsv> the only negative point of fdk is that it's not GPL-compliant atm
[12:59] <JEEBsv> :<
[12:59] <burek> for that you can use wav too
[12:59] <divVerent> previously I was using the aac or libfaac codecs for that, but these are both crap
[12:59] <JEEBsv> but not like nero is
[12:59] <JEEBsv> and not like faac is :D
[12:59] <divVerent> sure
[12:59] <JEEBsv> yeah, faac is derp, too by now
[12:59] <divVerent> I know I am creating an illegal ffmpeg binary
[12:59] <divVerent> sue me ;)
[12:59] <JEEBsv> nah, only illegal to distro
[13:00] <JEEBsv> IIRC
[13:00] <divVerent> sure
[13:00] <JEEBsv> and only if it's GPL
[13:00] <divVerent> is it then BTW also illegal to e.g. back up my notebook to an external HDD? ;)
[13:00] <JEEBsv> lol
[13:00] <divVerent> sure I am linking in some GPL stuff
[13:00] <JEEBsv> dunno, don't care
[13:00] <divVerent> but I consider it equally evil as the nvidia driver
[13:00] <JEEBsv> I wish fdk would've not written their own license
[13:00] <divVerent> except that it actually has SOME source
[13:01] <JEEBsv> I actually wonder how the fdk's license works within the Android framework
[13:01] <divVerent> also, anything is better than "aac" ;)
[13:01] <JEEBsv> vo-aacenc is close to that
[13:01] <JEEBsv> afaik
[13:01] <burek> better in what?
[13:01] <JEEBsv> some japanese guy tested
[13:01] <divVerent> burek: "aac" codec in ffmpeg is REALLY broken
[13:01] <divVerent> even does clipping
[13:01] <burek> recommended/better is usually used with some criteria
[13:01] <divVerent> it also is way below LAME for mp3...
[13:01] <divVerent> at same bitrate
[13:01] <JEEBsv> yes, LAME is actually pretty good
[13:01] <JEEBsv> encoder-wise
[13:01] <burek> what are your goals
[13:01] <divVerent> and being at least as good as LAME/mp3 is something I'd really expect of an aac codec
[13:02] <JEEBsv> while ffaac is well... something that semi-works
[13:02] <divVerent> burek: no audiophile stuff though
[13:02] <divVerent> just getting videos to work at all on the iCrap
[13:02] <divVerent> for that it has to be aac
[13:02] <burek> my point is that each of those codecs have something they are good at
[13:02] <divVerent> yes, faac seems to be not buggy
[13:02] <divVerent> but also below LAME level
[13:02] <JEEBsv> yeah
[13:02] <burek> so, it depends what exactly do you need
[13:02] <JEEBsv> faac is one of the LC-AAC reference implementations
[13:02] <JEEBsv> (with a small wrapper IIRC)
[13:03] <divVerent> why is it so bad then? ;)
[13:03] <divVerent> it should beat LAME then
[13:03] <burek> if by the "level" you refer to a perceived audio quality, then try libaacplus
[13:03] <JEEBsv> burek: fdk has HE-AAC as well
[13:03] <burek> at the same rate you would usually use aac
[13:03] <JEEBsv> so derp
[13:03] <burek> JEEBsv, we already covered that topic
[13:03] <divVerent> burek: sure, I always used 128k though
[13:03] <JEEBsv> did we?
[13:03] <divVerent> as that makes "aac" bearable and "libfaac" somewhat good
[13:03] <burek> it doesn't correctly produce HE-AAC v2
[13:03] <burek> aka aac+
[13:03] <JEEBsv> is it incorrect or just uses more features?
[13:04] <JEEBsv> I don't think anyone got an opinion from a person that has read the AAC spec
[13:04] <burek> well, I don't know, but it's not comparable to libaacplus
[13:04] <burek> if it doesn't produce aac+
[13:04] <JEEBsv> well, it does
[13:04] <JEEBsv> it does have problems with QuickTime tho it seems
[13:04] <JEEBsv> V:
[13:04] <burek> whatever
[13:04] <burek> divVerent if you like, try libaacplus at -ab 48k
[13:04] <burek> and see the results
[13:05] Action: JEEBsv sighs
[13:05] <burek> also, compare to whatever mp3s you have
[13:05] <JEEBsv> divVerent: vo-aacenc and faac are the two reference implementations lol
[13:05] <JEEBsv> they're both relatively bad
[13:05] <JEEBsv> to terribad
[13:05] <divVerent> 13:04:35    JEEBsv | it does have problems with QuickTime tho it seems
[13:05] <divVerent> ah, that means it's out ;)
[13:06] <divVerent> because one doesn't simply install another player on the iPhone
[13:06] <divVerent> (or rather, they all are either slow or based on Quicktime...)
[13:06] <JEEBsv> dunno if the hardware decoder fails
[13:06] <JEEBsv> just quicktime fails with its he-aacv2 output. he-aac seems to work
[13:06] <divVerent> thing is, an app can only access the HW decoder through quicktime
[13:06] <divVerent> sandbox madness...
[13:07] <JEEBsv> well, I'm not sure if anyone has tested it on the actual portable stuff
[13:07] <divVerent> and I've never seen an app that decodes video on HW and audio separately (with other formats supported)
[13:07] <divVerent> it may be quite hard to do that on iCrapOS
[13:07] <JEEBsv> the quicktime I meant was the OS X app IIRC
[13:07] <JEEBsv> not even sure which version
[13:07] <JEEBsv> anyways, that should be looked into
[13:07] <divVerent> yes, but I assume Apple uses quite similar code on iOS
[13:07] <divVerent> anything else would be madness
[13:07] <JEEBsv> well, if it uses a hw decoder on the portable things
[13:07] <divVerent> it still has to demux and verify some stuff
[13:08] <divVerent> but yes, the actual decoder may even be by some other company
[13:08] <JEEBsv> I don't think there's a HW decoder ASIC on the PC versions
[13:08] <JEEBsv> yes
[13:08] <JEEBsv> the iThings have been able to decode shit QT hasn't been for ages
[13:08] <JEEBsv> ever since 3GS
[13:08] <JEEBsv> or so
[13:08] <divVerent> but compat problems TYPICALLY are at container level, because someone wrote moov with too many os or stuff like that ;)
[13:08] <divVerent> or at "let me just annoy you" level, e.g. try feeding VFR video to an iPhone
[13:09] <JEEBsv> lol
[13:09] <divVerent> it displays the first frame then stays there
[13:09] <JEEBsv> the PSP handled VFR well, surprisingly
[13:09] <divVerent> and the best part: how does it even know it's VFR
[13:09] <divVerent> at the first frame already
[13:09] <divVerent> IIRC in mov/mp4, the time base is just the denominator
[13:10] <divVerent> so for NTSC stuff, the numerator goes up by 1001 each frame
[13:10] <divVerent> which looks quite close to VFR already ;)
[13:10] <JEEBsv> it does have the timestamps in the index methinks, no?
[13:10] <JEEBsv> yeah
[13:10] <JEEBsv> the parsers have to be crazy limited to fail at VFR and so
[13:10] <divVerent> well, if it actively wants to find out, it sure can check if all frame timestamps are divisible by 1001 in the index
[13:10] <divVerent> or stuff like that ;)
[13:10] <divVerent> but why should anyone do that
[13:11] <JEEBsv> I think I've mostly just found integer overflows with timestamp parsing with regards to MP4/MOV
[13:11] <divVerent> mp4 is basically an always-VFR format due to the encoding of timestamps
[13:11] <JEEBsv> nah, it can be exactly CFR, unlike MKV
[13:11] <divVerent> true :P
[13:11] <JEEBsv> but it naturally /can/ have VFR in there too
[13:11] <JEEBsv> MKV did lately get proper X/Y stuff too I think tho
[13:11] <divVerent> mp4 can be "totally sane CFR with timestamps going up by one each frame" only if fps are integer
[13:11] <divVerent> i.e. in Europe only ;)
[13:12] <JEEBsv> nah, I didn't mean like that
[13:12] <divVerent> but yes, I know that mkv issue
[13:12] <JEEBsv> in mkv before this you couldn't even get 24000/1001 there exactly
[13:12] <divVerent> which can't even represent many frame rates
[13:12] <divVerent> at least PAL works
[13:12] <JEEBsv> yeah
[13:12] <JEEBsv> anyways, they added stuff to make proper X/Y rates possible
[13:12] <JEEBsv> for mkv4 I think?
[13:12] <JEEBsv> not finished
[13:12] <JEEBsv> it's on the spec now tho
[13:13] <divVerent> does that mean... that they didn't even fix this when rebranding a subset of mkv as WebM?
[13:13] <JEEBsv> yeah
[13:13] <divVerent> morons at google...
[13:13] <divVerent> just to troll everyone, we should invent a container format that stores timestamps as float32
[13:14] <divVerent> or float16 ("half" like GPU vendors tend to call it), which is accurate for just long enough for a single rickroll
[13:14] <divVerent> not even that, probably :P
[13:15] <JEEBsv> but at least they finally fixed that now, and now they might want to do something about preroll
[13:15] <JEEBsv> for intra refresh in H.264 and opus
[13:16] <divVerent> BTW, I noticed ffmpeg got a lot better in the last year... I just managed to actually write a script to pick streams by language and render subtitles... even using attached fonts
[13:16] <divVerent> the latter was awful XDG_ vars hacks to tell fontconfig where the fonts are, though ;)
[13:17] <divVerent> just some TheDailyWTF worthy error messages...
[13:17] <divVerent> fontconfig: Selected font is not the requested one: 'CronosPro-BoldDisp' != 'Cronos Pro Display'
[13:17] <JEEBsv> lol
[13:17] <divVerent> wonder how that can even happen... I am amazed that it found the font, but is that fuzzy matching?
[13:17] <divVerent> why do we do fuzzy matching in a font system? ;)
[13:18] <divVerent> is that an old Microsoft hack so they could remap TmsRmn into Times New Roman in Windows 3.1? ;)
[13:18] <divVerent> and fontconfig had to perfect it
[14:14] <knoch> I wrote an application using libav* libraries, the goal is to stream MPEG2-TS over RTP. To accomplish that, I create an AVFormatContext to read an input file, which is in MPEG2-TS format, then I create another AVFormatContext to remux all streams from the file, and finally, I use an AVFormatContext RTP muxer. It works fine with the VLC client but not with another RTSP client. What can I do?
[14:14] <knoch> I basically followed this : http://libav-users.943685.n4.nabble.com/Output-mpeg-ts-to-rtp-td2234066.html
[15:44] <zule> Hello, still having trouble with avconv/ffmpeg dropping frames on me if I try to specify the framerate when encoding a stream of jpg's to a mp4
[15:44] <zule> Please see the following pastie:
[15:44] <zule> http://pastie.org/private/ojuply3xjyxuxxwcyiana
[15:45] <zule> I have no idea why this is happening, the only change I make is adding "-r 5" to the command line
[16:06] <iive> zule: if ffmpeg doesn't drop frames. the video would loose sync with the audio.
[16:07] <iive> zule: are you trying to change the speed of the video?
[16:07] <zule> there is no audio
[16:07] <zule> can I prevent the drop somehow?
[16:07] <zule> this is pretty much just a fast paced slide show
[16:09] <iive> i'm not that familiar with ffmpeg...
[16:09] <iive> wait... you encode jpg's? not another video?
[16:10] <zule> jpg to video
[16:10] <zule> no audio is present or mixed in however
[16:11] <iive> try to move the -framerate before the -i
[16:13] <zule> I'll be damned
[16:13] <zule> :D
[16:13] <iive> the position of the options matter, because it may apply to different system. for example, if you put it in the decoder section, then the demuxer (reading files) would use a default value and the codec section would try to change from one fps to another, while preserving same speed.
[16:13] <zule> why would it matter, I thought I was specifying output framerate to 5, why would a jpg stream need a framerate?
[16:14] <zule> regardless that does it
[16:14] <iive> because it is abstracted as virtual container, where each jpg is a frame.
[16:14] <zule> thank you!
[16:16] <iive> i'm glad i was able to help you :)
[16:25] <zule> pwd
[16:57] <divVerent> what am I doing wrong when I get
[16:57] <divVerent>     Stream #0:1[0xa0], 0, 1/90000: Audio: pcm_s16be, 48000 Hz, stereo, 1536 kb/s
[16:57] <divVerent> from the Input section
[16:57] <divVerent> but later
[16:57] <divVerent> [abuffer @ 0x1bf9400] Setting entry with key 'sample_fmt' to value '(null)'
[16:57] <divVerent> [abuffer @ 0x1bf9400] Setting entry with key 'channel_layout' to value '0x3'
[16:57] <divVerent> [graph 0 input from stream 0:1 @ 0x1bf9880] Invalid sample format '(null)'
[16:57] <burek> paste everything at pastebin
[16:57] <burek> don't let as guess the rest
[16:57] <tonsofpcs> how do I force audio codec on input?  is it ffmpeg -acodec foo -i input -acodec bar output , with input in format foo and output in format bar?
[16:58] <burek> tonsofpcs yes
[16:58] <tonsofpcs> thanks burek :)
[16:58] <tonsofpcs> I'm doing crazy now :D
[16:58] <divVerent> https://gist.github.com/3859359
[16:58] <divVerent> burek: not much more info, though
[16:58] <tonsofpcs> ( ffmpeg -acodec ac3 -i input.aiff -acodec copy output.ac3 )
[16:59] <burek> divVerent as I can see ffmpeg has a hard time recognizing your input
[16:59] <divVerent> ffplay plays it fine though
[17:00] <divVerent> including audio
[17:00] <burek> tonsofpcs, sorry, not -acodec foo
[17:01] <burek> but -f format
[17:01] <burek> the file is a container
[17:01] <burek> that contains multiple acodec/vcodec/scodec streams
[17:01] <divVerent> burek: do I maybe need special options to read directly from a DVD ISO?
[17:01] <burek> container/format is what you want to set
[17:02] <divVerent> burek: mounting the iso and playing from there is no option, because these stupid VOBs are split at every GB
[17:02] <burek> divVerent, not sure, never used it :)
[17:02] <tonsofpcs> burek: it seems to be working...
[17:02] <divVerent> burek: haha
[17:02] <divVerent> "by not using ffmpeg"
[17:02] <divVerent> is the answer there
[17:02] <tonsofpcs> ffmpeg -acodec ac3 -i input.aiff -acodec copy output.wav    makes an output that vlc can handle...
[17:03] <divVerent> okay, assume I have the iso mounted... how would I do it then?
[17:03] <tonsofpcs> I'll try ffmpeg -f ac3 -i input.aiff -acodec copy output.wav    and compare them
[17:03] <ubitux> < divVerent> burek: do I maybe need special options to read directly from a DVD ISO? // there is still no dvd input device :(
[17:03] <divVerent> the problem is that a title consists of multiple VObs
[17:03] <divVerent> but I want o "seams"
[17:03] <ubitux> so mount -o loop :(
[17:03] <divVerent> that alone doesn't help
[17:03] <divVerent> I want the second vob follow right after the first one without any clicks, stutter or the like
[17:04] <divVerent> just like on a real DVD player
[17:04] <divVerent> wonder if catting the vobs and encoding from that does the right thing?
[17:04] <divVerent> mc
[17:05] <ubitux> it's mpeg2 so it should work
[17:05] <ubitux> you have a concat protocol for this
[17:05] <divVerent> damn... that means large changes to the encoding script... :P
[17:05] <divVerent> as I rely on e.g. ffprobe on it
[17:06] <ubitux> divVerent: working on the dvd input device is welcome :)
[17:06] <divVerent> maybe it'll rather be dvdunauthor then...
[17:06] <divVerent> which eats over 9000 MB temp space
[17:06] <divVerent> oh wait, concat: is an URL scheme
[17:06] <divVerent> that works for me ;)
[17:07] <tonsofpcs> burek: interesting.  If I use -f ac3 , the output is a WAV header followed by an AIFF header (FORM, COMM chunk, ANNO chunk, etc.).  If I use acodec, it outputs just a WAV header and data.
[17:11] <knoch> hello ubitux, have you read my problem ?
[17:11] <ubitux> knoch: i have no idea about your problem
[17:12] <divVerent> can one hardsub a DVD subtitle somehow?
[17:12] <ubitux> yes
[17:12] <knoch> do you know someone who can help me ?
[17:12] <divVerent> okay, how?
[17:12] <ubitux> nicolas added something recently
[17:12] <ubitux> just a sec
[17:12] <divVerent> also, is it possible with -filter or only with -complex_filter?
[17:13] <tonsofpcs> is there a way to properly parse byte-reversed AC3?  If I try decoding it, I get endless frame sync errors.
[17:13] <burek> tonsofpcs, why are you using -f before -i at all?
[17:13] <burek> why dont you let ffmpeg auto-recognize the content
[17:14] <tonsofpcs> burek: because the content is byte-reversed AC3 in a container that flags it as PCM.
[17:14] <knoch> thank you again ubitux
[17:14] <ubitux> knoch: i can't help you :p
[17:14] <tonsofpcs> (well, technically the container flags it as none and the pointer to the container flags it as PCM... long story :)
[17:14] <ubitux> divVerent: mmh i can't find it again :(
[17:14] <knoch> but if you know someone who can, then you helped me :D
[17:15] <tonsofpcs> (read: AIFF is a horrible RIFF implementation)
[17:15] <ubitux> knoch: i don't know any
[17:15] <knoch> it's sad
[17:15] <ubitux> divVerent: ah, got it
[17:16] <ubitux> divVerent: http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=88fc1438c693ffb7793aeb111d89775440491840
[17:16] <burek> divVerent http://blog.rot13.org/2009/10/encoding_dvd_into_hq_flash_video_using_ffmpeg.html
[17:17] <divVerent> ubitux: okay, so this really needs a complex filter.... will do this tomorrow then :P
[17:19] <divVerent> ubitux: I take it, assuming I use -filter_complex with implicit output
[17:19] <divVerent> -filter_complex FOO gets a filter appended by -filter_complex "FOO [XYZZY] ; [XYZZY] BAR"
[17:19] <divVerent> where XYZZY is an unused label
[17:20] <divVerent> with that, it'll just be a job of string appending again ;)
[20:15] <KING_LEE> hello
[20:15] <KING_LEE> burek: you here?
[20:20] <KING_LEE> i changed the -deinterlace to -v yadif and testet my script with a longer clip, it still shortens the duration, seems like it plays it slightly faster
[20:45] <KING_LEE> http://pastie.org/5025636
[20:46] <KING_LEE> the original MTS had a duration of 8:08 and it came out 7 mins and 35 sec
[20:47] <KING_LEE> http://pastie.org/5025641
[20:47] <KING_LEE> thats the script
[20:47] <burek> KING_LEE, what is the ffmpeg's cmd line?
[20:47] <burek> try running: bash -x <scriptname>
[20:47] <burek> or sh .. something
[20:47] <burek> for debug
[20:47] <KING_LEE> ok
[20:49] <KING_LEE> http://pastie.org/5025652
[20:49] <KING_LEE> like this?
[20:53] <burek> just a sec
[20:54] <burek> hm
[20:54] <burek> Input #0, mpegts, from '00009.MTS': Duration: 00:07:35.04, start: 0.481967, bitrate: 23837 kb/s
[20:54] <burek> your input is 7:35 long
[20:55] <KING_LEE> but when i play it in my videoplayer it says 8:08
[20:56] <burek> which file exactly?
[20:56] <burek> 00009.MTS ?
[20:56] <KING_LEE> yes
[20:57] <burek> try: mediainfo 00009.MTS
[20:59] <KING_LEE> gotta install that first
[21:01] <KING_LEE> what that mediainfo? it says that it cannot be trusted
[21:01] <KING_LEE> WARNUNG: Die folgenden Pakete können nicht authentifiziert werden!
[21:01] <KING_LEE>   libtinyxml2.6.2 libzen0 libmediainfo0 mediainfo
[21:01] <burek> ?
[21:01] <burek> do you have vlc
[21:01] <KING_LEE> no
[21:02] <KING_LEE> eer yes
[21:02] <burek> do you have any other tool
[21:02] <burek> you can check that file with
[21:06] <KING_LEE> ok, when i play it in vlc it says 7:35
[21:10] <KING_LEE> seems like totem is not very reliable
[21:10] <KING_LEE> the mts plays wierd in vlc
[21:10] <KING_LEE> and i cant close it
[21:10] <KING_LEE> hmm,... vlc...
[21:10] <burek> well, where did you see 8:08?
[21:11] <KING_LEE> in totem
[21:11] <KING_LEE> in the info box and on the timeline
[21:16] <KING_LEE> how wierd, when i open it in gnome mplayer it says 11:46 on the timeline
[21:18] <burek> KING_LEE, your input might be damaged
[21:19] <burek> try extracting the video with various tools
[21:19] <burek> to get uncompressed video
[21:19] <burek> and check if all those are of the same length
[21:19] <burek> you'll surely figure out that something is wrong, using one of the tools
[21:20] <KING_LEE> i am not very experienced in taht
[21:21] <KING_LEE> my main goal is getting that stuff to work in cinelerra
[21:22] <burek> what do you mean to work in cinerella?
[21:22] <burek> you cant open .MTS files in that video editor?
[21:22] <KING_LEE> no
[21:24] <KING_LEE> it assumes a raw pcm and there is no video and the audio is plain noise
[21:25] <burek> try ffmpeg -i 00009.MTS -map 0 -c copy output.mkv
[21:25] <burek> just remux it to mkv and try opening in cin.
[21:25] <KING_LEE> getting MTS into cinelerra would be even better but i have no info about that.
[21:28] <KING_LEE> http://pastie.org/5025843
[21:37] <burek> hm
[21:37] <burek> KING_LEE, it should work if it's ffmpeg, but try this instead
[21:37] <KING_LEE> cant really read ffmpeg errors
[21:37] <burek> ffmpeg -i 00009.MTS -map O.0 -map 0.1 -c copy output.mkv
[21:40] <burbas> Hi. I'm having problems when capturing data from an rtsp stream. The output I get is http://pastebin.com/BrAB4zcx . The problem is that it takes forever in order to get that output. The command I use is "ffmpeg -i rtsp://192.168.1.90/11 output.mp4"
[21:40] <burek> why don't you paste entire cmd output, not just fragments burbas
[21:41] <KING_LEE> http://pastie.org/5025918
[21:41] <KING_LEE> same thing
[21:42] <burbas> burek: Sorry, thought it was that part that was interesting
[21:42] <burbas> I'll repost the entire thing
[21:45] <burek> KING_LEE try just: ffmpeg -i 00009.MTS -c copy output.mkv
[21:46] <KING_LEE> jep
[21:47] <KING_LEE> but it does not load into cinelerra
[21:47] <burek> why not
[21:47] <burek> same errors?
[21:48] <KING_LEE> yes
[21:48] <KING_LEE> assuming raw pcm
[21:49] <burek> then your audio/video stream is either broken or unsupported by cinerella
[21:49] <burek> so, find another video editor
[21:51] <KING_LEE> hmm,...  i guess the camera bring the same output in every file but i will check a newer one
[21:51] <KING_LEE> it's not mine and we have 4 of them
[21:53] <KING_LEE> hmm, i got kinda used to cinelerra and i want something good. not just a simple editor only for putting clips together
[21:53] <burek> btw KING_LEE, it's -map 0 (zero)
[21:54] <burek> not -map O (letter 'o')
[21:54] <burek> like 0-indexed input
[21:54] <burek> 0.0 = input 0, stream 0
[21:54] <burek> so -map 0.1 means input zero, stream one
[21:54] <burek> and -map 0 means all streams from input zero
[21:54] <burek> ok?
[21:56] <KING_LEE>  ffmpeg -i 00009.MTS -map 0.0 -map 0.1 -c copy output.mkv
[21:56] <KING_LEE> ?
[21:59] <burek> or just -map 0
[21:59] <burek> if you want all streams
[21:59] <burek> it's shorter to type
[22:00] <KING_LEE> you mean        ffmpeg -i 00009.MTS -map 0 -map 0 -c copy output.mkv
[22:00] <KING_LEE> or
[22:00] <KING_LEE>  ffmpeg -i 00009.MTS  -map 0 -c copy output.mkv
[22:00] <burek> ffmpeg -i 00009.MTS -map 0 -c copy output.mkv
[22:01] <burek> -map 0 = -map 0.0 -map 0.1 -map 0.2 -map 0.3 ...
[22:04] <KING_LEE> still no cinelerra input
[22:05] <burek> try kdenlive
[22:13] <KING_LEE> i'll give it a try
[22:13] <KING_LEE> but i  would really like to stick with cinelerra
[22:14] <KING_LEE> a reasonable script to turn the MTS to something that fits into it all i need
[22:15] <burek> it's not MTS that cinellera has problem with
[22:15] <burek> it's either audio or video
[22:16] <burek> it doesn't support ac3 I guess
[22:16] <burek> since h264 is widely supported
[22:16] <burek> by most popular video editors
[22:25] <KING_LEE> well, the script i have now is ok
[22:26] <KING_LEE> the size of the output is a bit large sometimes but i guess i have to live with that
[22:38] <KING_LEE> the 0009.MTS is 1.4 gb and the 0009.mov gets up to 9.7 gb
[22:44] <KING_LEE> kdenlive looks ok but cinelerra seems to have more functions to me
[22:49] <KING_LEE> is there a way to cut down some size?
[00:00] --- Wed Oct 10 2012