[Ffmpeg-devel-irc] ffmpeg.log.20121025

[burek]

[06:16] <mancha> hi folks. i am getting an error when i try to convert to -acodec libmp3lame -ab 96000 -ar 22500
[06:17] <mancha> Error while opening encoder for output stream #0:1 - maybe incorrect parameters such as bit_rate, rate, width or height
[06:18] <mancha> is 22.5k not allowed?
[07:20] <V1nce> hello
[07:21] <V1nce> question: how can i start ffplay fullscreen? i tried -fs but no go
[07:21] <V1nce> in fact it was like going into fullscreen for a half second but it got back windowed
[07:22] <V1nce> anyone alive? :-O
[07:32] <zap0> ffplay --help?
[07:33] <V1nce> don't see any full screen on start zap0
[07:33] <V1nce> i want it to play in fullscreen immediately
[07:34] <V1nce> i checked docs as well, ofcourse
[07:34] <V1nce> but no traces of fullscreen on start, only "f" key while playing
[07:36] <zap0> ffplay -fs file.ext
[07:36] <V1nce> -fs doesn't work
[07:36] <V1nce> tried on 2 boxes
[07:38] <V1nce> oh wait, it works, just tried with ordinary pal video
[07:38] <mark4o> ffplay -fs works for me
[07:38] <V1nce> but on 720p it doesn't work
[07:38] <V1nce> with 720p video
[07:38] <V1nce> good god
[07:39] <V1nce> yes yes it works, just tried with another video
[07:39] <V1nce> 720p mp4 doesn't though
[07:39] <V1nce> it enters fullscreen for half second and gets to windowed mode
[07:41] <mark4o> what os and screen resolution
[07:42] <V1nce> win 7 1920 x 1080
[07:43] <V1nce> okay i just found out that ffplay actually switches the monitor resolution to the size of the video
[07:43] <V1nce> when played in fullscreen
[07:44] <V1nce> no scaling involved
[07:45] <V1nce> so something, somehow, is in conflict with this 1280x720 piece
[07:45] <V1nce> and it enters windowed
[07:45] <V1nce> no probs with lower resolution though
[07:45] <V1nce> but shouldn't be the problem with this one either. hmm
[07:48] <V1nce> Frame changed from size:0x0 to size:1280x720
[07:49] <V1nce> all seems good, but it starts windowed :-)
[07:49] <mark4o> 720p mp4 works in fullscreen for me, although I'm on a mac not win 7.  perhaps an issue with the resolution change
[07:49] <V1nce> yes
[07:49] <V1nce> if i can just tell it to scale it to the native res
[07:49] <V1nce> which will be perfect
[07:52] <V1nce> weird thing i press "f" and it is all good
[07:52] <V1nce> fullscreen plays like there's no tomorrow
[07:52] <V1nce> change res, scales, all great
[08:07] <V1nce> solved
[08:07] <V1nce> very elegantly
[08:07] <mark4o> cool
[08:08] <mark4o> how did you fix it?
[08:08] <V1nce> you know how? i command line VLC, works like charm :-)
[08:08] <V1nce> also i could hide mouse pointer
[08:09] <mark4o> ok no ffplay then
[08:09] <V1nce> no :-/
[08:10] <V1nce> it is desktop thing, fscommand it from flash projector, so all good
[08:11] <V1nce> wanted to open video externally, so it plays and when finished gently exits to the menu
[08:11] <V1nce> vlc is very discreet with this
[08:12] <mark4o> well glad you got something working
[08:12] <V1nce> me too
[08:12] <V1nce> ffmpeg is the beast with server related things
[08:13] <V1nce> batch it, tell it anything and it will do it
[08:14] <mark4o> it doesn't rebel too often
[08:17] <V1nce> keeps his mouth shut and does the job
[08:17] <V1nce> mark4o, are you related to ffmpeg development?
[08:17] <mark4o> no
[08:19] <V1nce> okay, more time for bourbon, chicks and horses
[08:20] <mark4o> :) just a user and fan
[08:29] <V1nce> gtg, ciao mark4o
[09:54] <zro> on Debian, do i need to install a seperate package for ffmpeg presets? i dont see them in /usr/share... ? I thought they were there as ffmpeg/*.ffpreset files or something
[10:55] <brontosaurusrex> zro: what presets do you need?
[12:12] <Yulth> join #zfs
[12:18] <spaam> Yulth: no thanks
[12:19] <jameshowe> there's no details in the current -map documentation about how to map from multiple input files to a single output file
[12:20] <jameshowe> e.g. taking a video stream from the first and an audio stream from the second
[12:22] <jameshowe> oh wait, maybe it does, except that "the second -map option specifies the source for output stream 1" is highly misleading
[12:29] <skifreak> hey guys, is there a way to copy truehd from one video to another? Just trying to do a -c:a copy from one video file to another. I know demuxing truehd is supported, but remuxing isn't. Trying to figure out a way to get around this problem, thanks!
[12:34] <skifreak> should I convert the truehd track instead to another format, such as flac, which would carry all of the channels and quality of the original audio encode?
[12:35] <divVerent> it depends on what you are trying to do
[12:35] <jameshowe> skifreak: that happens with a lot of things, usually you have to find a bsf to stick in
[12:35] <divVerent> what is your output format?
[12:35] <jameshowe> but then that isn't guarenteed to work
[12:36] <jameshowe> the mp4 muxer can't read it's own aac output for example
[12:36] <skifreak> my output format is mkv, but I can do m2ts
[12:37] <skifreak> here's the info on the track I'm trying to encode from http://dpaste.com/818100/
[12:38] <divVerent> what is the error you get for trying to mux it into mkv?
[12:39] <skifreak> [NULL @ 0x16b1d00] Unable to find a suitable output format for 'ashowinfo:fmt'
[12:39] <JEEB> <jameshowe> the mp4 muxer can't read it's own aac output for example <- huh? for me all of my AAC-in-"MP4" muxes have gone just nice, and then read with libavformat
[12:39] <skifreak> not a bad idea to use -bsf, just trying to preserve as much noticeable audio quality as I can. My objective is always tranparent encodes.
[12:39] <divVerent> skifreak: what if you don't use ashowinfo?
[12:40] <divVerent> well, one lossless to another lossless is also guaranteed transparent
[12:40] <divVerent> skifreak: I mean, you're using an audio filter, this basically isn't really consistent wiht using -codec:a copy
[12:40] <jameshowe> JEEB then if you try to make another mp4 with it using -c copy, you get the "malformated aac bitstream" message
[12:40] <JEEB> huh...
[12:41] <JEEB> lemme try
[12:41] <skifreak> divVerent, I didn't think I was using an audio filter if I was just copying the audio stream
[12:41] <jameshowe> or maybe it's make a mpegts first (with libfaac encoder) then streamcopy to an mp4
[12:41] <divVerent> well, you got an error from ashowinfo, which is an audio filter
[12:42] <jameshowe> anyway, googling that error message shows everyone else that gets that problem
[12:42] <divVerent> or is that just a VERY misleading error and your command line doesn't mention ashowinfo?
[12:42] <JEEB> jameshowe, lavf mp4 -> mp4 aac WorksForMe, so it's something else
[12:43] <skifreak> sorry, wrong line of code, [matroska @ 0x2116d00] Application provided invalid, non monotonically increasing dts to muxer in stream 1: 2 >= 2
[12:43] <divVerent> that tells more :P
[12:44] <skifreak> yea....bit late on this side of the sphere
[12:44] <divVerent> basically, whatever is your input is not suitable for mkv, not sure if this is caused by mkv rounding to milliseconds
[12:44] <divVerent> if your target format is mkv, I'd however transcode to flac anyway
[12:44] <jameshowe> JEEB, just checking my logs from yesterday, looks like (mpegts -> mp4 libfaac) and (mp4 -> mpegts libx264) are the streamcopies that needed bsfs
[12:44] <divVerent> when going between different multichannel formats, you may have to test the output
[12:45] <divVerent> as channel layouts may differ
[12:45] <JEEB> if the input is mkv or mpeg-ts he can try using mkvtoolnix
[12:45] <JEEB> mkvmerge from there
[12:45] <divVerent> and there may be bugs causing bad conversions
[12:45] <jameshowe> at least for the data I was playing with
[12:45] <skifreak> divVerent that's why I've been going with just copying the audio stream
[12:45] <skifreak> I could go with a different container, if you think one might be more suitable
[12:45] <divVerent> but other losses shouldn't happen when toggling between lossless formats
[12:46] <divVerent> you can try
[12:46] <skifreak> what if my input file was a m2ts instead, might I get different results?
[12:46] <divVerent> maybe
[12:46] <divVerent> does it go wrong with multiple such videos?
[12:47] <skifreak> nope, this is the first
[12:47] <divVerent> as it MIGHT also be an actually "defect" encode
[12:47] <divVerent> as this is a defect decoders would usually cope with, except for maybe a small click
[12:47] <divVerent> nobody would notice in the silence at the start anyway
[12:47] <divVerent> but muxers really HATE non monotonous dts
[12:48] <skifreak> ok, well, I have version 0.10.3
[12:48] <skifreak> why is that?
[12:48] <divVerent> as the dts is what they sort packets by
[12:49] <JEEB> jameshowe, the latter being a H.264 stream copy? yeah, mp4->mpeg-ts needs the h264_mp4toannexb bitstream filter, because mp4 contains the H.264 stream in another format, and mpeg-ts want it in another
[12:49] <JEEB> it wouldn't really make sense with libx264 re-encoding unless ffmpeg is being really dumb
[12:49] <divVerent> skifreak: BTW, does ffmpeg continue encoding apart from the error message?
[12:50] <JEEB> because libx264 can output either length prefixed /or/ start code prefixed
[12:50] <jameshowe> yep, and the other one needs aac_adtstoasc
[12:50] <skifreak> nope, it stops, dea in its tracks, divVerent
[12:50] <skifreak> *dead
[12:50] <divVerent> skifreak: so, no error, it doesn't abort, but also not encode
[12:50] <JEEB> ok, that doesn't mean anything is really broken, just that ffmpeg isn't for some reason "smart" enough to insert those bit stream filters in there
[12:51] <jameshowe> although I'm yet to determine whether the resulting file actually works properly on my target devices (whereas re-encoding with libfaac does)
[12:51] <skifreak> divVerent: when I try to encode, it gave me the above error, and wouldn't encode one frame. I was back to stdout
[12:51] <divVerent> can you quote all messages you got?
[12:51] <JEEB> not sure about the adts to asc filter, but in general it's the same thing, two ways of a bit stream and for some reason a certain muxer wants it in a certain way
[12:51] <divVerent> just asking, as I also once wrote code using ffmpeg, and did get this error due to a bug
[12:51] <JEEB> jameshowe, if you're encoding for hardware boxes or whatever I recommend you use a more tested mpeg-ts muxer
[12:51] <divVerent> but it continued encoding, with frames missing due to the error
[12:51] <JEEB> either kierank's libmpegts or videolan's thingy
[12:52] <JEEB> I'm still surprised that ffmpeg hasn't jumped on either
[12:52] <jameshowe> JEEB, no LCD android and iOS devices
[12:52] <jameshowe> which is "fun"
[12:52] <JEEB> well those things in general get fed quite the broken shit
[12:52] <skifreak> divVerent: this is my exact output http://dpaste.com/818104/
[12:52] <JEEB> so I guess libavformat's mpeg-ts muxing should be "OK"
[12:52] <jameshowe> the mpegts muxing is good
[12:52] <jameshowe> the mp4 muxing is good
[12:53] <JEEB> hahaha, now you said a good joke
[12:53] <JEEB> it's parse'able as mpeg-ts, but it really, really doesn't AFAIK follow the specs
[12:53] <kelnoky> I have several video files here and some of them don't play on my android device (but ofc fine on my pc). the only difference I found between them (via mediainfo) was that the working files have 90k maximum fps while the files that don't work have a lower maximum frame rate than they have minimum frame rate
[12:53] <divVerent> skifreak: ah, I see, this error is intentionally fatal in ffmpeg.c
[12:53] <JEEB> the only ones that I know that are freely available and follow the specs are kierank's libmpegts and videolan's whatever it was
[12:53] <skifreak> divVerent why is it intentionally fatal?
[12:54] <jameshowe> it's the aac that goes a bit heywire sometimes
[12:54] <divVerent> skifreak: because whenever this happens, the output file IS broken
[12:54] <JEEB> you should say "the mpegts muxing is 'good enough for my use case'"
[12:54] <divVerent> just, it may be not broken enough to not play any more ;)
[12:54] <JEEB> because saying it's "good" will lead to incorrect assumptions :)
[12:54] <skifreak> wow, ok
[12:54] <divVerent> skifreak: if you are fine with changing code, you could try changing the code there
[12:54] <divVerent> by removing that exit_program(1) line in ffmpeg.c
[12:54] <jameshowe> it's all relative when you're working with android devices,
[12:55] <divVerent> and then trying if you get only one, or lots of these errors
[12:55] <jameshowe> it's astonishing how many ways you can implement graphics drivers wrong
[12:55] <divVerent> if you get lots, then it sure won't work. If you only get one or two lost audio frames, chances are your encode MAY play back fine anyway, because of the first frames typically being silent anyway
[12:55] <divVerent> don't know how TrueHD copes with that though
[12:55] <skifreak> well, I can play back the video
[12:55] <skifreak> mplayer handles it fine
[12:56] <divVerent> skifreak: it did write video?
[12:56] <divVerent> but the dts is 2, that's very low...
[12:56] <divVerent> so your output file should have neither much audio nor video
[12:56] <divVerent> when it aborts at this point
[12:56] <JEEB> jameshowe, well the idea is that libavformat's mpegts muxer afaik doesn't create a mux that is fully compliant, while there are now at least two contenders for doing that
[12:56] <skifreak> yep, you got it
[12:57] <skifreak> It wrote the file, but it's extremely small
[12:57] <divVerent> ok
[12:57] <divVerent> then try that code change
[12:57] <divVerent> if you think you can
[12:57] <divVerent> if it leads to a successful encode, ffmpeg MAY even get an option for this in the future (strict -3 maybe? :P)
[12:57] <skifreak> I should be able to with this Gentoo box, it's been a really long time since I did it last though
[12:58] <jameshowe> right, so what I'm actually here for is mapping the video stream from a filter chain and the audio from a second input into the same file
[12:58] <jameshowe> "-map [out] -map 1:a" gives only audio
[12:59] <jameshowe> "-map 0:v -map 1:a" gives both streams, but the video is just solid grey (it's not supposed to be)
[12:59] <jameshowe> this was previously achieved by "-newaudio", but that option has been removed
[13:01] <JEEB> I think the map command is nowadays /really/ powerful but hell if I can remember the syntax :P
[13:01] <JEEB> and yeah, just tested on a mp4 mux I've done and I need the mp4-to-annexB filter for a mpeg-ts remux
[13:01] <JEEB> for H.264
[13:04] <JEEB> it's not broken or incorrect, but it's funny that you have to manually set it, since I would think it's relatively automatic that if you have a "MP4" file that you have the H.264 stream not in annexB format
[13:04] <skifreak> divVerent: all of the exit_program(1) lines?
[13:04] <skifreak> in ffmpeg.c?
[13:11] <divVerent> skifreak: only the one near a call to av_interleaved_write_frame
[13:11] <divVerent> you also can remove all and see what happens ;)
[13:12] <skifreak> got it, figured it out, compiling
[13:15] <jameshowe> and not using any -map gives the correct video, but no audio stream
[13:19] <skifreak> http://dpaste.com/818114/
[13:20] <skifreak> divVerent: received a series of errors, as seen above
[13:22] <divVerent> skifreak: okay... quite likely it won't work with matroska then
[13:23] <divVerent> what if you now try another output format?
[13:23] <skifreak> so I should use a different containter?
[13:23] <divVerent> yes
[13:23] <divVerent> try .nut
[13:23] <skifreak> k
[13:23] <JEEB> jameshowe, does it show which video track gets selected without the -map ?
[13:23] <skifreak> same
[13:23] <divVerent> with similar values too?
[13:23] <divVerent> increasing in increments of 5?
[13:24] <skifreak> nope, increments of 5
[13:24] <jameshowe> so without any map I get Stream #0:0 -> #0:0 (png -> libx264)
[13:24] <divVerent> now wondering if there is a way to print ALL pts values...
[13:24] <skifreak> 2788 >= 2788 is on one line, the next is 2793 >= 2793
[13:24] <divVerent> *dts
[13:24] <skifreak> good point
[13:24] <jameshowe> but that's the correct filter output (the filters use some movie=)
[13:24] <divVerent> skifreak: try again with -debug_ts
[13:25] <divVerent> and again quote the first pages of output errors ;)
[13:25] <JEEB> jameshowe, so -map 0:0:0 would be correct for that, m'kay. What about audio?
[13:25] <JEEB> (at least I just had -map 0:0:0 work to get the track 0 of file 0)
[13:27] <skifreak> divVerent: here http://dpaste.com/818117/ thanks for the help btw
[13:27] <JEEB> (and set it to be track 0 of output)
[13:27] <divVerent> skifreak: all you changed was -debug_ts?
[13:27] <skifreak> yep
[13:27] <skifreak> and .nut output
[13:27] <divVerent> weird, so debug_ts doesn't work with stream copy at all?
[13:27] <jameshowe> JEEB, when I did that it did not do the filterchain
[13:28] <jameshowe> hang on, now I'm getting "Output with label 'out' does not exist in any defined filter graph"
[13:28] <jameshowe> arrgh
[13:28] <divVerent> skifreak: what if you use matroska again?
[13:28] <divVerent> you should get the previous errors back :)
[13:28] <divVerent> the error here is about the video stream BTW
[13:29] <skifreak> hunh, I was able to disable the audio stream before and the video started to encode just fine
[13:29] <divVerent> [NULL @ 0x1ae0b20] [Eval @ 0x7fffcf4dcb20] Undefined constant or missing '(' in 'ts'
[13:29] <divVerent> [NULL @ 0x1ae0b20] Unable to parse option value "ts"
[13:29] <divVerent> [NULL @ 0x1ae0b20] Error setting option debug to value ts.
[13:29] <divVerent> ah, there's the problem
[13:29] <divVerent> it's -debug_ts, not -debug ts
[13:30] <skifreak> ah, sorry, will try again with .nut
[13:31] <skifreak> actually, i double checked, I didn't use _ts
[13:31] <skifreak> this is what I got for a .mkv output file http://dpaste.com/818120/
[13:32] <divVerent> you still have -debug ts
[13:32] <divVerent> not -debug_ts
[13:32] <jameshowe> but only when not attempting to add the audio
[13:32] <skifreak> well, it doesn't recognize debug_ts
[13:32] <JEEB> I just tried with two files
[13:33] <divVerent> skifreak: odd
[13:33] <skifreak> Unrecognized option 'debug_ts'
[13:33] <JEEB> -i derp1.mp4 - derp2.mp4 -map 0:0:0 -map 1:0:0 out.mp4
[13:33] <divVerent> my version has -debug_ts
[13:33] <divVerent> try -fdebug ts?
[13:33] <JEEB> and that seemingly works
[13:33] <JEEB> argh
[13:33] <JEEB> 1:0:1
[13:33] <JEEB> I mean
[13:33] <jameshowe> JEEB, yes that works, now try with a filter chain
[13:33] <skifreak> nope, same error
[13:34] <jameshowe> JEEB, you want my full command?
[13:34] <JEEB> somehow I feel like I don't, but pastebin it if you want :P
[13:34] <divVerent> skifreak: weird
[13:34] <skifreak> divVeren: here is what both of them look like http://dpaste.com/818126/
[13:34] <divVerent> for me, this works:
[13:34] <divVerent> ffmpeg -debug_ts -i file.mkv file.nut
[13:35] <JEEB> divVerent, his ffmpeg is version 0.10.3, wasn't -debug_ts newer than that?
[13:35] <divVerent> OH... that is it then
[13:35] <divVerent> does 0.10.3 have a similar option?
[13:35] <JEEB> I would guess not
[13:35] <skifreak> ok, well, which version should I get?
[13:36] <divVerent> either 1.0 or git
[13:36] <skifreak> ok, I can go with the hardmasked 1.0
[13:36] <divVerent> this MAY actually also fix your issue... stupid me, haven't tried that before
[13:36] <divVerent> beware, it has different library names
[13:37] <divVerent> if you have programs depending on ffmpeg, these will need recompiling
[13:37] <skifreak> deps?
[13:37] <divVerent> it may be easier to rather just compile ffmpeg in your home dir for a test
[13:37] <divVerent> and not install it, just use the ffmpeg binary you get after compiling
[13:37] <JEEB> in which case just doing a git clone is simpler
[13:37] <divVerent> sure
[13:37] <divVerent> you will need some --enable commands, like for libx264
[13:37] <skifreak> haha, it's no problem, I compile everythign with gentoo as it is, good workout for my 3770k
[13:38] <skifreak> so my ffmpeg command will change?
[13:38] <JEEB> can you thank gentoo for disabling asm optimizations on x264 for 32bit
[13:38] <jameshowe> http://pastebin.com/nYzs9Vnp
[13:38] <JEEB> and thus bringing a nice amount of stupid people saying x264 is slow onto #x264
[13:38] <divVerent> jeeb: that's only to force people to FINALLY use 64bit distros on 64bit systems ;)
[13:38] <JEEB> heh
[13:38] <divVerent> I so hate people using a 32bit OS on 64bit systems
[13:38] <divVerent> it's annoying
[13:38] <divVerent> especially if they have over 2 GB RAM ;)
[13:39] <skifreak> yea....have to agree with you there divVeren
[13:39] <jameshowe> JEEB, and other, more complex filters may also be involved with a few more movie=
[13:39] <skifreak> ok, going with 1.0 then
[13:39] <divVerent> just, this likely isn't why Gentoo does that
[13:39] <JEEB> ^
[13:39] <JEEB> I also found that macports did something similar
[13:39] <JEEB> thankfully nobody uses macports
[13:39] <JEEB> so it didn't create an influx of idiots
[13:39] <divVerent> Gentoo probably rather does it because the asm optimizations for 32bit don't work with a certain bleeding edge version of some compiler gentoo users love to install ;)
[13:40] <divVerent> like, maybe it was buggy for a week in git checkouts of clang ;)
[13:40] <JEEB> heh, looking at the randomness of compiler versions and browsers that are tagged 'stable', I really don't know about that either
[13:41] <divVerent> but still, x264 IS slow ;)
[13:41] <JEEB> lol
[13:41] <divVerent> unless you tell it to be fast
[13:41] <JEEB> the default is --preset medium
[13:41] <divVerent> but I want it to be slow and yield good quality, don't care if it takes longer
[13:41] <JEEB> which is far from slow
[13:41] <divVerent> (still encodes way faster than realtime)
[13:41] <divVerent> I still call it slow
[13:41] <JEEB> sure, it can be slow depending on your machine, but then again placebo is much slower
[13:41] <divVerent> "slower than transferring the file to the iPhone is" ;)
[13:42] <divVerent> but anyway
[13:42] <divVerent> my point here is, your software can be as fast as you want
[13:42] <divVerent> there's always people who'd like it even faster
[13:42] <divVerent> on a related note: "next iOS version shortened my battery lifetime"
[13:42] Action: skifreak not surprised
[13:43] <divVerent> basically, I call x264 "slow", but the slow performance is worth the results
[13:43] <divVerent> there is faster codecs, but they suck, and if you want worse quality, you can typically outperform these by using a faster x264 preset anyway ;)
[13:44] <skifreak> gotta love -preset
[13:44] <divVerent> also, I wouldn't complain if x264 from today to tomorrow got 50% slower. I'd either deal with it, or pick a faster preset.
[13:44] <JEEB> -preset and -crf are the best things that have happened to using video encoders lately :P
[13:45] <divVerent> -crf, mainly
[13:45] <JEEB> yeah
[13:45] <divVerent> preset, not so much
[13:45] <divVerent> or rather, LAME already does this too :P
[13:45] <divVerent> but is an audio encoder
[13:45] <JEEB> well, preset is nice for easymode speed/compression selection
[13:45] <divVerent> LAME's -q option is quite similar to x264 -preset
[13:45] <skifreak> speaking of which should I change the ffmpeg command I've been using in the dpastes? I switch could to something along the lines of -crf 18 -preset for tranparent 1080p?
[13:46] <JEEB> crf is nice because you don't have to start guessing random bit rate values for clips
[13:46] <divVerent> just that LAME named the option so wrong...
[13:46] <divVerent> I mean, how many people have once thought that -b is for CBR mode and -q for VBR mode? ;)
[13:46] <divVerent> while -V is VBR, and -q is just a general algorithmic quality setting (similar to x264 -preset)
[13:47] <divVerent> skifreak: as for the command I don't know
[13:47] <divVerent> you sure shzouldn't need that many options
[13:47] <divVerent> I'd just go with -crf 18 for a start
[13:47] <skifreak> ok
[13:47] <skifreak> deblockalpha wasn't recognized by 1.0
[13:47] <JEEB> argh
[13:47] <divVerent> it's probably called different
[13:47] <JEEB> manual libx264 settings make me cry
[13:48] <divVerent> -deblock           <string> E..V.. Loop filter parameters, in <alpha:beta> form.
[13:48] <divVerent> but you really shouldn't need to manually tune these
[13:48] <skifreak> ....and we're up and running!
[13:48] <divVerent> you can use all x264 -preset -profile -tune
[13:48] <skifreak> 1.0 worked
[13:49] <divVerent> skifreak: even the TrueHD stuff?
[13:49] <skifreak> well, it's encoding
[13:49] <divVerent> nice
[13:49] <skifreak> 8.8 fps with -crf and -slower 18, copying the audio
[13:49] <divVerent> so all we found out while debugging was that your ffmpeg version was old and broken ;)
[13:50] <skifreak> erm, -crf 18 and -preset slower
[13:50] <divVerent> however, BTW
[13:50] <divVerent> there is no "really" allowed way to put TrueHD in mkv
[13:50] <divVerent> it will go through that hack to use quicktime or avi fourccs
[13:50] <divVerent> wait, sorry
[13:50] <divVerent> no
[13:50] <skifreak> the quality and channels will still be preserved though...right?
[13:50] <divVerent> A_TRUEHD exists in matroska, so you can consider it officially supported
[13:50] <divVerent> if you copy, sure
[13:51] <divVerent> I was just saying, ffmpeg allows you to put "anything" into matroska, no matter how insane
[13:51] <divVerent> but it then uses weird "compatibility" hacks to encode quicktime or avi codec tags
[13:51] <divVerent> which I wouldn't assume to be a great idea to use
[13:52] <divVerent> but anything not "too odd" should be fine
[13:52] <divVerent> and I thought TrueHD was "too odd", but it's not ;)
[13:52] <skifreak> well, does bluray have any formats which are "too odd"?
[13:52] <divVerent> no idea
[13:52] <skifreak> hahaha, yep, that's dolby
[13:53] <divVerent> in "worst case" you may get a file that will work now but may fail in 5 to 10 years
[13:54] <divVerent> like, don't put Intel Indeo in mkv ;)
[13:54] <divVerent> not that anyone would want to use this codec anyway ;)
[13:55] <JEEB> well, at least the AVI compatibility mode is official, so those who have the decoder will be able to decode it
[13:55] <divVerent> jeeb: semi-official ;)
[13:55] <JEEB> just that if the codec gets a native muxing mode, that's of course better
[13:55] <divVerent> the compat mode itself is
[13:55] <divVerent> but the fourccs typically not so much
[13:55] <divVerent> It'd be new to me if AVI FOURCCs were always officially registered at Microsoft
[13:56] <JEEB> now that's a different level of pedanticity :P
[13:56] <divVerent> hehe
[13:58] <skifreak> divVerent: on using tune and profile: don't see a reason to use profile since I'll only play these files on my computer with mplayer.
[13:58] <divVerent> is the avi compatibility mode any more than a simple header info?
[13:58] <divVerent> skifreak: profile still kinda sets decoding CPU usage
[13:58] <JEEB> the default is high and most people want high for pc playback
[13:59] <divVerent> you MAY find out that using baseline profile may keep your fans quiter while playing back
[13:59] <divVerent> but you have to find that out yourself
[13:59] <divVerent> normally high is good
[13:59] <divVerent> jeeb: from what I see in ffmpeg's muxer, the most evil part of the AVI compat mode is that dts, not pts, get written into the timestamp field
[14:00] <divVerent> or is there more evil involved?
[14:00] <skifreak> divVerent: sound doesn't matter at all to me with my Noctua DH-14, video encoding is the reason I OCed my 3770k
[14:01] <iive> avi have timestamp field?
[14:01] <divVerent> iive: no
[14:01] <divVerent> iive: just seeint ghis in matroskaenc.c that when the AVI compat mode is invoked, write_dts is set
[14:01] <divVerent> which means that the matroska muxer will put dts, not pts, in the timestamp fields
[14:01] <divVerent> no idea why, of course there are no comments :P
[14:02] <iive> oh... i got scared for a moment.
[14:03] <jameshowe> so, any ideas for my -newaudio replacement?
[14:03] <divVerent> and a similar thing happens on demuxing, when "ms_compat" is set, the demuxer outputs dts, otherwise pts
[14:03] <skifreak> divVerent: seems the only tune options I would use is animation, film, and grain. Should I use grain only when the bluray source has grain in the video?
[14:04] <divVerent> skifreak: I never tested the latter two... so no idea
[14:04] <divVerent> I would assume that too
[14:05] <divVerent> grain: for grainy (or really foggy) source material where the grain should be kept and is not filtered out before encoding (see: https://en.wikipedia.org/wiki/Film_grain); the movie 300 is a good example of a grains source
[14:05] <divVerent> film: ment for normal film material encoded at a decent datarate (lowers the inloop-deblocking and tweaks the psychovisual settings slightly)
[14:05] <skifreak> http://forum.videohelp.com/threads/344451-the-meaning-of-various-tunings-in-x264?p=2148087&viewfull=1#post2148087 yep
[14:05] <divVerent> skifreak: that's the URL I just wanted to post as reference
[14:06] <skifreak> yea, already found the page before I asked, wanted to see if you had any more insight or could elaborate
[14:07] <skifreak> one of the many aspects of bluray I don't understand, is why film grain is present. Why can't better film/masters be made?
[14:07] <divVerent> they want to annoy you
[14:08] <divVerent> the grain is in fact digital watermarking to track down the movie pirates
[14:08] <divVerent> j/k
[14:08] <skifreak> lolz
[14:08] <divVerent> more seriously: bluray version of an old movie can only be as good as the film material of the movie
[14:08] <divVerent> they can't shoot the movie again
[14:09] <jacobs> Hello all I am trying to capture an android device screen with this command: /data/local/ffmpeg -y   -vcodec rawvideo -f rawvideo -pix_fmt rgb32 -s 320x480 -i /dev/graphics/fb0   /sdcard/output2.avi 2> /sdcard/out.txt . this creates a file with a single ( unclear ) frame and stops. this is the output: http://pastebin.com/LtfZ1Amg  . any idea what am I doing wrong ?
[14:09] <skifreak> true, but we're talking about movies such as 300 and the Passion
[14:09] <skifreak> they're new enough
[14:09] <divVerent> and e.g. The Matrix I'd assume to be a bit grainy :P
[14:09] <skifreak> nope
[14:09] Action: skifreak checks
[14:09] <divVerent> or, they may have saved $500 on equipment :P
[14:09] <divVerent> because there just was no budget for better stuff
[14:10] <divVerent> or, they like grain
[14:10] <divVerent> from working in an IT company, I know how it is with these budgets
[14:10] <skifreak> The Matrix has a very, very incredibly small amount of grain
[14:10] <skifreak> looking at 1.5' from my 55"
[14:10] <divVerent> there is always budgets to buy new servers, but never for a single DVD-R for backing up stuff
[14:11] <skifreak> I know, but still films such as 300 and the Passion have tons of grain
[14:11] <skifreak> 10x more than the Matrix, easily
[14:11] <divVerent> nobody knows
[14:11] <divVerent> it had a budget of $65 million
[14:11] <divVerent> maybe with less grain, it'd have needed $65,050,000 instead
[14:11] <divVerent> ;)
[14:11] <skifreak> yep, and not enough to buy decent film
[14:11] <skifreak> rofl, yep, probably right
[14:12] <skifreak> the next bothersome point for me is that there are not any films out at 60fps, and bluray cannot accomedate 60fps 3D
[14:12] <divVerent> just like I can't get a single DVD-R for backing up stuff at work here
[14:12] <divVerent> there just is no process to get cheap stuff
[14:12] <skifreak> wow, makes me glad I left the business world for awhile
[14:12] <skifreak> not to say grad school doesn't have its shortcomings...
[14:13] <divVerent> luckily, I am external here, we suffer a LOT less from the bureaucrazy[sic]
[14:13] <skifreak> nice
[14:14] <divVerent> the few "actual employees" mostly have to redeclare operational expenditures (running costs) as capital expenditures (project costs) just to make upper level management happy
[14:14] <divVerent> because there is no op-ex budget
[14:14] <divVerent> that's why e.g. systems don't get upgraded, they get replaced
[14:14] <skifreak> hunh, ok
[14:14] <divVerent> in the end it's way more expensive that way, but the upper management can say "we reduced operational costs by 50% again"
[14:14] <skifreak> doesn't sound too awfully bad..
[14:15] <skifreak> yea, but what is that to you
[14:15] <divVerent> I'm fine, my wages is cap-ex because I am external ;)
[14:15] <skifreak> my point :-D
[14:15] <divVerent> but it really looks like managing a company into destruction
[14:16] <skifreak> true, can't disagree
[14:16] <divVerent> and of course this all started since the company was bought by another country
[14:16] <divVerent> surprisingly not in USA, but UK
[14:17] <divVerent> *company in another country
[14:18] <skifreak> yea...definitely wouldn't want to be part of the complicated upper management in that
[14:19] <skifreak> the biggest problem I've been trying to get around presently is 3D playback. I am trying to play full3D, don't know if I should go with sbs 3800x1080 or frame sequential. I was thinking frame sequential, unless I could get my tv to recognized my computer as a bluray player and then send framepacked video to it
[14:20] <skifreak> I have options for sbs in 3D mode on my tv, but I don't know if it will be able to interpret the different between 1920x1080 video and 3880x1080. Most likely not (or unless it thinks the computer is a bluray player) So I have thought of going with frame sequential
[14:21] <skifreak> though how I do make mplayer pick up on the flagged left frames and send these flags to the video card and then the TV? haven't been able to find much of any help on #mplayer or ##hardware
[14:22] <skifreak> just thought if anyone would know how to get around the software decoding aspect of the problem, it might be here
[14:23] <skifreak> encoding the files doesn't seem to be very hard, it's just the plackback aspect
[14:28] <zap0> skifreak, i have pondered the same issues.  i have LG 3D TV.
[14:28] <zap0> it does SBS and OU, not sure if it does frame seq.
[14:29] <zap0> i have heard the TV will not recog 3800x1080 :(
[14:29] <skifreak> I have any LG as well zap0, LV9500. It can do frame sequential playback
[14:30] <skifreak> ah, yea, I don't know about mine, I would think it should....but I have never connected a bluray player to it
[14:30] <zap0> if you did frame seq,  how does the TV determine which is left, which is right?
[14:30] <skifreak> I think frame sequential would be the best for me, and I can tag the left frames...
[14:31] <skifreak> well, there are flags in the 3D file identifying the left frames. These flags should be passed through the video card to the tv
[14:31] <skifreak> if the tv wanted the right frames to be tagged instead I can simply change one of the 3d settings on the tv
[14:32] <skifreak> same as changing sbs L|R to R|L
[14:32] <zap0> back in 5, going out to TV to have a look
[14:33] <skifreak> zap0: cool, thanks
[14:39] <hendry> anyone have a recipe for shaving the first few seconds from a raw .mkv capture?
[14:41] <divVerent> mkvmerge can do this via --split
[14:42] <hendry> divVerent: hmm, mkvmerge doesn't seem to be packaged in Archlinux
[14:43] <hendry> wondering if there is a general equiv of imagemagick, but for videos, to fade-ins and audio bg tracks programatically
[14:43] <divVerent> it is packaged in arch
[14:43] <divVerent> the package is called mkvtoolnix-cli
[14:44] <divVerent> ffmpeg can use frei0r filters, that's as close as you will get
[14:47] <hendry> what does "frei0r" mean?
[14:48] <skifreak> zap0: any thoughts
[14:52] <jameshowe> re-wrote as a complex filtergraph and it works
[14:52] <jameshowe> going to be fun re-writing the ones with even more input...
[15:05] <natrixnatrix89> Hi guys I'm struggling with ffmpeg to read mjpeg video stream from an ip camera..
[15:05] <natrixnatrix89> it says "could not find codec parameters" although I've specified -vcodec mjpeg and -s 640x480 http://pastebin.com/vgsQsm2r
[15:05] <natrixnatrix89> if I try the same url in VLC, it does play fine..
[15:06] <natrixnatrix89> (although the network connection of camera is very intermittent..)
[15:13] <skifreak> zap0: let me know if you find anything, this is the most I've been able to find out http://www.avsforum.com/t/1281356/how-to-create-3d-frame-sequential-video/30
[15:18] <durandal_1707> natrixnatrix89: fetching file with wget and displaying with ffplay works fine
[15:19] <jameshowe> does the documentation for alphamerge imply that setpts=PTS-STARTPTS on the inputs is pointless?
[15:20] <durandal_1707> natrixnatrix89: use -f mjpeg  and not image2
[15:21] <durandal_1707> and you do  not need to set vcodec
[15:38] <natrixnatrix89> durandal_1707: Thanks!
[15:45] <skifreak> seems ffmpeg-1.0 has broken gst-plugins-ffmpeg, for me, can't seem to compile it http://bpaste.net/show/53566/ has anyone else ran into this problem?
[15:47] <JEEB> skifreak, most probably it just needs some older API
[15:47] <JEEB> can't really do much else but to ask the using application/library to update their usage
[15:47] <JEEB> or use the internal copy they seemingly have in there
[15:48] <skifreak> JEEB any recommendations? how would I use the internal copy?
[15:49] <JEEB> also it seems like that version you're pulling off is made for ffmpeg-0.10.13 ?
[15:49] <JEEB> see if your distro has customized it for newer versions
[15:50] <JEEB> skifreak, removing --with-system-ffmpeg might do it :P
[16:07] <skifreak> JEEB: thanks for the help, it seems this is an oustanding bug (423717)
[16:08] <Sashmo_at_work> I could really really use some help.  I am at my wits end with my streams going out of sync after 5+ hours.  Anyone willing to pick up where I left off, and fix it?  I would be willing to pay for the assistance.
[16:15] <Silicium> hi there
[16:15] <Silicium> i just compiled ffmpeg/ffserver because i had a problem using ffmpeg and h264
[16:16] <Silicium> but i still getting a segfault on:
[16:16] <Silicium> ffmpeg -v verbose -r 5 -s 640x480 -r 29.97 -f video4linux2 -i /dev/video0 http://localhost:8080/webcam.ffm
[16:16] <Silicium> ffserver is running
[16:17] <Silicium> Port 8080
[16:17] <Silicium> BindAddress 0.0.0.0
[16:17] <Silicium> MaxClients 10
[16:17] <Silicium> MaxBandwidth 50000
[16:17] <Silicium> NoDaemon
[16:17] <Silicium> <Feed webcam.ffm>
[16:17] <Silicium> file /tmp/webcam.ffm
[16:17] <Silicium> FileMaxSize 10M
[16:17] <Silicium> </Feed>
[16:17] <Silicium> <Stream webcam.mp4>
[16:17] <Silicium> Feed webcam.ffm
[16:17] <Silicium> Format h264
[16:17] <Silicium> #VideoSize 640x480
[16:17] <Silicium> #VideoFrameRate 10
[16:17] <Silicium> VideoBitRate 2000
[16:17] <Silicium> VideoQMin 1
[16:17] <Silicium> VideoQMax 10
[16:17] <Silicium> </Stream>
[16:17] <Silicium> oops, sorry
[16:17] <Silicium> any idea?
[16:17] <Sashmo_at_work> DUDE use pastebin!!!!!!!
[16:17] <Silicium> yes i'am sorry
[16:40] <tikasukaz> hello every one :)
[16:50] <tikasukaz> in presets what is flags2=+wpred+dct8x8+fastpskip
[16:50] <tikasukaz> googled can't find any proper documents for "flag2"
[16:50] <JEEB> you generally shouldn't care about those, they are minor things and either on or off depending on the preset
[16:50] <JEEB> (and profile)
[16:52] <tikasukaz> problem is i have two identical server on 1st server encoding is fast (300 fps+ ) on 2nd server encoding doesn't go above 150 fps
[16:52] <tikasukaz> 2nd server is new, and much lower cpu load thn 1st server.
[16:53] <JEEB> well, use a faster preset in that case, or check for other kinds of things that are limiting the frame rate
[16:53] <JEEB> also make sure your libx264 is compiled with assembly optimizations
[16:55] <tikasukaz> my friend compiled ffmpeg for ACC3 support he also tweaked few things i guess.. can i use those codes on new server ?
[16:55] <tikasukaz> CentOS version are different.
[16:56] <tikasukaz> sorry for noob question..
[16:56] <zap0> skifreak, got a bit distracted!
[16:57] <zap0> skifreak, only have antenna input, and 3d options would not permit me to move the cursor to the frame-seq icon,  only SBS and OU
[16:57] <JEEB> uhh, you most probably will want to build your own up-to-date libx264 and ffmpeg
[16:57] <JEEB> and probably yasm, since the one that comes with centos will probably be very awfully old
[16:57] <zap0> skifreak, maybe frame-seq only works once a compatible HDMI input is active.
[16:57] <tikasukaz> JEEB thanks alot.. i will try.
[17:01] <JEEB> tikasukaz, http://4.bp.blogspot.com/_D_Z-D2tzi14/S8TRIo4br3I/AAAAAAAACv4/Zh7_GcMlRKo/s400/ALOT.png
[17:03] <Tjoppen> I like this alot
[17:06] <tikasukaz> lol :P
[17:06] <tikasukaz> actually it should be "a lot"
[17:07] <tikasukaz> i was bad in English classes, mam was too hot :P
[20:06] <Yulth> Hi everyone
[20:11] <Yulth> I'm encoding HE-AAC and HE-AACv2 to make M4A files, but when I play this files under players like Foobar2000, their properties are the same: AAC audio with SBR for both, but HE-AACv2 doesn't seem to have PS. Anyone have any idea on what is happening?
[21:25] <knob> Good afternoon everyone =)
[22:06] <zro> im having trouble converting some mp4 files to ogv w/ ffmpeg2theora. it creates files that just blip open in vlc which then reports the message [0x1116ee8] main input error: ES_OUT_RESET_PCR called
[22:07] <zro> ffplay open the file and plays the first fraction of a second really fast, then shows a still frame for the duration.
[22:07] <zro> any id ea what my issue might be?
[22:20] <undercash> hi
[22:20] <undercash> it s possible to stream subtitles with ffmpeg?
[22:26] <zro> ffmpeg -i input.mp4 -acodec libvorbis -vcodec libtheora  output.ogv appears to be making a file that is huge and hiddeous, but plays
[22:42] <zro> ya this is ridiculous. i can make it work on certain files, but not tothers, but i cant see what the difference would possibly be
[22:46] <zro> here's the diff in poutput from the works and fails streams : http://www.diffb.in/by
[23:29] <llogan> zro: ffmpeg2theora is not a ffmpeg project.
[23:29] <llogan> also it's hard to see what you're doing without seeing the command and complete console output.
[23:30] <llogan> and yes, the defaults aren't always great (excluding libx264).
[23:36] <zro> thanks llogan
[23:43] <llogan> zro: by "i can make it work on certain files, but not tothers" are you referring to ffmpeg or ffmpeg2theora?
[00:00] --- Fri Oct 26 2012