[Ffmpeg-devel-irc] ffmpeg.log.20130104

[burek]

[00:03] <seablade_mbp> Ok so I gave up for the time being on skipping the file, I am recording to the file and trying to come up with a good command line for ffmpeg to stream in realtime a MPEG2 TS h.264/aac file so that I can preview it in VLC for now(Getting it to my server will come later)
[00:03] <seablade_mbp> What I have right now is… `ffmpeg -i atem.ts -re -acodec libfaac -ab 192k -ar 44100 -ac 2 -strict experimental -vcodec libx264 -vb 2000k -f flv udp://127.0.0.1:58440`
[00:03] <seablade_mbp> While it seems to be working, I can't seem to open it in VLC to confirm it is working
[00:04] <seablade_mbp> ideally I would like to just copy that video and audio encoding rather than reencode, but for some reason just copying didn't seem to work
[00:05] <seablade_mbp> More correctly running with -vcodec copy and -acodec copy seems to try to do it as fast as it can, even if -re is given
[00:05] <seablade_mbp> Obviously that doesn't work so well when I need it delivered in realtime
[00:06] <seablade_mbp> Moving the -re to be the first thing on the command line causes ffmpeg to lock up here as well, so I don't think that will work to well:)
[00:10] <BtbN> ffmpeg -re -i test.mkv test.avi works fine here.
[00:10] <BtbN> with -re it limits to the real framerate, without it uses as much CPU power as possible
[00:12] <seablade_mbp> BtbN, Sadly it doesn't seem to be working here.  Either ffmpeg locks up (-re located before the input file) or it moves as fast as possible(Located after the input file) git version from September
[00:13] <BtbN> it has to be before the input to work.
[00:13] <seablade_mbp> BtbN, I should note that I am trying to encode to a stream, not to a file
[00:13] <BtbN> that's the same
[00:13] <seablade_mbp> BtbN, I thought I remembered that yes, but see my comment about it being before:)
[00:13] <BtbN> you're doing something else wrong then if it locks up with -re
[00:14] <seablade_mbp> BtbN, Doesn't seem to be anything different, lemme grab the command I am using...
[00:14] <seablade_mbp> BtbN, ffmpeg -re -i atem.ts -acodec copy -vcodec copy -f flv udp://127.0.0.1:58440
[00:15] <BtbN> you're sure flv supports the codecs you want to copy?
[00:15] <seablade_mbp> BtbN, h.264 and aac, yes
[00:16] <seablade_mbp> BtbN, Standard codecs I have used for other purposes, wouldn't have any problem switching off fly but haven't yet only because I know it works
[07:31] Action: Boon aww
[08:15] <macmichael01> Whenever I try muxing a video that contains Dolby TrueHD to the mp4 container I seem to get an error. Is it possible to mux Dolby TrueHD to mp4 container? The spec claims to support it: http://mp4ra.org/filetype.html.
[08:50] <thalsharif> hello there, I have question about ffmpeg, so can someone help on that ?
[08:52] <JEEB> macmichael01, basically stuff listed by the mp4ra means that someone wrote a specification for it, and registered the name for the thing... Too bad Dolby be Dolby
[08:52] <JEEB> I guess the specification isn't exactly easily available
[08:53] <JEEB> lemme see if L-SMASH has implemented it
[08:55] <JEEB> yeah... nope
[08:55] <thalsharif> My question is: I receive from local port video stream which is UDP packet that contains only RGB data, no header no specific protocol used, just only RGB 24bit over UDP packet. My goal is to receive and play my stream on PC using ffmpeg with vlc or ffplay, anybody can guide me how to do that. Thanks
[08:56] <JEEB> there was a DTS specification that was somehow gotten by certain people, so that's why L-SMASH could implement DTS-in-"MP4"
[08:56] <JEEB> macmichael01, find us the specification and you'll get your support, basically \o/
[09:02] <thalsharif> anybody can help me on my question.
[09:02] <thalsharif> hello
[09:02] <macmichael01> JEEB:  I saw a google summer of code project which looked like they were developing something related to this.  http://bit.ly/Us7BEb (MLP/TrueHD encoder) To bad the links are dead :(
[09:08] <macmichael01> JEEB: I know MKV supports TrueHD. How hard would it be to translate their specification to work within MP4?
[12:10] <thalsharif> hello everyone, I have question and I relley need some help.
[12:10] <thalsharif> My Problem is: I receive from local port video stream which is UDP packet that contains only RGB data, no header no specific protocol used, just only RGB 24bit over UDP packet. My goal is to receive and play my stream on PC using ffmpeg with vlc or ffplay, anybody can guide me how to do that.
[12:11] <saste> thalsharif, nothing special, it should be like reading from input, but you need to specify -pix_fmt rgb24 and size since there is no header information
[12:13] <thalsharif> i am working on windows, can you give command line to do that.
[12:13] <saste> thalsharif, what command are you using?
[12:15] <thalsharif> i am begginer with ffmpeg but i try to construct the command. ffmpeg -i udp://ip:port, then i don't know how to complete this command
[12:15] <thalsharif> upd://ip : port
[12:16] <saste> thalsharif, ffmpeg -pix_fmt rgb24 -s SIZE -r RATE -i udp://ip:port OUTPUT
[12:16] <saste> check the wiki/documentation for more info
[12:18] <thalsharif> ok i will check that. but as you said it's possible to do that with ffmpeg.
[12:19] <thalsharif> i just want to make sure ffmpeg has capabilty to do my work.
[12:20] <saste> thalsharif, it should, but checking is better
[12:21] <thalsharif> saste, thanks for your help :)
[12:22] <saste> thalsharif, you should also specify the input codec, ffmpeg -codec:v rawvideo -pix_fmt rgb24 -s SIZE -r RATE -i udp://ip:port OUTPUT
[12:24] <thalsharif> good, and what about playing the stream??
[12:24] <saste> ffplay -codec:v rawvideo -pix_fmt rgb24 -s SIZE -r RATE -i udp://ip:port
[12:27] <thalsharif> nice, i can use ffplay without ffmpeg to receive and play stream??
[12:28] <saste> thalsharif, why not?
[12:31] <Fjorgynn> 688x288 isn't 16:9
[12:31] <iive> it is recommended that you use some container. udp can loose packets and the raw demuxer won't notice it, leading to image shifting
[14:35] <An_Ony_Moose> what's a good intermediate format for videos? I have a recording here of which the resolution is a bit too high and intend to scale it down before further processing - what should I output it as, for little loss of quality while keeping a manageable filesize?
[14:39] <lnb> trying to convert a .mov to .flv with winff, the preset is: -vcodec flv -f flv -r 29.97 -s 860x484 -aspect 16:9 -b 300k -g 160 -cmp dct -subcmp dct -mbd 2 -flags +aic+cbp+mv0+mv4 -trellis 1 -ac 1 -ar 22050 -ab 56k
[14:40] <lnb> when it runs, errors are: requested bitrate is too low. Error while opening encoder for output stream #0.0 - maybe incorrect parameters such as bit_rate, rate, width or height
[14:40] <lnb> no flv is made
[15:02] <klaxa> An_Ony_Moose: either use a lossless codec like ffv1, huffyuv (might get big) or lossless settings for libx264, or use a high quality libx264 settings
[15:04] <lnb> it keeps exiting with bitrate too low
[15:04] <klaxa> increase bitrate?
[15:04] <lnb> -vcodec flv -f flv -preset veryslow  -s 860x484 -aspect 16:9
[15:04] <lnb> thats what i have in preset
[15:04] <lnb> the file is 2.2gb
[15:05] <klaxa> increase the value after -b mabye?
[15:05] <klaxa> *maybe
[15:08] <lnb> trying again with -vb 400k
[15:09] <lnb> trying to convert .mov to .flv
[15:09] <sacarasc> Old version of ffmpeg?
[15:09] <klaxa> that might be too
[15:10] <lnb> ffmpeg version 0.8.4-4:0.8.4-0ubuntu0.12.04.1
[15:10] <lnb> is that old?
[15:10] <klaxa> that probably isn't even ffmpeg :P
[15:10] <klaxa> yeah ffmpeg is at version 1.0+
[15:11] <lnb> i typed in shell ffmpeg -version
[15:14] <limpc> hi guys. is gps info available in video?
[15:15] <klaxa> pretty sure there is
[15:15] <klaxa> would be weird if photos had it and videos hadn't
[15:17] <lnb> when converting it shows Duration: 00:15:38.02, start: 0.000000, bitrate: 19763 kb/s
[15:18] <lnb> does that mean i have to set bitrate to 19763?
[15:19] <limpc> well the reason i ask is because unalike photos, video is 'on the move'. so its not like you can just add 1 gps coordinate to the entire video
[15:19] <saste> limpc, in general: no
[15:20] <limpc> as in no, theres no gps info?
[15:20] <saste> in theory you could add metadata per frame, but i'm not aware of any format which supports per-frame metadata
[15:20] <saste> geotiff does, but is only for images
[15:21] <saste> limpc, as in, in theory possible, but i don't know if some formats make use of it
[15:21] <limpc> hmm thats a bummer
[15:21] <saste> and for sure we only support per-frame metadata only in the Tiff decoder
[15:22] <limpc> we were hoping to make use of gps attached to video files for a project.  I came across an interesting idea by one person who encoded gps info into the audio stream
[15:22] <limpc> so it had me wondering
[15:22] <saste> "in the audio stream"?
[15:23] <saste> limpc, how was that?
[15:23] <limpc> yeah. let me find the page hd a min
[15:23] <saste> limpc: BUT: you can store such info in a DATA stream
[15:24] <limpc> http://www.sivann.gr/software/gps2dtmf/
[15:25] <lnb> anyone have a preset line for .mov to .flv so it can be used full screen good quality?
[15:26] <saste> limpc, yes in theory you can encode anything in audio data, but is that a good idea?
[15:26] <lnb> [flv @ 0x1bd8700] requested bitrate is too low
[15:26] <limpc> saste, probably not, it'll probably add an odd noise, but I imagine sometimes options are limited
[15:27] <limpc> obviously thats not a usable solution by us. we need something thats already there on mobile phones and modern dslr cameras/videocams.
[15:28] <limpc> so i guess I'm SOL :)
[15:32] <espr3ss0> hi, me again ;-p
[15:32] <espr3ss0> Q: does "no symbols" or, "nothing to do" cause any problems ?
[15:33] <espr3ss0> installing "gettext" & "glib" @ the moment for ffmpeg
[15:34] <luc4> Hi! I'm using avformat_find_stream_info(formatContext, NULL) to get info on the streams. I tried with an file containing an aac stream, but when I get the duration I get a wrong value. Is this supposed to happen?
[15:34] <Diogo> hi this is possible use variable frame rate ??
[15:35] <Diogo> (http://pastebin.com/48qcjqJZ)
[15:35] <Diogo> media info said always Constant
[15:36] <Mavrik> luc4, that depends on the container
[15:36] <Mavrik> luc4, but for some the duration is not reliable :\
[15:37] <luc4> Mavrik: no container, it is a aac stream I extracted from a mov file.
[15:37] <Mavrik> yeah, that would explain why probing the start can't give you reliable duration :)
[15:37] <Mavrik> duration is usually noted in container metadata
[15:38] <luc4> Mavrik: so there is no way to get a duration or maybe that value is modified while decoding?
[15:39] <Mavrik> well, you can read the file and calculate duration that way
[15:39] <luc4> Mavrik: thanks!
[15:39] <Mavrik> IIRC that value doesn't change while decoding
[15:55] <espr3ss0> Q: how do u compile "Glib" with flags ? o_O
[15:55] <espr3ss0> is there a tutorial ?
[15:56] <lnb> is -b = bitrate?
[15:56] <lnb> as in -vcodec flv -f flv -s 640x360 -aspect 16:9 -qmin 6 -qmax 11 -b 800k
[15:57] <lnb> [scale @ 0x17e2720] w:1920 h:1080 fmt:yuvj420p -> w:640 h:360 fmt:yuv420p flags:0x4
[15:57] <lnb> [flv @ 0x17e1700] [lavc rc] Error: bitrate too low for this video with these parameters.
[15:57] <espr3ss0> lnb: "http://forum.videohelp.com/threads/277807-Useful-FFmpeg-Syntax-Examples"
[15:59] <lnb> it says -b int
[15:59] <lnb> do you add a 'k' after the number?
[15:59] <lnb> or just -b 800
[16:01] <sacarasc> You add a k.
[16:01] <lnb> ok
[16:05] <espr3ss0> Q: is compiling "./configure | make | install" ?
[16:05] <An_Ony_Moose> klaxa: so what filename extension should I use?
[16:05] <sacarasc> espr3ss0: More or less.
[16:05] <An_Ony_Moose> why use pipe?
[16:06] <An_Ony_Moose> && is preferred to my knowledge
[16:06] <An_Ony_Moose> so ./configure && make
[16:06] <An_Ony_Moose> make install is installation, no longer compilation
[16:07] <klaxa> An_Ony_Moose: i guess .mkv for matroska should be fine for either of those codecs, i'm not sure though
[16:07] <klaxa> if you are using ffv1 or huffyuv, i'm not sure it'll work with matroska
[16:07] <espr3ss0> sacarasc: i need to 'compile' including ... "LIBFFI_CFLAGS=-I/usr/include/ffi LIBFFI_LIBS=-lffi"
[16:07] <An_Ony_Moose> klaxa: so just .mkv and add a -b (somethingsuitable) option?
[16:08] <luc4> Hi! I'm trying to decode audio streams and I'm reading the documentation for avcodec_decode_audio4. But can I ask a specific format for the decoded samples? Like the number of channels, but per sample etc... Who is deciding that?
[16:08] <luc4> but per sample = bits per sample :-)
[16:08] <klaxa> using bitrate based quality measures on any of the codecs i named is not wise
[16:09] <klaxa> if you are going lossly (libx264) do some test encodes with -crf
[16:09] <klaxa> lower crf values mean higher quality
[16:09] <klaxa> start with something like 20 or so
[16:09] <klaxa> if it looks like shit reduce the value
[16:09] <klaxa> don't encode the whole video just a part of it
[16:09] <klaxa> to check quality
[16:12] <Fjorgynn> crf seems more used than 2 pass encoding for H.264. Why?
[16:13] <JEEB> because most people don't have a set file size limit to abide to
[16:13] <An_Ony_Moose> klaxa: ok, thanks
[16:13] <lnb> all this does is fail
[16:13] <lnb> (#*&@&
[16:13] <Fjorgynn> JEEB: but it seems like the "scene standard" also is crf today
[16:13] <JEEB> yes, motherfucking finally!
[16:14] <JEEB> it took them how many years?
[16:14] <Fjorgynn> but then I don't have to do 2 pass encoding, right?
[16:14] <JEEB> yes, you don't
[16:14] <Fjorgynn> http://scenerules.irc.gs/t.html?id=2012_SDTVx264u1.nfo
[16:14] <JEEB> and because the encoder isn't trying to hit a certain average rate, rate control isn't limited either
[16:14] <JEEB> (unless you set vbv limits)
[16:15] <JEEB> in other words, you are not losing quality/compression at all
[16:15] <JEEB> while you are gaining by the fact that you are only doing one pass
[16:15] <JEEB> granted, that one pass is not a "fast first pass" like it is with -pass 1, but I think that just makes sense :P
[16:15] <Fjorgynn> can I do crf even if I rescale the video?
[16:15] <Fjorgynn> I must
[16:15] <JEEB> naturally, rescaling etc. doesn't affect the encoder
[16:16] <JEEB> just that you might want to tweak your crf value of choice depending on the resolution
[16:16] <JEEB> because lower-resolution stuff generally tends to get upscaled more during playback
[16:16] <JEEB> and thus it is naturally easier for you to see possible artifacts
[16:16] <Fjorgynn> I find stuff somewhere which are 4:3 but they release it as 1280x720...
[16:17] <An_Ony_Moose> klaxa: changing the crf doesn't seem to have any effect at all
[16:17] <JEEB> anyways, as I have already stated o9k times, you should under no circuimstances look at the "scene rules" as any kind of hint at "what should I do"
[16:17] <klaxa> you may just not see the difference
[16:17] <Fjorgynn> ;)
[16:17] <JEEB> An_Ony_Moose, do you use bit rate at the same time or are your values just so low that you can't see the difference?
[16:17] <Fjorgynn> JEEB: so if I am only rescaling I don't have to do -crf at all?
[16:18] <Fjorgynn> only -c:v libx264 so it rescales and -c:a copy, right?
[16:18] <JEEB> Fjorgynn, why are putting together pre-encoding stuff (deint/resize/crop etc.) and encoding!?
[16:18] <An_Ony_Moose> JEEB: only crf
[16:18] <JEEB> An_Ony_Moose, what encoder are you using?
[16:18] <Fjorgynn> JEEB: I thought you must encode if you rescale?
[16:18] <An_Ony_Moose> JEEB: whatever's the default for file extension mkv
[16:18] <JEEB> An_Ony_Moose, depends on your build
[16:18] <JEEB> there is no single default, see the log
[16:19] <JEEB> note: you should never. NEVER imply that ffmpeg is gonna always pick the right one
[16:19] <JEEB> (that you want)
[16:19] <Fjorgynn> lol default
[16:19] <JEEB> Fjorgynn, yes if you resize it implies that you have to re-encode
[16:19] <JEEB> but still you were putting apples and oranges together just now
[16:20] <An_Ony_Moose> JEEB: ah... it's using mpeg4. I assume I have to change it to libx264 then?
[16:20] <JEEB> An_Ony_Moose, yup
[16:20] <JEEB> crf is only in libx264
[16:20] <Fjorgynn> dunno
[16:20] <JEEB> Fjorgynn, look at it this way: 1) Resizing implies that you have to re-encode 2) Re-encoding does not imply that you have resize
[16:21] <JEEB> *have to
[16:21] <An_Ony_Moose> bah. Is my problem that I'm using "ffmpeg version 0.8.4-6:0.8.4-0ubuntu0.12.10.1, Copyright (c) 2000-2012 the Libav developers"? Should I be using a version that's not from libav?
[16:21] <Fjorgynn> I must resize a file that's 1280x720 and 1GB > because I want it to be 4:3 and 300 MB
[16:21] <JEEB> An_Ony_Moose, you should be using the 'avconv' binary in that case, and you want to install the libavcodec-extra package to get libx264
[16:23] <JEEB> Fjorgynn, anyways I hope you understand more how you took apples and oranges as the same with <Fjorgynn> JEEB: so if I am only rescaling I don't have to do -crf at all? <Fjorgynn> only -c:v libx264 so it rescales and -c:a copy, right?
[16:23] <lnb> http://pastebin.com/KEND5PW0
[16:23] <An_Ony_Moose> what's actually the deal with avconv/ffmpeg? Why are they split?
[16:23] <JEEB> An_Ony_Moose, in early 2011 there was this split, both projects are in active development and ffmpeg merges most if not all of libav's change
[16:23] <JEEB> *changes
[16:24] <JEEB> it's better not to go into the possible politics back then because both just seem to want to put that behind themselves
[16:24] <An_Ony_Moose> ok
[16:24] <Fjorgynn> ffmpeg -i input -c:v libx264 -vf crop=960:720:160:0,size=640:352 -crf 21 -c:a copy output.mp4
[16:24] <An_Ony_Moose> well I think I'll get ffmpeg because I have no idea how the options for avconv work (or are they the same?)
[16:24] <JEEB> well, the new commands rewrite mostly came from libav
[16:25] <JEEB> so using avconv should work mostly with the avconv binary
[16:25] <JEEB> there are some differences
[16:25] <An_Ony_Moose> I'll get ffmpeg then
[16:25] <JEEB> http://x264.fushizen.eu/random/psp_encoding.txt <- see how libx264 specific settings (you don't generally have to use these at all) get used f.ex.
[16:25] <JEEB> well, suit yourself :) That works too
[16:25] <espr3ss0> Q: glib needs pkg-config & pkg-config needs glib ???
[16:26] <An_Ony_Moose> 16:25 < JEEB> http://x264.fushizen.eu/random/psp_encoding.txt <- see how libx264 specific settings (you don't generally have to use these at all) get used  f.ex.
[16:27] <An_Ony_Moose> oops
[16:27] <Fjorgynn> lol
[16:29] <espr3ss0> used "./configure --with-internal-glib" and ok, so why need glib ?
[16:30] <espr3ss0> pkg-config i'm referring to
[16:31] <An_Ony_Moose> ooh it wants yasm... What one doesn't do for speed :D
[16:33] <An_Ony_Moose> why do glib and glibc have to have such similar names?
[16:38] <JEEB> An_Ony_Moose, yes -- yasm is needed for compilation because x264/ffmpeg have a LOT of asm in there
[16:38] <mpfundstein> its not needed
[16:38] <An_Ony_Moose> I can imagine
[16:38] <mpfundstein> but its advised :-)
[16:38] <JEEB> mpfundstein, well yes -- but it's not really worth it to make a build without them
[16:38] <An_Ony_Moose> http://dpaste.org/dsCCD/ damn#
[16:39] <JEEB> also neither will let you configure without an extra option
[16:39] <lnb> anyone have a preset to convert .mov to .flv so when watching video in full screen its good quality
[16:39] <JEEB> without an assembler
[16:39] <mpfundstein> JEEB: yes --disable-yasm
[16:39] <mpfundstein> JEEB: but one should not do it
[16:39] <JEEB> ...I HAVE SAID THAT QUITE A FEW TIMES
[16:39] <JEEB> like, just now
[16:39] <An_Ony_Moose> http://dpaste.org/dsCCD/ not very related but how do I fix that?
[16:40] <An_Ony_Moose> export PATH=$PATH
[16:40] <An_Ony_Moose> fixed it x)
[16:44] <corda> hey there
[16:45] <Ben_> I'm getting a duration change afte reencoding with ffmpeg
[16:46] <Ben_> so my input is  Duration: 00:00:02.85, start: 0.013333, bitrate: 1039 kb/s
[16:46] <espr3ss0> so installing Glib, is for an update ?
[16:46] <Ben_> and after ffmpeg -i myinput.mp4 -acodec copy -vcodec mpeg4 -b:v 500k output.mp4
[16:46] <Ben_> I'm getting
[16:46] <Ben_> Duration: 00:00:02.92, start: 0.000000, bitrate: 485 kb/s
[16:47] <Ben_> any idea where the pb could be?
[16:50] <Fjorgynn> mpeg4 :o
[16:50] <espr3ss0> Q: what's FAM ? o_O
[16:50] <Fjorgynn> Ben_: no I have no idea
[16:51] <Ben_> Fjorgynn: do you think it's a bug?
[16:51] <Ben_> I believe it hads an empty frame at the end
[16:51] <Fjorgynn> dunno, stay in here and wait for someone to answer
[16:51] <Ben_> *adds
[16:52] <Ben_> for more details: https://ffmpeg.org/trac/ffmpeg/ticket/2108
[16:52] <Ben_> I'm gonna fetch the git and try to recompile it
[16:53] <espr3ss0> Glib "make", "make[4]: *** [gconvert.lo] Error 1" ?
[16:53] <Ben_> I used brew to install it so not sure if it uses the HEAD of the git repo
[16:57] <seece> hello, any tips how to stream raw video over LAN?
[16:58] <seece> i'm on windows, using directshow to capture input
[17:13] <espr3ss0> Q: Glib still won't configure properly, even used "./configure --with-libiconv=native" (after updating libiconv)
[17:15] <espr3ss0> "#error GNU libiconv not in use but included iconv.h is from libiconv"
[17:17] <Ben_> hey the new ffmpeg ask for libavutil.52
[17:18] <Ben_> anybody know the diff between this one and 51?
[17:22] <Diogo> any tool to compare quality of 2 video and generate a graphic? command line tool please..
[17:24] <espr3ss0> Q: do i need to specify "--prefix=/usr/local" for libiconv ?
[17:29] <Diogo> any tool to compare quality of 2 videos and generate a graphic? a command line tool please..
[17:30] <espr3ss0> what's the latest version of "libconv" ?, i have 1.9.2
[17:36] <An_Ony_Moose> anyone know a good tool for deshaking videos? The video filter included with ffmpeg gives me terrible results.
[17:36] <Mavrik> hqdn3d?
[17:37] <ubitux> hqdn3d for deshake? wtf
[17:37] <ubitux> An_Ony_Moose: https://ffmpeg.org/trac/ffmpeg/ticket/1430 :(
[17:37] <espr3ss0> specified "--prefix:/usr/local" still no change
[17:38] <espr3ss0> :-9
[17:38] <An_Ony_Moose> http://public.hronopik.de/vid.stab/ seems quite effective but transcode doesn't like me. Or the ubuntu package maintainer for it doesn't.
[17:39] <An_Ony_Moose> ah well I guess I'll build transcode from source then
[17:40] <An_Ony_Moose> am I in the future? I have transcode 1.1.7 but the website says the latest is 1.1.0
[17:47] <saste> An_Ony_Moose, you are of course in the future, w.r.t. to that website
[17:48] <An_Ony_Moose> saste: haha
[17:55] <An_Ony_Moose> https://github.com/georgmartius/vid.stab now if only this were an ffmpeg plugin
[17:58] <espr3ss0> Q: anyone know how to fix the "libiconv" problem ?
[18:06] <Mista_D> Is the fatest way to decode x264 "-f null"? Can I force decoder to read every 10th frame or so?
[18:07] <JEEB> no
[18:07] <JEEB> if you want proper decoding that is
[18:07] <JEEB> you could make it skip everything but IDR frames, but no idea how to do that :P
[18:10] <Mista_D> JEEB: thanks, I'll see if I can read IDR only.
[18:23] <grepper> An_Ony_Moose: transcode in debian is badly broken due to leaving some modules out (license issues).  Dunno about ubuntu.
[18:44] <Fjorgynn> what does -f null do?
[18:45] <Fjorgynn> what do you use it for? Decode to what?
[18:46] <Fjorgynn> Just read what Mista_D wrote
[18:50] <extractor> Hello, do anyone know how to extract only the last image of a video?
[19:31] <ubitux> Fjorgynn: -f null demux and decode, and output to nothing
[19:36] <Fjorgynn> aha?
[19:46] <seece> can ffmpeg stream any codec with UDP? or does it need to be in a mpegts container?
[19:50] <ubitux> should not be container specific
[19:50] <ubitux> it's a protocol like another
[19:50] <klaxa> udp is just the protocol you should be able to send anything over udp
[19:51] <seece> ok great
[20:12] <seece> i get "real-time buffer 258% full! frame dropped!" errors even though i can see the CPU load is really low. any ideas what could be causing it with x264?
[20:46] <espr3ss0> Q: "libiconv" requires enabling, is this done by "--enable-static" ?
[20:49] <klaxa> doubt it
[20:49] <espr3ss0> how can i enable it ? o_O
[20:49] <Mavrik> what's libiconv?
[21:12] <espr3ss0> Q: am i ok with "/usr/local/" or only "/usr/local" or even "usr/local" ? :-)
[21:13] <espr3ss0> i have a feeling it's what cause "/usr/local//lib/.new"
[21:14] <sacarasc> Just /usr/local should work. If you do /usr/local/ it will put do /usr/local//bin, which is the same as /usr/local/bin, but looks ugly.
[21:14] <espr3ss0> ok thanks
[21:21] <llogan> espr3ss0: what distro?
[21:21] <llogan> ...not that it matters too much
[21:22] <llogan> eh, nevermind. i misread your question
[21:28] <espr3ss0> llogan: distro of ?
[21:28] <klaxa> gnu/linux distribution
[21:28] <klaxa> but doesn't matter regarding the question
[21:29] <espr3ss0> os x 10.6
[21:37] <luc4> Hi! I'm getting a segfault when using the avcodec_decode_audio4 function. What are the possible reasons? I'm running ffmpeg on an armv6 hardfp platform.
[21:39] <llogan> luc4: not many API answers here. try the libav-user mailing list if you don't get an answer.
[21:39] <llogan> http://ffmpeg.org/mailman/listinfo/libav-user
[21:48] <jadams> hello
[21:48] <jadams> I'm a noob trying to stream video with as little lag as possible over the network
[21:48] <jadams> I can only get it down to around 2s lag with this: `ffmpeg -framerate ntsc -video_size 320x240 -f x11grab -i :0.0 -vcodec libx264 -tune zerolatency -crf 26 -b 900k -probesize 1000 -analyzeduration 0 -g 10 -f mpegts udp://aye-seven.local:5003`
[21:48] <jadams> opening the stream in vlc (ffplay is laggier by far)
[21:49] <rkt> which is the best codec to use with ffserver for 8bit graphics? i mean vmv1 feels hardcore for an application only running at 8bit anyway? :D
[21:56] <jadams> it would appear the problem was in the client in fact (caching)
[21:56] <jadams> thanks for the help
[22:01] <hub2> Hi,I'm making a program which uses ffmpeg in C++,and when I'm compiling it I get UINT64_C was not declared in this scope error. Is there any way to remove it?
[22:04] <Mavrik> hub2: IIRC it's -D__STDC_CONSTANT_MACROS to the compiler to declare those macros also in C++
[22:07] <hub2> Thanks Marvik. That solved the problem
[22:19] <luc4> Hi! Is ffmpeg supposed to work on armv6 hardfp?
[23:18] <sesam> Hi
[23:19] <sesam> What application do u guys to split a quicktime before using two or more instances of ffmpeg?
[23:21] <sacarasc> I didn't know people did that.
[23:21] <sesam> and what is the fastest distro for ffmepg? arch maybe the way to go but it takes some time to get up and running, maybe its worth it?
[23:23] <sesam> I make large lots of large files that needs to compress into a h264, and a faster way would be to chop it up and process each slice on different computers and then stich them together again.
[23:32] <sesam> best distro for ffmpeg?
[23:38] <sesam> guess ill compile it on centos then =)
[23:38] <luc4> Hi! Anybody who knows what attribute_align_arg means?
[23:40] <llogan> sesam: the fastest distro is the one you're most familiar with
[23:40] <Mavrik> luc4: if you grep through the code you'll find out it's defined as force_align_arg_pointer
[23:41] <Mavrik> compiler macro.
[23:41] <sesam> llogan: thx
[23:41] <llogan> but arch usually does keep their repo ffmpeg more up to date than any other distro that i've checked
[23:41] <luc4> Mavrik: yes, I found that. My problem is that I'm getting a crash in a strange point of avdec_decode_audio4, I'm trying to understand if that might be related.
[23:42] <llogan> not that it's hard to compile ffmpeg (with the compile guides if you need help)
[23:42] <sesam> llogan: then ill go for arch instead. I want it to be as small as i can get it.
[23:42] <llogan> centos or arch would work just fine.
[23:43] <Mavrik> luc4: what does the debugger say about the crash?
[23:43] <luc4> Mavrik: the crash I'm getting is when avctx->codec->type != AVMEDIA_TYPE_AUDIO is executed.
[23:43] <Mavrik> hmm
[23:43] <Mavrik> are you passing a correct context pointer?
[23:43] <luc4> Mavrik: I'm working on an embedded platform. Really hard to get a gdb debugger working.
[23:43] <Mavrik> did you try to break and see the struct?
[23:43] <sesam> llogan: im currently installing a centos minimal on my proxmox node. i will try arch as well.
[23:44] <klaxa> sesam: splitting up a file and running multiple instances of ffmpeg doesn't increase encoding time
[23:44] <luc4> Mavrik: I'm doing the same instruction just before invoking that function in my code, and I get no segfault.
[23:44] <llogan> https://ffmpeg.org/trac/ffmpeg/wiki/CentosCompilationGuide
[23:44] <klaxa> libx264 is multithreaded already
[23:44] <klaxa> and should use all available cores
[23:44] <sesam> llogan: thx
[23:44] <klaxa> if you are spreading the split files across multiple computers it's a different story of course
[23:45] <klaxa> i wrote some code for that in python, it only works with matroska containers though, you'd also have to modify it heavily to match your usecase
[23:45] <sesam> klaxa: im spreadding them acress multible computers. Need the speed.
[23:46] <sesam> does libx264 use the GPU as well?
[23:46] <klaxa> in that case it makes sense to do that yeah
[23:46] <klaxa> not that i know of
[23:46] <klaxa> there have been approaches to use OpenCL with x264
[23:46] <klaxa> not sure if anything official has been released yet
[23:46] <luc4> Mavrik: oh wait wait sorry, seems it is segfaulting also in my code.
[23:47] <sesam> that would have been awsum if they could get opencl running
[23:47] <Mavrik> not really
[23:47] <klaxa> from what i read the usefulness is somewhat limited
[23:47] <Mavrik> the quality of GPU encoding is pretty much very bad across the board
[23:47] <sesam> ive tried the most commercial apps that are on the market and only episode pro does it but its crazy expensive
[23:48] <sesam> bad quality in like picture quality?
[23:48] <klaxa> Mavrik: huh? i thought they are using the GPU for general purpose computations
[23:48] <Mavrik> hard to be general-purpose with a highly specialized core
[23:48] <klaxa> mmh true that
[23:48] <Mavrik> memory barriers still hurt quality of any video encoding done on GPUs
[23:48] <Mavrik> which is the reason why x264 efforts have stalled
[23:49] <Mavrik> you can't do efficient searches when you're locked into small image areas
[23:49] <sesam> then what is the fastest way to encode a 264 file?
[23:50] <sesam> and still ensure the quality
[23:50] <Mavrik> a 4-core Intel CPU with huge clock
[23:50] <sesam> sayway?
[23:50] <sesam> hugh clock
[23:50] <sesam> ?
[23:51] <Mavrik> frequency.
[23:51] <Mavrik> as much GHz as you can get
[23:51] <sesam> =)
[23:51] <sesam> im swedish, sorry for not understanding
[23:51] <Mavrik> ^^
[23:51] <Mavrik> no problem
[23:52] <sesam> why 4 and not 8? is it because the frequency gets lower as more cores there are crampt into a single cpu?
[23:52] <klaxa> are there 8-core intel CPUs?
[23:52] <Mavrik> because you get diminishing returns when paralelizing video encoding
[23:52] <Mavrik> more than 4 cores (with HT) doesn't bring significant speed-ups anymore
[23:53] <sesam> more like 8 threads =)
[23:53] <Mavrik> klaxa: yeah, you can get 8-core Xeons with HT (16 threads total)
[23:53] <klaxa> oh nice
[23:53] <Mavrik> expensive beasts though
[23:53] <luc4> Mavrik: I suspect I found it. I'm missing the codec init part. Sorry.
[23:53] <luc4> Mavrik: thanks
[23:53] <Mavrik> luc4: no problem, glad you found a bug :D
[23:53] <sesam> i have 40x i7 3.2ghz computers that ive built.
[23:54] <sesam> so splitting a file up in 40 peaces will go fast?
[23:54] <klaxa> lol they have 10-cores
[23:54] <sesam> hehe
[23:54] <Mavrik> hmm
[23:55] <Mavrik> sesam: rather run 10 files at once :)
[23:55] <sesam> true that
[23:55] <klaxa> sesam: do you know of map-reduce?
[23:55] <klaxa> it's exactly for this usecase
[23:55] <sesam> klaxa nope never heard of it
[23:55] <sesam> ill do a search
[23:56] <klaxa> it's a well... design-technique? for extremely scalable distributed computing
[23:56] <klaxa> i don't know how well existing software can be used to optimize your usecase
[23:56] <Mavrik> yeah, trouble with map-reduce with video is that you have to move those huge files around
[23:56] <sesam> sounds interesting, not sure if it will work with ffmpeg though
[23:56] <Mavrik> I've found that wastes alot of time :\
[23:57] <klaxa> mmh
[23:57] <klaxa> Mavrik have we discussed this earlier last year? because i remember i was talking to someone from #ffmpeg about this
[23:57] <Mavrik> hmm, don't remember
[23:57] <klaxa> and i got some more or less usable code working for that
[23:58] Action: t4nk637 knocks, asks if there is anyone home who is familiar w/ ffmpeg build in space-constrained settings & for a narrow band of conversions...
[23:58] <Mavrik> we have a similar use case and just do a full file encode per 4 cores of a machine
[23:58] <sesam> maybe with a fast raid-0 system with 2-4 nics working in link-aggregation it will run goot
[23:58] <Mavrik> t4nk637: details! :)
[23:58] <Mavrik> but we have about 1-2 hr long files that have to be encoded to several formats
[23:58] <t4nk637> thanks @ Mavrik
[23:58] <t4nk637> Question to power-users and those familiar w/ embedded contexts:  I want to use ffmpeg for a narrow subset of conversions,  -- only converting jpeg images to a time-lapse mp4 file --
[23:59] <sesam> mavrik: u work at a post-house?
[23:59] <sesam> mavrik: do u do all conversions in ffmpeg?
[23:59] <t4nk637> What is the minimal subset/choice of muxers, demuxes, decoders, encoders, filters, etc that I'd need to do this?
[23:59] <t4nk637> just jpegs to mp4 timelapse
[23:59] <Mavrik> sesam: post-production (cutting, etc.) in Premiere, release encodes in ffmpeg yeah
[00:00] --- Sat Jan  5 2013