[Ffmpeg-devel-irc] ffmpeg.log.20120322

[burek]

[00:05] <JEEB> yes
[00:05] <JEEB> libavformat can demux matroska
[00:05] <JEEB> and you can DECODE H.264 with libavcodec
[00:06] <JEEB> libx264 is for encoding H.264
[00:07] <JEEB> (and libavformat can mux matroska as well)
[00:07] <tekk> thanks JEEB
[00:08] <tekk> i basically wanna demux and output the video and audio streams into separate files
[00:08] <hi117> ffmpeg -i bla.mkv -vn -acodec copy bla.bla
[00:09] <hi117> and ffmpeg -i bla.mkv -an -vcodec copy bla.(x264|mkv)
[00:09] <tekk> thanks (will be doing it via the C level API though)
[00:09] <JEEB> have le fun
[00:09] <hi117> ah
[00:09] <tekk> but thats a good command to test if my ffmpeg build is working as required :)
[00:09] <tekk> so still useful
[00:11] <hi117> JEEB: is ff going to add mkvmerge like capabilities? where you can add arbutrary files into it
[00:11] <hi117> ive seen some mkvs with fonts in them
[00:11] <JEEB> ask elenril
[00:11] <JEEB> he's the one who was working with stuff like that
[00:12] <hi117> hes not in channel, ty
[00:12] <hi117> and its not really important, the problem is solved but it would be a nice addition to ff
[00:12] <tekk> JEEB, do you know what kind of performance the demuxer has& say on a standard workstation CPU
[00:12] <tekk> i.e. in comparison to re-encoding
[00:13] <hi117> normally decoding is faster than encoding on my laptop, and IIRC demuxing is trivial
[00:14] <tekk> demuxing is just I/O right?
[00:14] <tekk> in theory
[00:14] <hi117> in theory ya, with a little bit of cpu to find the relevent data in the file
[00:15] <hi117> when i split files like that its always the disk thats the bottleneck
[00:15] <tekk> ok thats interesting :)
[00:15] <tekk> does ffmpeg easily remux into a mp4 container also?
[00:16] <hi117> ask JEEB that one lol, i havent had problems but i only use it a little
[00:17] <hi117> also mp4 can get complicated, like modifying it for streaming
[00:18] <tekk> JEEB ? thoughts
[00:18] <tekk> i remember a very long time ago (years) i had sync issues when doing this
[00:18] <tekk> i.e. mkv had some strange sync concept
[00:22] <judget>  I am attempting to use ffmpeg to stream from an hdpvr to a crtmpserver so I can live stream to a JW player in a web page.
[00:22] <judget>  here is the pastebin that shows what I am trying http://pastebin.com/A6xEdHds
[00:22] <judget>  it seems to go through about 4 cycles of transcoding and stops
[00:22] <judget>  I note one error about the audio
[00:22] <judget>  any suggestions or helpm wpould be greatly appreciated
[00:36] <tekk> any libavformat guru's here?
[03:08] <RHKratos> anyone tell me what's wrong with this setting? "libaacplus: bad aac setting: br:128000, AACch:2, AACsr:24000"
[03:57] <pythonirc101> how does ffmpeg compare with adobe media encoder for encoding a bunch of RGB Images into a movie and compressing it to h264.
[04:04] <dashcloud> it should compare very well, since x264 (the library & program) is in all likelyhood the finest h264 encoder available
[04:10] <pythonirc101> dashcloud: what about speed? are there any comparisons on the web somewhere about this?
[04:13] <dashcloud> speed is very good- and here's the results from a comparison in 2010- things have only gotten better since then http://www.compression.ru/video/codec_comparison/h264_2010/
[04:15] <dashcloud> if you have questions about x264 itself, and not so much ffmpeg, the folks over at #x264 can answer all of your questions far better than I could- but if you just have some general questions here, I can try to answer them for you
[04:17] <dashcloud> here's the 2011 codec comparison result: http://compression.ru/video/codec_comparison/h264_2011
[04:29] <dashcloud> hope that helps out- have a great day/night/etc!
[07:26] <jhericurl> if  720x405 is perfect 16/9 resolution, why is not possible to create 720x405  video?
[09:46] <JEEB> jhericurl, well it has to do with this thing called chroma subsampling :) Now, most video around you is chroma subsampled (most popular thing that end-users can get is 4:2:0, which means that luminocity is sampled at every sample, and chromacity is sampled at one value per 2x2 area)  -- this is why, when you deal with chroma subsampled video, you have to adhere to the limitations of the chroma subsampling of your colorspace (divisible by two for 4:2:0).
[09:55] <IamTrying> How to get ride of this error? https://gist.github.com/2157228
[10:03] <grepper> doesn't look like an error, the command appears to have finished.  Probably just a warning.
[10:26] <burek> IamTrying, if you mean this "non-existing SPS 32 referenced in buffering period", that's just a warning that your input file is probably damaged, but not too much
[11:14] <MehdiLauters> hi
[11:15] <MehdiLauters> I'm trie to read a mp4 video using 34 slices so i had to recompile ffmpeg
[11:15] <MehdiLauters> why the initial value is 16?
[12:08] <burek> MehdiLauters,
[12:08] <burek> can you please use pastebin.com, to show your command line and its output?
[12:15] <MehdiLauters> http://pastebin.com/cL6ayt4V
[12:16] <MehdiLauters> but initially it was "Possibly too many slices (34 >= 16), increase MAX_SLICES and recompile if there are artifacts"
[12:17] <mystica555> why in all the world would there be a 34 slice video *cry&
[12:27] <MehdiLauters> I'm working on a cuda mp4 encoder, and the default value is 34 (i don't know why), i tried to down it to 16 but then the encoder didn't work
[12:57] <grepper> I am looking to keep some scripts both backwards compatible as well as supporting the latest syntax such as -b:v -c:v etc.  Can anyone suggest the best way to determine this ?  Right now parsing "ffmpeg -h" for the string -b:v  would probably work, but seems a bit dodgy approch.
[13:03] <globus> How can I use the ffmpeg-api to stream x264 NAL units ? I want to use it only for streaming not encoding. I have checked the ffserver.c but I haven't got it yet
[14:18] <cbreak-work> ffmpeg rocks
[14:24] <Tjoppen> woo
[14:27] <cbreak-work> but it's not well documented at all :(
[14:28] <Tjoppen> google helps. some online guides are out of date though
[14:28] <cbreak-work> most I found were
[14:28] <cbreak-work> I wonder what the semantics of the error codees returned by av_read_frame are
[14:32] <Tjoppen> grep the headers for AVERROR
[14:57] <juanmabc> i got a flv pts that goes from 0 to 33,67 no 0,1,2,3 what does that mean ?
[14:57] <Tjoppen> variable frame rate. introductory screen?
[14:58] <juanmabc> yep vfr could be it, how could i solve it?
[14:59] <Tjoppen> there's a problem?
[15:00] <juanmabc> i display based in fps (so fixing frame rate is assumed ;O)
[15:00] <Tjoppen> don't? use the pts instead
[15:01] <juanmabc> problem with that it could be NOPTS_VALUE in some files
[15:01] <Tjoppen> look at how ffplay handles that
[15:01] <juanmabc> damn, okay, but it's too mess for me
[15:01] <Tjoppen> ^Fpts
[15:01] <Tjoppen> :)
[15:06] <Tjoppen> else you could just transcode using -r 25, but that feels like the wrong thing to do here
[15:31] <anttipe> Hi! I am developing ffmpeg based player on Android device. I have problems opening https hls urls. I think the problem is somehow related to self-signed certificates. I can open the url on firefox 11.0 on my desktop / Android browser without problems. Cert is self signed. Http urls play fine with android. I use ffmpeg's trunk. Can anyone give direction togo? :)
[16:07] <juanmabc> yep, -r 25 did the trick, i have to restrict support to fixed br until i can solve the ffplay.c sync mess ;P
[16:08] <Tjoppen> fffmpeg's lavf exposes best_effort_timestamp, which works well enough IIRC
[16:09] <sprzybilla> if I am encoding a video that is letterboxed, should I crop and then add padding to not waste any encodeing resources on the letterboxing?
[16:10] <Tjoppen> you should set aspect information correctly, not pad
[16:10] <Tjoppen> if it already has padding then cropping it should help not wasting bits on it
[16:11] <sprzybilla> Tjoppen: so, jsut to clarify, the source has padding, and yes, I don't want to waste any bits, but I do want to keep the padding. Would setting the aspect info prevent bit waste, or will I have to crop then add padding?
[16:12] <Tjoppen> you don't want any padding
[16:13] <Tjoppen> just crop and hopefully the sample aspect ratio will be intact in which case you're good to go
[16:14] <Tjoppen> encode a few seconds and see what ffprobe has to say about the aspect of the resulting file
[16:16] <sprzybilla> will do
[16:44] <na85> hey guys, I'm converting a .vob to quicktime .mov on windows and am seeing a lot of duplicate frames in the output.  Is this bad?/Can I eliminate them?
[16:51] <burek> na85,
[16:51] <burek> can you please use pastebin.com, to show your command line and its output?
[16:54] <na85> sure
[16:54] <na85> one sec
[16:56] <na85> err
[16:56] <na85> i'm used to linux, how do i route the output to a txt file on windows?
[17:01] <na85> burek: here is the paste: pastebin.com/YDg4zAhY
[17:03] <mgeary> hey, i'm strugging with removing letterboxing on a video file. If i just want to take off 50 pixels from the top and bottom, what's the right way to do that with -vf crop=foo?
[17:05] <mgeary> aha. got it: -vf crop=in_w:in_h-100:0:50
[17:20] <juanmabc> Tjoppen:: i am using best effor timestamp but to convert to time i multiply by time_base (1/fps) so it still has to be fixed br
[17:25] <na85> hey guys, I'm converting a .vob to quicktime .mov on windows and am seeing a lot of duplicate frames in the output.  Is this bad?/Can I eliminate them?
[17:25] <na85> pastebin.com/YDg4zAhY
[17:37] <burek> na85, that's a clean output
[17:37] <burek> no errors
[17:43] <na85> burek: yep, but a lot of duplicated frames... is that bad?
[17:43] <na85> or avoidable?
[17:48] <burek> na85, your input is at 30 fps
[17:48] <burek> and output is at 60 fps if i see correctly
[17:48] <burek> so duplicate frames are good in this case
[17:48] <burek> they are the only way to fill up those gaps
[17:49] <na85> ok that makes sense
[17:49] <na85> thanks
[19:56] <niro> Hi, is this the right place to ask why v4l2 changes the driver from 1024x768 to 800x600
[19:56] <Freakshow> can someone clue me in on what dup and drop mean on the output log data?
[19:56] <Freakshow> dup=3468 drop=866305
[19:58] <relaxed> duplicated and dropped frame count
[20:00] <Freakshow> hmm...
[20:00] <Freakshow> great! thanks... I suspected as much, but wanted to be sure
[20:00] <Freakshow> is this actually in the wiki?
[20:01] <relaxed> I'm not sure
[20:01] <Freakshow> I'll double check
[20:31] <tekk> any libavformat guru's in here?
[20:33] <tekk> i need to demux an mkv& to get the h264 and audio streams separate via the C level API&
[20:33] <tekk> to be honest& I'm not sure where to begin
[20:37] <cbreak> tekk: avformat.h? :)
[20:37] <cbreak> tekk: also, ffplay.c does that.
[20:38] <tekk> what is the simplest way to do it ?
[20:38] <tekk> i.e. which would you recommend
[20:38] <tekk> or your suggesting to look at ffplay.c to see how its implemented from avformat?
[20:40] <cbreak> the latter
[20:40] <tekk> i've seen this previously http://libav-users.943685.n4.nabble.com/demux-remux-files-with-libavformat-td945728.html
[20:41] <tekk> but it looks like if you have a 100mb video you need 100mb of ram ?
[21:05] <TACPILOT> can ffmpeg convert yuv420p to yv12 ?
[21:12] <Tjoppen> uhm, yv12 *is* 4:2:0
[21:17] <sprzybilla> does anyone know where I might be able to find an uncompressed movie trailer? Probably a weird question, but I am looking for the new spiderman movie trailer as pristine as possible
[21:18] <sprzybilla> since they are promotional i thought there may be a place to find em
[21:18] <mgeary> hey, where 0.10 defaults to libx264, how do i specify a preset? I want to use -vpre lossless_fast
[21:19] <PurplePeter> sprzybilla: What do you want with it?
[21:20] <sprzybilla> PurplePeter: I am applying for a job, and their prerequsite is to compress that trailer, however, the source they provide is H.264, so I thought I would try and find a better one ;)
[21:20] <sprzybilla> not to mention it doesn't even have CABAC enabled!
[21:23] <sprzybilla> although, the employer is probably trolling this IRC so keep my master compression plan on the down-low
[21:24] Action: sprzybilla throws on shades
[21:26] <PurplePeter> sprzybilla: Good luck, totally-not-Shawn-Przybilla. I won't tell them. Also, I don't know where such a thing could be obtained, my apologies.
[21:36] <sprzybilla> PurplePeter: Zing! Itunes trailers lets you download AND it seems to be a better encode than I was equipped with
[21:38] <sprzybilla> shit it has a watermark - oh welllll
[21:43] <PurplePeter> sprzybilla: Write code that replaces the region with the watermark with the lower quality video you were given -- or, if the water mark is additive, attempt to remove it?
[21:45] <sprzybilla> PurplePeter: hehe, in comparison it's not THAT much better than the original. So it's probably not worth it. If I could find an uncompressed source I might go down that road, but this one is just a bit higher of a bitrate
[21:45] <sprzybilla> still no CABAC either, whats up with apple?!
[21:55] <TACPILOT> Tjoppen: similar but not the same for its layout of U and V are reversed
[21:56] <Tjoppen> you're talking file format then
[21:56] <TACPILOT> though it is still considered a 420 fromat
[21:56] <Tjoppen> you420p is what's in RAM
[21:57] <Tjoppen> AFAIK no codec deal with yv12 directly
[21:57] <TACPILOT> avisynth plugins require yv12
[21:58] <TACPILOT> at least the ones I am needing
[21:58] <Tjoppen> so pack the data as such then
[21:58] <TACPILOT> hence my original question
[21:59] <Tjoppen> you're writing an avfilter?
[22:01] <TACPILOT> i need avchd to yv12 so I can use QTGMC
[22:11] <Aegwyn11> I'm trying to compile FFmpeg with x264 and keep getting an error message that libx264 isn't found. I'm using the same flags that work compiling FFmbc. The flags are: configure --enable-gpl --enable-memalign-hack --enable-runtime-cpudetect --enable-avisynth --enable-nonfree --enable-libx264 --enable-libmp3lame --extra-cflags=-I/usr/local/include --extra-ldflags=-L/usr/local/lib
[22:11] <Aegwyn11> I'm on windows and using the ffmpeg snapshot from today
[22:16] <Mista_D> while checking a file with "-f null" I get 2 errors, but exit status code is "0", shoudn't it be "1" ?http://pastebin.com/LRxb8iZK
[22:25] <Mista_D> still the same even with "-er expolde -"...
[23:09] <Nagy> I seem to be unable to play 1080i FFV1 files in real-time, I've tried ffplay, VLC and my own software, none of it manages it, even though my cpu is only 14% utilized
[23:09] <Nagy> why?
[23:11] <sacarasc> I/O swamped?
[23:11] <Nagy> on a ssd drive?
[23:11] <cbreak> Nagy: what cpu?
[23:11] <Nagy> I'm playing the same file encoded with qtrle, which is 4 times larger
[23:11] <Nagy> Xeon W3520
[23:12] <Nagy> Quad Core 2.67Ghz
[23:19] <cbreak> maybe with hyper threading?
[23:19] <cbreak> then 100% on one cpu is 12.5%
[23:19] <cbreak> so ~14% cpu load
[23:19] <cbreak> or do you mean it uses only 14% of one cpu?
[23:20] <Nagy> its 14% total
[23:20] <Nagy> but according to the task manager no thread is over 50%
[23:27] <Nagy> no cpu*
[00:00] --- Fri Mar 23 2012