[Ffmpeg-devel-irc] ffmpeg.log.20140906

burek burek021 at gmail.com
Sun Sep 7 02:05:01 CEST 2014


[02:11] <gigicarlo> hello
[02:11] <gigicarlo> i am getting crazy ... trying to compile ffmpeg with openjpeg support
[02:11] <gigicarlo> its on os x
[02:11] <gigicarlo> 10.9
[02:11] <gigicarlo> "ERROR: libopenjpeg not found"
[02:13] <gigicarlo> anybody here ??? at all ?
[02:20] <relaxed_> gigicarlo: is it version 1.5.2?
[02:21] <relaxed_> libopenjpeg 2.* isn't supported
[02:21] <gigicarlo> good to know ;-)
[02:21] <gigicarlo> ok .. i will try
[02:22] <gigicarlo> so ....
[02:22] <gigicarlo> the one that is installed with "brew! ... should be ok
[02:23] <gigicarlo> i installed it with cmake ...
[02:23] <gigicarlo> is there an uninstall ??
[02:27] <relaxed_> gigicarlo: maybe you need --extra-cflags="-I/path/to/prefix/include" --extra-ldflags="-L/path/to/prefix/lib" to point to its install prefix. Post your config.log if that doesn't work.
[02:27] <gigicarlo> still looking for 1.5.2 source code
[02:31] <gigicarlo> how do i install openjpeg ....
[02:31] <gigicarlo> its not the usual configure .....
[02:39] <gigicarlo> like that :"./configure --extra-cflags="-I/Users/gigi_mini/tmp/openjpeg-1.5.1-Darwin64" --extra-ldflags="-L/Users/gigi_mini/tmp/openjpeg-1.5.1-Darwin64" --enable-libopenjpeg"
[03:02] <gigicarlo> https://gist.github.com/anonymous/102c697b561ffb9d98a5
[03:02] <gigicarlo> here is my config.log
[03:03] <gigicarlo> i am not sure about "-L/path/to/prefix/lib"
[03:03] <gigicarlo> where to point that
[03:03] <c_14> Where did you install libopenjpeg?
[03:03] <gigicarlo> "pc" ... "la" files ? what should i look for
[03:04] <c_14> You configured it and built it, right?
[03:04] <gigicarlo> i just ran .configure
[03:04] <gigicarlo> make make install
[03:04] <c_14> right, where did it install it to?
[03:05] <c_14> Under UNIX, the default is usually /usr/local/, but I don't know if that directory exists under OSX, you could check?
[03:05] <gigicarlo> "/usr/local/include/openjpeg-1.5/"
[03:05] <gigicarlo> there is the openjpeg.h
[03:06] <gigicarlo> but i dont know what kind of "libs" to look for ...
[03:06] <gigicarlo> "dylib" ??
[03:06] <c_14> Right, the extra-cflags should point to /usr/local/include then. The ldflags should probably point to /usr/local/lib/
[03:06] <c_14> Either .so or .a
[03:06] <c_14> (probably)
[03:06] <gigicarlo> ok
[03:07] <gigicarlo> do you know what that means " --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include/c++/4.2.1"
[03:08] <gigicarlo> Ok .. it worked
[03:08] <c_14> ehhh, the prefix is where it'll be installed
[03:08] <c_14> and the gxx-include-dir is where the c++ includes are located
[03:08] <gigicarlo> its fine ... it worked
[03:09] <gigicarlo> i want to convert a "j2k" mov file to prores
[03:11] <gigicarlo> but it doesnt work ;-(
[03:11] <gigicarlo>  Duration: 01:28:31.60, start: 0.000000, bitrate: 89801 kb/s
[03:11] <gigicarlo>     Stream #0:0(eng): Video: none (AVj2 / 0x326A5641), 1920x1080, 87496 kb/s, 25 fps, 25 tbr, 25k tbn, 25k tbc (default)
[03:11] <gigicarlo>     Metadata:
[03:11] <gigicarlo>       creation_time   : 2014-09-05 10:13:08
[03:11] <gigicarlo>       handler_name    : Apple Alias Data Handler
[03:11] <gigicarlo>       encoder         : Unknown
[03:11] <c_14> Eh, use a pastebin service, please.
[03:11] <gigicarlo> sorry
[03:11] <gigicarlo> https://gist.github.com/anonymous/746f664f0042a48d1da4
[03:12] <c_14> What's the command you're trying to use?
[03:13] <gigicarlo> i did a mistake
[03:13] <gigicarlo> its still an old one
[03:13] <gigicarlo> not the fresh one
[03:18] <gigicarlo> grrrrr
[03:18] <gigicarlo> libavcodec/libopenjpegdec.o: No such file or directory
[03:18] <gigicarlo> the "make" failed
[03:20] <gigicarlo> trying sudo make now
[03:20] <gigicarlo> its quite tricky ....
[03:22] <gigicarlo> do i have to add some dependencies for openjpeg ?
[03:27] <c_14> Did openjpeg compile correctly?
[03:27] <c_14> If yes, it should be fine.
[03:32] <gigicarlo> and what about this ? :"libavcodec/libopenjpegdec.o: No such file or directory"
[03:32] <gigicarlo> this means it did not compile fine .... doesnt it ?
[03:33] <gigicarlo> within ffmpeg
[03:33] <onyx> Hey guys, is it possible to create a time lapse video from a live stream?
[03:34] <c_14> gigicarlo: It might just mean that the linker needs more flags, does the file libopenjpedcec.o exist on your fs somewhere?
[03:34] <gigicarlo> nope
[03:35] <c_14> Can you pastebin your entire configure output + command?
[03:36] <gigicarlo> from openjpeg ?
[03:36] <gigicarlo> or from ffmpeg
[03:36] <c_14> ffmpeg
[03:37] <c_14> onyx: Change the fps to something ridiculously small and then increase it again, or even better, use the select filter to only take every xth frame.
[03:37] <gigicarlo> last try was with  --enable-shared --disable-static
[03:37] <gigicarlo> but thats useless i think
[03:37] <gigicarlo> sudo ./configure --extra-cflags="-I/usr/local/include/" --extra-ldflags="-L/usr/local/lib/" --enable-libopenjpeg
[03:38] <gigicarlo> https://gist.github.com/anonymous/0827c08178cf54c74999
[03:39] <gigicarlo> running make again right now ....
[03:40] <c_14> can you paste your make output if it errors out again?
[03:41] <gigicarlo> takes a while ... its still running
[03:41] <gigicarlo> All ????
[03:41] <gigicarlo> ist really long ...
[03:43] <gigicarlo> thank you for your help ... anyway
[03:45] <gigicarlo> It's 2909 lines ..... is that ok, to paste ????
[03:45] <gigicarlo> https://gist.github.com/anonymous/78fe4a43c480defa93e5
[03:47] <onyx> c_14: could you provide me with an example command to create the time lapse?
[03:48] <gigicarlo> dont understand ....
[03:50] <gigicarlo> "time lapse" ???
[03:51] <onyx> yep gigicarlo Im trying to create a time lapse video from a livestream
[03:51] <gigicarlo> sorry .. no idea
[03:51] <c_14> onyx: ffmpeg -i file -vf select='isnan(prev_selected_t)+gte(t-prev_selected_t\,60)'
[03:51] <c_14> Should grab one frame every 60 seconds.
[03:51] <c_14> gigicarlo: It looks like FFmpeg isn't compiling the object files for libopenjpeg...
[03:52] <gigicarlo> sure ;-)
[03:52] <gigicarlo> but why ....
[03:52] <gigicarlo> libavcodec  .....
[03:53] <onyx> c_14: ok, so how do I set the time, example end result be a time lapse from 5am til 8pm?
[03:55] <c_14> onyx: start the program at 5am and add -t [(8pm-5am)*3600/(output fps * frame grab interval)]
[03:55] <c_14> or something
[03:55] <c_14> My math might be whack.
[03:56] <c_14> gigicarlo: Do the OSX static builds perchance include libopenjpeg?
[03:56] <c_14> I know the linux ones do.
[03:57] <gigicarlo> how do i check that ?
[03:58] <c_14> Download the fitting one and then use ./ffmpeg -codecs
[03:59] <gigicarlo> @fflogger ... but with openjpeg ???
[03:59] <c_14> Have you tried those before? I don't know if they build with openjpeg or not, but they might. It's worth trying.
[04:00] <c_14> But other than that, I have no idea why FFmpeg is failing to build with libopenjpeg, you could always try the mailing list.
[04:00] <c_14> Someone on there might know.
[04:02] <gigicarlo> .. yes i tried before ...
[04:02] <c_14> Ah, ok. Didn't know that.
[04:02] <c_14> As I said, you might want to try the mailing list next because I have no clue.
[04:02] <gigicarlo> and openjpeg is not included
[04:03] <gigicarlo> https://gist.github.com/anonymous/8be5d2ab8ae36cb3a902
[04:09] <jjohn> Hello, guys. I got a question about ffmpeg and hope you guys can help me. http://pastebin.com/BJJ77yJU
[04:09] <HDTime> oh good cause i got a question too about ffmpeg and x264
[04:12] <jjohn> Well, why don't you post it then, HTTime?
[04:12] <jjohn> HDTime+
[04:13] <c_14> jjohn: Can you pastebin the output of ffprobe -show_streams for one of the files with the [Sar..] and for one of the files without?
[04:14] <gigicarlo> c_14 ... thanks a lot ......
[04:14] <gigicarlo> need to rest now ....
[04:14] <gigicarlo> bye
[04:15] <jjohn> Can do, just gimme a sec.
[04:16] <HDTime> I know you wanted yours answered first
[04:16] <HDTime> but heres mine
[04:16] <HDTime> http://forum.doom9.org/showthread.php?t=171086
[04:17] <jjohn> c_14 here's the SAR one: http://pastebin.com/PDxZpTtr
[04:18] <jjohn> and here is the non-SAR one: http://pastebin.com/5jj3k5Aw
[04:21] <HDTime> Any idea about my question? http://forum.doom9.org/showthread.php?t=171086
[04:22] <jjohn> HDTime you did try "cat $file > ffmpeg -i -"?
[04:22] <sacarasc> HDTime: What's your question?
[04:22] <jjohn> Exactly as I wrote it? ffmpeg<SPACE>-i<SPACE>-?
[04:23] <Hello71> useless use of cat
[04:23] <HDTime> that command i put in the first post is what i've used
[04:23] <HDTime> stdin gives the pipe error
[04:23] <jjohn> Ahm, I meant '|'.
[04:23] <HDTime> i havent tried anything else
[04:23] <LanDi> hey guys, I'm a little noob on ffmpeg.... Can I use it to record a internet livestream like NHK world ?
[04:23] <HDTime> what would the command line look like with those settings?
[04:24] <c_14> jjohn: the non-sar files don't have a sample aspect ratio or display aspect ratio set...
[04:24] <jjohn> Well, foxyshadis already suggested to pass the file via command line.
[04:24] <HDTime> i tried his command he suggested and it didnt work
[04:24] <jjohn> c_14 OK, and how to set them?
[04:25] <c_14> jjohn: That's what the setsar/setdar filters are for, but you'll need to reencode to use them.
[04:26] <jjohn> That's what I wanted to avoid - especially because the files are running fine in VLC.
[04:26] <jjohn> Just not merged together.
[04:26] <HDTime> The command i posted works fine for files already fully recorded
[04:27] <HDTime> but ones that are still recording stop at the point of what was airing when i gave ffmpeg the command
[04:28] <jjohn> Can
[04:28] <c_14> You can either try reencoding during the concat step by adding -vf setsar=1:1,setdar=4:3 and not using codec copy, or convert the non-sar with the same -vf or what _might_ work is converting the sar videos with setsar=0:1,setdar=0:1 but I don't about that.
[04:28] <jjohn> Sorry, can't I just get rid of SAR and DAR?
[04:29] <c_14> Maybe if you like having fun with hex editors? But I'm not sure how well that'll work.
[04:30] <jjohn> In fact I use hex editors on a regular basis. :) But I'm not sure I that would realy help.
[04:30] <jjohn> really*
[04:31] <c_14> I can't really see a way around this without reencoding at least some of the files.
[04:32] <jjohn> With -vf setsar=0:1,setdar=0:1, I reencode the videos, I just tested it ...
[04:34] <jjohn> Would the video quality suffer from such an reencoding?
[04:34] <c_14> Depends on the settings you use.
[04:34] <c_14> With what you're trying to do, it'll always be lossy, but it might not be visually lossy.
[04:34] <c_14> ie, use a crf around 18
[04:35] <jjohn> It's still "/opt/local/ffmpeg/bin/ffmpeg -i 15.mp4 -vf setsar=0:1,setdar=0:1 test.mp4"
[04:35] <jjohn> Crf?
[04:35] <c_14> ye -c:v libx264 -crf 18
[04:36] <c_14> Might also want -pix_fmt yuv420p -profile main
[04:36] <c_14> But those 2 should be default iirc.
[04:39] <HDTime> so about my problem?
[04:39] <HDTime> Can you give a command line to use?
[04:42] <jjohn> Is it OK that the tbn of a converted video is just 12800 and not 90k as the non-converted SAR videos?
[04:44] <c_14> The tbn is the time base, I don't think it should break anything...
[04:46] <jjohn> OK, I am reencoding all vids, should take a little while.
[04:47] <c_14> If you're going to reencode them all, you could just use the concat filter instead.
[04:47] <jjohn> No, just the non-SAR ones.
[04:48] <jjohn> I want them all in the exact format and then concat them again, just to be sure.
[04:53] <jjohn> Well, that didn't work out. If I try to concat them now, I get this error: "Non-monotonous DTS in output stream 0:0; previous: 58312800, current: 8292877; changing to 58312801. This may result in incorrect timestamps in the output file." just over and over again.
[04:54] <jjohn> But there is no real difference in the stream information, as far as I call tell. Apart from the increased bitrate of the reencoded files and the different tbn.
[04:55] <jjohn> And the files are running file separately in VLC.
[04:55] <jjohn> fine*
[04:56] <jjohn> are running fine*
[04:56] <c_14> Those messages are (usually) just warnings and not errors, does the output file play correctly if you let the process finish?
[04:58] <jjohn> No, it doesn't. Breaks on the same time stamp as before.
[04:58] <jjohn> Should I paste the stream informations on pastebin?
[04:58] <c_14> Sure.
[05:00] <jjohn> c_14: http://pastebin.com/gHM5Apxz
[05:00] <jjohn> The files beginning with an underscore (_) are the new reencoded ones.
[05:03] <jjohn> So, what'd you suggest I try doing?
[05:04] <c_14> Can you paste the show_streams for 15 and _16 ? I want to see if there's a difference in anything other than the tbn.
[05:04] <c_14> You could also try adding -auto_convert 1 as an input option.
[05:04] <c_14> Just in case it's the h264_mp4toannexb magic missing...
[05:05] <jjohn> Sure thing: http://pastebin.com/bkZ0xwJd
[05:08] <jjohn> Ahm, auto_convert is not found by ffmpeg: "Option auto_convert not found."
[05:09] <c_14> As an input option? ie before the -i ?
[05:10] <jjohn> 'for i in 16.mp4 17.mp4 18.mp4; do /opt/local/ffmpeg/bin/ffmpeg -auto_convert 1 -i "$i" ...' and so on.
[05:10] <rule_2> hey c_14 are you there?
[05:10] <rule_2> i solved my problem from earlier after banging my head against the wall.
[05:11] <jjohn> rule_2 he is and bangs his head on the wall of my problem as well. :D
[05:11] <rule_2> So if you are trying to use V4L2 & ffmpeg to capture input from an HDMI to USB device.. don't use kernel 3.16
[05:11] <rule_2> just avoid :)
[05:11] <rule_2> kernel 3.14.14 is gold
[05:11] <c_14> ok, good to know
[05:11] <rule_2> c_14: not only do I have no more buffer errors, no more V4L errors
[05:11] <rule_2> but i went from 130% CPU to record lossless h.264 in realtime
[05:11] <rule_2> to 32%!!!!
[05:11] <jjohn> 3.16 is broken in more than a way. I cannot have my laptop running a night with it.
[05:11] <rule_2> that's insane
[05:11] <rule_2> 3.16 can suck it!!! :)
[05:12] <jjohn> Without freezing totally.
[05:12] <rule_2> i was actually going to upgrade the CPU in my laptop so I could do two streams.One for my microscope camera and one for my desk camera, since I plan on recording both simultaneously
[05:12] <rule_2> so I can edit a split screen video showing both cameras
[05:12] <rule_2> now I can use this and probably watch porn in the background at the same time with the remaining 30% CPu
[05:13] <rule_2> one thing that worries me is it says fps=8.1 or fps=11
[05:13] <c_14> jjohn: just try adding -bsf:v h264_mp4toannexb as an output option instead
[05:13] <rule_2> but when i load the video into vlc it says 60 fps
[05:13] <c_14> rule_2: That's the encoding fps
[05:14] <c_14> How many frames it's encoding per second
[05:14] <rule_2> if i am capturing at 60 fps, and encoding at 4 fps
[05:14] <rule_2> how does that work?
[05:14] <rule_2> when i hit q on that terminal it isn't still working
[05:14] Action: rule_2 scratches head
[05:14] <rule_2> wouldn't it still have work to do once the capture is done if it is encoding so slow compared to the capture?
[05:14] <jjohn> Now it gives me errors on encoding: "[libx264 @ 0x1a1abe0] Packet header is not contained in global extradata, corrupted stream or invalid MP4/AVCC bitstream"
[05:14] <jjohn> "Failed to open bitstream filter h264_mp4toannexb for stream 0 with codec libx264: Invalid argument"
[05:14] <c_14> Is it encoding realtime? can you play the file while it is recording and never hit the end?
[05:15] <c_14> Hmm, so that won't do it...
[05:17] <c_14> The only difference I can find between the streams is the frame rate (which is so close it shouldn't matter) and the time_base (which also shouldn't matter) at least I think the tbn shouldn't matter.
[05:18] <rule_2> c_14: I am going to record a clock at 60 fps that has seconds and see what i get when i count frames
[05:20] <jjohn> I am no expert in video edition, I gotta depend on your judgement, c_14.
[05:25] <c_14> Hmm, you can try the command again once with -copytb 1 and once with -copytb 0 in case the autodetect is messing up there. (the concat one)
[05:28] <c_14> hmm
[05:29] <c_14> Can you try the concat again with this commandline?: ffmpeg -auto_convert 1 -f concat -i concat.txt -c copy out.mp4 ?
[05:30] <rule_2> c_14: I recorded a one minute long clip of a clock counting. Throughout the clip, I have 60 frames per second as confirmed by me moving frame by frame 60 frames to count 1 second on the clock, and there is video in the file all the way up to the second I cut off the video
[05:30] <rule_2> so if it says it is recording at 11 fps.. what is that all about?
[05:30] <rule_2> if it was, if I cut off the process at 11:28:00, then that means there should be video of the clock hitting 11:28:00, since there'd still be stuff left to encode
[05:30] <rule_2> I guess I'm complaining about something working which is pretty god damn silly :)
[05:31] <c_14> jjohn: If neither the copytb nor the auto_convert work, an alternative could be to remux all the input files to mpeg-ts, concat those and then convert back to mp4.
[05:32] <c_14> (according to this mailing list thread from may.
[05:32] <c_14> )
[05:33] <c_14> rule_2: I have no idea, but as long as it works I guess?
[05:33] <c_14> jjohn: https://ffmpeg.org/pipermail/ffmpeg-user/2014-May/021186.html
[05:33] <c_14> The relevant thread.
[05:36] <jjohn> With that auto_convert command you gave me the DTS "warnings" still show up and I still do get errors, but different ones this time: "[h264 @ 0x7ffac8e843c0] mmco: unref short failure"
[05:37] <jjohn> The copytb 1/0 one - as input or output option?
[05:37] <c_14> output, I think
[05:37] <c_14> yeah, output
[05:38] <jjohn> OK, concating again ...
[05:38] <jjohn> OK, Still the same error. I should try disable auto_convert again ...
[05:39] <jjohn> Without auto_convert and copytb 1 the very first errory appear. I'll try out copytb 0 now.
[05:41] <jjohn> Ok, no matter what I pass for copytb, it does not work out. I will read the mail now.
[05:41] <jjohn> How do I "demux" files exactly?
[05:42] <c_14> That's done automatically.
[05:42] <jjohn> Actually, why should I do that? There's no second stream in the container.
[05:42] <jjohn> There's just video. Nothing more.
[05:42] <c_14> Where does it say anything about a second stream?
[05:43] <c_14> What I found interesting (considering the auto_convert failed) was this part:  AFAIK, nothing short of remuxing to MPEG-TS and concatenating the bitstream.
[05:43] <jjohn> I always thought a demuxer is responsible for reading the different streams in a container - streams which got "muxed".
[05:43] <c_14> The demuxer demuxes any and all streams out of the container.
[05:43] <c_14> In your case the h264 stream from the mp4 container.
[05:44] <jjohn> Hm, OK. But I already did /opt/local/ffmpeg/bin/ffmpeg -auto_convert 1 -f concat -i v_list -c copy test.mp4
[05:44] <c_14> But back to the mpegts, try converting all the files to mpegts, ffmpeg -i file -c copy file.mpegts, and then just cat those together.
[05:45] <jjohn> It does recognize the format by the file extention, right?
[05:45] <c_14> You might need to add -bsf:v h264_mp4toannexb, but ffmpeg will spit out an error telling you to (usually).
[05:45] <c_14> Not really.
[05:45] <c_14> There are magic bits in the beginning of the file.
[05:46] <jjohn> OK, gimme a second.
[05:49] <jjohn> No, does not work, neither with nor without the -bsf option: [NULL @ 0x1a15560] Unable to find a suitable output format for '_16.mp4.mpegts'
[05:49] <jjohn> _16.mp4.mpegts: Invalid argument
[05:49] <jjohn> I did: for i in _1*.mp4; do /opt/local/ffmpeg/bin/ffmpeg -i "$i" -c copy -bsf:v h264_mp4toannexb "$i.mpegts" ; done
[05:49] <c_14> add -f mpegts before it?
[05:49] <c_14> ie -f mpegts "$i.mpegts"
[05:49] <onyx> any idea why this command fails? ffmpeg -t 6 -i http://liveurl -f image2 -vf fps=fps=6/60 images%04d.jpg
[05:49] <jjohn> No, trying so.
[05:50] <onyx> I mean it works but it doesnt do what Im trying to do...
[05:50] <c_14> it usually does output format detection via extension, maybe it just doesn't like that one...
[05:50] <jjohn> OK, that worked out nice. I got all the mpegts files now.
[05:51] <c_14> cat *.mpegts > concat.mpegts
[05:51] <jjohn> I also tried just "test.mpegts", wasn't recognized either.
[05:51] <c_14> Then pray to your favorite deity and try playing it.
[05:52] <onyx> I mean it works but it doesnt do what Im trying to do...
[05:52] <c_14> onyx: It should give you 14.4 frames
[05:52] <c_14> If my math is right.
[05:52] <c_14> Assuming the input is 24fps
[05:52] <c_14> And I think it rounds down.
[05:52] <c_14> So 14 frames.
[05:52] <jjohn> c_14 ... the raw bitstream is running.
[05:52] <onyx> no, it actually it only gave me me 2 frames
[05:53] <onyx> both about the same second too
[05:53] <c_14> So the concat.mpegts works? just convert it to mp4 then
[05:53] <onyx> Im trying to capture 6 images per minute for a period of time
[05:54] <c_14> did you try that thing with the select that I posted earlier?
[05:54] <onyx> no it went way over my head
[05:54] <jjohn> /opt/local/ffmpeg/bin/ffmpeg -i concat.mpegts -c copy test_stream.mp4?
[05:55] <jjohn> That worked, but I am not sure I did it as you wanted me to do it.
[05:55] <c_14> As long as it works, does it matter?
[05:56] <onyx> c_14: what was that command you posted earlier?
[05:56] <c_14> onyx: scroll down to examples: https://ffmpeg.org/ffmpeg-filters.html#select_002c-aselect
[05:56] <c_14> The one with the isnan
[05:56] <jjohn> FUCK YEAH! The video is not distorted now.
[05:57] <jjohn> OK, not entirely. That euphorism might have been a bit too early.
[05:58] <jjohn> You know I told you there are two files with original SARs and DARs? Well, the video does not break on 10:47 anymore ...
[05:58] <c_14> It breaks before the last one?
[05:58] <jjohn> But on something like 42:xx ... the time at which the last video.
[05:58] <jjohn> starts. Yeah.
[05:59] <jjohn> On exact the last one. But the transition from the first to the second one is fine.
[05:59] <c_14> Only in the mp4 or also in the mpegts?
[06:00] <jjohn> Lol I didn't convert the original SAR files in mpegts, just the reencoded ones. Converted that stream into a mp4, and then concated as usual.
[06:00] <jjohn> I will try that out now.
[06:01] <onyx> c_14: this one right?
[06:01] <onyx> select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)'
[06:01] <c_14> yep
[06:02] <c_14> The 10 should be good for you as well.
[06:02] <onyx> yeah
[06:02] <c_14> One frame every 10 seconds, ie 6 frames per minute.
[06:02] <onyx> yep
[06:02] <onyx> now how would I integrate that to the command?
[06:03] <onyx> like this? ffmpeg -i http://liveurl select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)' images%04d.jpg
[06:03] <c_14> ffmpeg -t 60 -i http://fooobar -vf filterfoo -f image2 images%04d.jpg
[06:03] <jjohn> OK ... this time, I used ALL THE ORIGINAL files (not the three encoded ones), demuxed them, concated the bitstreams and muxed it back into a mp4. And it worked ...
[06:04] <c_14> Yay for mpegts?
[06:04] <jjohn> No distortion. But the VLC now tells me for each "break" "Reinit context to 640x480, pix_fmt: 0".
[06:04] <jjohn> But you know ... currently I am just glad it works somehow.
[06:04] <jjohn> :D
[06:05] <jjohn> I guess I will do another attempt with the reencoded ones, but I will take a short break before that. It's 6:04 AM where I live, and I should get a little bit of sleep.
[06:05] <onyx> c_14:  the -t 60 means that it will create a total of 60 images?
[06:05] <c_14> nah
[06:05] <c_14> It'll record 60 seconds of video.
[06:05] <c_14> Ie 6 images
[06:06] <c_14> or maybe 5
[06:06] <c_14> depending on rounding foo
[06:06] <onyx> oh I get it
[06:06] <onyx> ok let me try that
[06:07] <onyx> [NULL @ 0x1e5a1c0] Unable to find a suitable output format for 'image2'
[06:08] <jjohn> Fuck it, I lied. I just did the attempt with the three converted videos, works like a charm without any problems from VLC.
[06:08] <onyx> image2: Invalid argument
[06:08] <c_14> eh, just get rid of that part
[06:08] <onyx> ok
[06:08] <jjohn> c_14, I thank you. I learned a hell of stuff today.
[06:09] <c_14> np
[06:09] <jjohn> Thank you so much. I think I might take on the next few videos as well.
[06:09] <c_14> Next time I'll just use mpegts from the get-go.
[06:09] <onyx> At least one output file must be specified
[06:09] <jjohn> Good night, then. Thanks again.
[06:10] <c_14> onyx: what's your current commandline?
[06:10] <onyx> ffmpeg -t 60 -i http://liveur -vf select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)' -f images%04d.jpg
[06:11] <c_14> eh, get rid of the -f
[06:12] <onyx> ok
[06:12] <onyx> seems to be working
[06:13] <onyx> mmm... nope
[06:14] <onyx> it got like a 100 jpgs
[06:14] <c_14> is the -t working correctly?
[06:14] <onyx> yeah it ran for 1 min
[06:14] <c_14> add -r 1 as an output option
[06:15] <onyx> right before the images.jpg?
[06:15] <c_14> ye
[06:18] <onyx> now there is 51 files
[06:18] <c_14> Make that -r .1
[06:19] <onyx> ok trying
[06:20] <c_14> .1 might be wrong, let me check my maths
[06:20] <c_14> nah, .1 should be it
[06:21] <onyx> now there is 6 files
[06:21] <onyx> wierd 4 images have the same creation time stamp
[06:22] <onyx> and 2 have the same creation time stamp
[06:22] <c_14> Just the same time stamp or also the same content?
[06:23] <onyx> idk, hold on let me check
[06:29] <Redb3ard_> Hi, I have a m3u8 file which lists a single rtmp url. This file plays in VLC. Is it possible to play it with ffplay or transcode it with ffmpeg? I have the file wrong or the command line syntax wrong, because I get error messages like this: test.m3u8: Invalid data found when processing input
[06:30] <onyx> no they are all different
[06:31] <c_14> onyx: Well, then it looks like it's working.
[06:31] <onyx> yeah,
[06:32] <c_14> Oh, and just in case you ever need to change how many frames you want. The value in the select is how many seconds between frames and the -r is how many frames per second you want.
[06:32] <onyx> is the a way to -t 5h?
[06:32] <onyx> like 5 hours?
[06:32] <c_14> -t 5:00:00
[06:32] <onyx> ok great!
[06:33] <c_14> And with that I'm going to go pass out.
[06:33] <c_14> \o
[06:33] <onyx> thanks man!
[07:13] <Redb3ard_> Anyone awake?
[07:17] <LanDi> lol
[07:21] <HDTime> yes
[07:21] <HDTime> and i still need help
[07:21] <HDTime> two hours later...
[07:22] <HDTime> [10:22] <HDTime> Any idea about my question? http://forum.doom9.org/showthread.php?t=171086
[07:24] <Redb3ard_> Well, it's hard to tell sometimes. It is a Friday night, after all.
[07:25] <Redb3ard_> Sorry HDTime. I've only just installed ffmpeg today.
[07:26] <HDTime> its a good command line for encoding though there are some bugs
[07:26] <HDTime> but its pretty good
[07:27] <Redb3ard_> I mostly use Plex, and I know it has ffmpeg buried in it somewhere.
[07:27] <Redb3ard_> The idea that I could make m3u8 files of streams, put those in Plex, and get it to play live streams is tempting... but I don't think it likes that.
[07:28] <HDTime> i use staxrip which has it buried somewhere probably
[07:28] <Redb3ard_> I can get it to play/transcode the stream urls directly, but not from the file.
[07:28] <HDTime> i cant encode live from a file that is still recording from TV
[07:28] <HDTime> it stops at what is airing when i start encoding
[09:27] <roughana> hi have been trying to compile ffmpeg on QNAP and have had a segmentation fault. not sure what I did wrong and a bit out of my depth. Hoping someone could help me unwind the mess and start again
[09:31] <roughana> I've been following this thread on what to do: http://forum.qnap.com/viewtopic.php?f=50&t=52358
[09:32] <roughana> someone else suggested that the bottom part of that thread that deals with changing /etc/ld.so.conf was not best practice and I should be using another paramter to configure
[09:33] <roughana> we're not quite sure what the paramater change should be though. guessing --extra-ldflags='-L/opt/lib -Wl,-rpath/opt/lib'
[09:34] <roughana> does the -Wl part of that look ok?
[09:36] <roughana> I just tried to re-issue ./configure ... and received error:
[09:36] <roughana> gcc is unable to create an executable file.
[09:36] <roughana> C compiler test failed.
[09:36] <roughana> etc
[09:36] <roughana> it seemed to work last time, so I'm wondering if I should clear out the directory and start again
[09:38] <roughana> anyone?
[09:47] <Zahlii> Hello there, I am having a problem with silence detect on a mp3 with libmp3lame I asked a question on stackexchange including the CLI command and console output and the results: http://stackoverflow.com/questions/25697596/ffmpeg-silence-detect-with-libmp3lame-queue-is-empty
[09:54] <roughana> Just had a look in config.log and found this
[09:54] <roughana> gcc -L/opt/lib -Wl,-rpath/opt/lib -o /tmp/ffconf.YeZReqfa /tmp/ffconf.yJP1GaVn.o
[09:55] <roughana> ....    /bin/ld: bad -rpath option
[09:55] <roughana> collect2: ld returned 1 exit status
[09:55] <roughana> C compiler test failed.
[09:55] <roughana> which tells me that the option was incorrect.
[09:56] <roughana> ok, I'll try with a couple of other ways and see if there's any progress
[10:02] <roughana> ./configure worked when I gave it --extra-ldflags='-L/opt/lib -Wl,-rpath=/opt/lib'
[10:02] <roughana> make is running now
[10:23] <roughana> AS      libavcodec/arm/aacpsdsp_neon.o
[10:23] <roughana> gcc: Internal error: Segmentation fault (program as)
[10:23] <roughana> make: *** [libavcodec/arm/aacpsdsp_neon.o] Error 1
[10:23] <roughana> any ideas?
[10:25] <relaxed_> roughana: --enable-rpath
[10:27] <relaxed_> wait
[10:30] <relaxed_> which distro and arch?
[10:31] <roughana> ./configure --arch=arm --enable-armv5te ...
[10:31] <roughana> git://git.videolan.org/ffmpeg.git
[10:32] <relaxed_> is that a mirror of git://source.ffmpeg.org/ffmpeg.git ?
[10:33] <roughana> I don't know.
[10:33] <relaxed_> then use git://source.ffmpeg.org/ffmpeg.git
[10:35] <roughana> http://git.videolan.org/?p=ffmpeg.git;a=summary
[10:35] <relaxed_> ok, that looks fine
[10:37] <roughana> so I should add --enable-rpath to the ./configure arguments?
[10:38] <roughana> does the order of the arguments make any difference?
[10:39] <relaxed_> I don't think so. Are you cross compiling?
[10:39] <roughana> no
[10:39] <roughana> (well, not that I know about)
[10:39] <roughana> I am compiling on the box that should run it
[10:40] <relaxed_> what are you passing to configure?
[10:40] <roughana> ./configure --arch=arm --enable-armv5te --prefix=/opt --extra-cflags='-I/opt/include' --extra-ldflags='-L/opt/lib -Wl,-rpath=/opt/lib' --enable-static --disable-shared --disable-ffplay --disable-ffserver --enable-libmp3lame
[10:42] <relaxed_> remove --arch=arm --enable-armv5te
[10:42] <relaxed_> make distclean
[10:43] <roughana> add --enable-rpath ?
[10:43] <relaxed_> yes
[10:45] <roughana> oh make distclean before running ./configure?
[10:45] <roughana> or after?
[10:45] <relaxed_> before
[10:46] <roughana> ah. ok
[10:46] <roughana> ./configure working
[10:47] <roughana> well, that was ambiguous.
[10:47] <roughana> processing
[10:51] <Zahlii> relaxed_ , do you have an idea why silencedetect detects the silence in a mp3, but does not remove it saying "Trying to remove 1152 samples, but the queue is empty"
[10:53] <relaxed_> Is that the purpose of the filter? To remove it?
[10:55] <Zahlii> Uh. I suppose so :o
[10:55] <relaxed_> I don't think so
[10:56] <Zahlii> Hm. That means I need to use the output of silence detect to tell my program to cut out the seconds?
[10:56] <relaxed_> yes
[10:56] <Zahlii> Alright, maybe I should read the doc more carefully :x
[10:57] <roughana> ./configure completed. make processing
[11:06] <vlatkozelka> hi , im trying to record a live udp stream containing more than one Program . The way im doing it atm is open a new process for each Program and do -map 0:p:<PID>   ... is that wrong , should i do it with one process ?
[11:07] <Zahlii> relaxed_ , which would be the best way to cut several parts out of the audio? I'd try using -ss first_silence_end -t (last_silence_start-first_silence_end), but this will only remove the silence at the start or end.
[11:16] <relaxed_> I guess you'll have to do multiple passes
[11:17] <relaxed_> vlatkozelka: -map 0 should include everything
[11:18] <vlatkozelka> yeah i just tried
[11:18] <vlatkozelka> -map 0: ... first output -map 0: ... 2nd output
[11:18] <vlatkozelka> but
[11:18] <vlatkozelka> is that anyway better than first way ?
[11:19] <vlatkozelka> which is to run a seperate ffmpeg process for each one
[11:19] <relaxed_> ffmpeg -i input -map 0 output
[11:20] <vlatkozelka> i think i should wxplain better
[11:20] <relaxed_> indeed
[11:20] <roughana> make failed
[11:20] <roughana> <roughana> AS      libavcodec/arm/aacpsdsp_neon.o
[11:20] <roughana> <roughana> gcc: Internal error: Segmentation fault (program as)
[11:20] <roughana> <roughana> make: *** [libavcodec/arm/aacpsdsp_neon.o] Error 1
[11:20] <roughana> same problem
[11:21] <vlatkozelka> my first method is : ffmpeg -i input -map 0:p:1202 output.ts ... then another one ffmpeg -i input -map 0:p:1201 output.ts
[11:21] <roughana> Please submit a full bug report.
[11:21] <roughana> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
[11:21] <roughana> I guess that's the next step.
[11:21] <vlatkozelka> 2nd method :ffmpeg -i input -map 0:p:1201 output1.ts -map 0:p:1202 output2.ts
[11:22] <vlatkozelka> 1st one is a ffmpeg process for each program , 2nd is all programs recorded in same ffmpeg process
[11:23] <vlatkozelka> the thing is i need to implement that in a java app .. so re-writing that to the 2nd method will take time , so im asking if 2nd method is better or not
[11:25] <relaxed_> roughana: make clean;make distclean and add --arch=armv5te --cpu=armv5te to configure. If that fails you can use --disable-neon
[11:29] <roughana> configure processing
[11:32] <roughana> make processing
[11:51] <roughana> same error
[11:51] <roughana> ok, add --disable-neon
[11:58] <relaxed_> roughana: which distro are you using?
[12:30] <Zahlii> Alright, relaxed_, my program is working now, thank you!
[13:23] <roughana> ok, compile has completed without error
[13:23] <roughana> hope that works.
[13:24] <roughana> thanks relaxed_ for your help
[15:49] <ikarus> quick question is there any lossless codec with higher compression then FFV1 on animated/CG video, that's included in ffmpeg, thinking perhaps Snow or Dirac ? but I am not at a machine that is able to test it right now
[15:54] <JEEB> long GOPs and slow preset x264
[15:54] <JEEB> that should do it
[15:54] <ikarus> JEEB: lossless, so x264 is not my favorite
[15:55] <JEEB> ikarus, x264 can do lossless just fine
[15:55] <JEEB> just set quantizer to zero :P
[15:55] <JEEB> that enables lossless coding
[15:56] <ikarus> JEEB: I am not sure if Dirac or FFV1 won't outperform it then
[15:56] <JEEB> dirac and snow won't outperform it for sure, and I'm pretty sure ffv1 won't either
[15:57] <ikarus> JEEB: in lossless mode, interesting
[15:57] <ikarus> performance isn't an issue, because I record in ffvhuff, then do a nightly transcoding
[15:58] <JEEB> ffmpeg -i input -c:v libx264 -q:v 0 -preset veryslow (or placebo) -g (whatever value is correct for "infinite")
[15:58] <JEEB> and then output
[15:58] <JEEB> I wonder if it was -g -1
[15:58] <JEEB> dunno :V
[15:59] <ikarus> JEEB: I guess I'll have to test suite it at some point myself then
[15:59] <JEEB> it should output the x264 parameters set in the encoder after it initializes the encoder, so you should see
[15:59] <JEEB> you need to have qp=0 and keyint=infinite
[15:59] <JEEB> that gives you basically maximum compression
[16:00] <JEEB> because x264 will only put intra frames where it thinks it's most efficient, and then qp=0 is lossless mode
[16:00] <JEEB> (the only case in which you want to use quantizer)
[16:00] <ikarus> JEEB: keyint infinite is going to piss off my NLE to no end (sorry, I guess I should have mentioned I need to edit the resulting footage at some point)
[16:00] <JEEB> well ffv1 already does long'ish GOPs IIRC, you should see what is the maximum you can take :P
[16:01] <ikarus> I know ffv1 GOP length works (I am using it now as a stopgap)
[16:01] <JEEB> well, then you could use something similar
[16:02] <JEEB> this reminds me that a guy I know once made a test for lossless coders with quick seek in mind
[16:05] <JEEB> don't remember the original URLs and this repost doesn't seem to contain the last version :V
[16:06] <JEEB> http://www.animemusicvideos.org/forum/viewtopic.php?f=46&t=101581&start=90#p1381176
[16:06] <JEEB> and this was in 2011 :D
[16:07] <JEEB> 3 frame GOP without threads and preset ultrafast is ~3x ffv1's encoding speed yet still beat it in compression :3
[17:32] <Redb3ard_> Good morning.
[19:24] <Redb3ard_> I have an m3u8 file containing a single url to streaming audio (http://2583.live.streamtheworld.com/KFMXFMAAC). This m3u8 file will play in VLC. ffplay will (with warnings) play the url directly, but it errors out if I try to play the m3u8 file... is it not possible for it to play the file? Or for ffmpeg to transcode using the m3u8 as input?
[20:27] <aleb> I'm trying to use -vf ...;color=enable='between(t,2.5,8)',black:100x200;... but I get: "Timeline ('enable' option) not supported with filter 'color'" - any idea how I can make it work?
[21:58] <pmart> i've already compiled ffmpeg but now i want to recompile adding new library. does it mean i have to recompile /everything/?
[21:59] <JEEB> you will have to recompile FFmpeg, but if you still have the old dir structure, then you don't necessarily have to compile all of FFmpeg again
[21:59] <JEEB> but yes, you need to recompile FFmpeg again
[21:59] <pmart> last time i've noticed that after running ./config for second time every source file is recompiled
[22:00] <pmart> (with other options)
[22:00] <JEEB> only what would need to be compiled in general, there might be other reasons for everything to get re-compiled
[22:01] <pmart> it would a lot more pleasent if i could find a way to do it the smart way
[22:02] <pmart> because compilation is quite long'
[22:02] <JEEB> in general I would say it should only recompile what is needed :P
[22:02] <JEEB> also if you have multiple cores make -jX is your friend
[22:03] <JEEB> where X is how many things you want to compile at the same time
[22:05] <pmart> so you can confirm that after running ./config for the second time in the same source tree (with object files from previous compilation) make didn't recompile everything?
[22:05] <pmart> i'll try with -j2
[22:10] <Redb3ard_> Just run make already.
[22:11] <Redb3ard_> This is a problem that they were well aware of 40 years ago and solved in the early days of unix.
[00:00] --- Sun Sep  7 2014


More information about the Ffmpeg-devel-irc mailing list