[Ffmpeg-devel-irc] ffmpeg.log.20140220

[burek]

[00:03] <BtbN> Is there any less cpu intensive useable deinterlacer in ffmpeg?
[00:07] <klaxa> as opposed to what?
[00:07] <BtbN> yadif
[00:07] <BtbN> it's just too much at 1080i
[00:07] <klaxa> hmm... i don't think there is
[00:08] <klaxa> what you could try is -vf fieldmatch,decimate
[00:08] <BtbN> what does that do?
[00:08] <klaxa> basically the same, it matches fields (interlaced frames) and removes duplicate frames
[00:08] <klaxa> at least that is how i understand it
[00:09] <klaxa> i use it for .ts files from TV streams
[00:09] <klaxa> i might use even more cpu though
[00:09] <klaxa> worth a try though maybe?
[00:09] <BtbN> than yadif? oO
[00:10] <BtbN> it uses like one full core on its own
[00:10] <klaxa> how do you know?
[00:10] <BtbN> because it does?
[00:11] <klaxa> where do you take that information from?
[00:11] <BtbN> the cpu usage?!
[00:11] <rjp421> anyone on windows using -f dshow -i "WebcamMax Capture"? im using the latest ffmpeg 2.1.3 win64 build, and keep getting "real-time buffer not full, frame dropped" errors.. i dont want to have to set -rtbufsize to 2 gig, and -r 10000/1001 before -i stopped the errors, but produced 10fps video.. i tried 24000/1001 for 24fps but still got the error flooding the screen.
[00:12] <klaxa> but you must be re-encoding if you are using yadif so i wouldn't think yadif is using up most of the cpu usage
[00:12] <BtbN> it clearly does, because without it it's not using one full core on its own...
[00:16] <Mavrik> yadif does burn alot of CPU
[00:16] <Mavrik> but it's pretty much the best thing you have
[00:16] <BtbN> well, it's worth nothing if the cpu can't keep up
[00:16] <Mavrik> hmm,
[00:16] <Mavrik> just how crappy is your CPU then_
[00:16] <BtbN> i3 3220
[00:17] <Mavrik> even at it's worst, yadif doesn't really go over 10% of cpu usage for x264 encode
[00:17] <BtbN> 1080i25 is quite taxing
[00:17] <BtbN> the output format doesn't matter for the deinterlacing
[00:17] <Mavrik> *shrug*
[00:18] <Mavrik> we have no problems doing 1080i/50 -> 1080p/25
[00:18] <Mavrik> 4-5 streams or so on a quad
[00:18] <BtbN> one is the max
[00:19] <BtbN> how do i pass a named parameter to a filter?
[00:22] <llogan> which filter?
[00:23] <BtbN> any filter, the parameters all have names, but i haven't found out how to set them by name, only via a : seperated value list
[00:23] <Mavrik> filter=parameter=bla:paramter2=bla2
[00:24] <BtbN> huh, it just errord to me about that, that it wouldn't know the parameter interl at the scale filter
[00:27] <BtbN> scale=trunc(oh*a/2)*2:min(720\,ih):interl=1: Invalid argument
[00:27] <BtbN> that's what i get
[00:29] <BtbN> nope, also doesn't work if i name them all
[00:36] <BtbN> i don't get it, according to google that's how it should just work
[00:37] <BtbN> "scale=2*iw:2*ih:interl=1" even the docs say it, but it doesn't work
[00:41] <BtbN> even just scale=w=1280:h=720 fails with Invalid argument.
[00:43] <BtbN> it's using it as a filename, what?
[00:44] <renihs_> thanks for the help! gn8
[00:47] <Mavrik> BtbN, make sure you're properly escaping stuff for your shell.
[00:47] <BtbN> the string is printed by ffmpeg correctly and as one piece, i don't think that's the issue
[00:48] <BtbN> got it working now without named params, but encoding interlaced video seems to just fail. Only get green output
[01:07] <llogan> without your actual command and console output we can not help
[01:19] <BtbN> llogan, seems like libx264 or something in ffmpeg can't handle interlaced video propperly. It's not very well tested i guess
[04:55] <aji> #ffmpeg
[05:01] <sacarasc> This is #ffmpeg.
[08:21] <njoyard> Hello everyone
[08:21] <njoyard> I'm looking into using ffmpeg to transcode videos on the fly to serve to a HTML5 player
[08:22] <njoyard> As browsers do byte-based Range requests to the web server, I'm wondering if there is a way (under certain codec conditions of course) to determine/enforce the total transcoded file size using ffmpeg ?
[08:24] <njoyard> (if possible I would like to avoid waiting for the whole video file to be transcoded before being able to serve it)
[08:36] <hendry> hi, i'm trying to make an mp4, given a bunch of jpegs, is this the correct way of going about it? ffmpeg -r 1/10 -i $tmpdir/%03d.jpg -c:v libx264 -r 30 -pix_fmt yuv420p out.mp4
[08:36] <zap0> tias
[08:44] <hendry> it seems to most a whole bunch of frames though. what am I missing in this log? http://ix.io/aCG
[08:59] <Keshl> hendry: One thing I notice, the -s is wrong. you want -vf scale=360:240.
[08:59] <Keshl> I forget what -s is. I'm pretty sure it's "seek ahead inaccurately to nearest keyframe", though.
[09:07] <hendry> i think the problem i have is the images are in different resolutions https://trac.ffmpeg.org/wiki/Create%20a%20video%20slideshow%20from%20images
[09:08] <hendry> i'm not quite sure how I get the images all to 1440x1080 http://ix.io/aCJ
[09:10] <Keshl> Yes, that actually /is/ the issue. I've run into it before. It'll take the resolution of the first frame, then only encode identically-sized frames. ANnoying as heck. <.<
[09:10] <Keshl> I'm not sure of any fast way to fix it either. I ended up doing it by hand with Gimp. Took like five minutes for ~300 frames. (Not that hard to do it /really/ fast if you're good with keyboard shortcuts.)
[09:11] <hendry> been trying imagemagick, neither -resize 1440x1080 not -scale 1440x1080 works
[09:22] <hendry> Keshl: think I nailed it http://ix.io/aCM
[09:23] <Keshl> Shiny. -É-.
[09:54] <brontosaurusrex> how do i say what fps a y4m should have?
[09:54] <brontosaurusrex> input is png sequence
[10:17] <zap0> brontosaurusrex, http://ffmpeg.org/ffmpeg.html   very first mention of 'fps'
[10:20] <brontosaurusrex> zap0, yeah, that doesn't seem to work
[10:21] <brontosaurusrex> i did -r 24 before input and -r 24 before output and it still seems like 25 fps files
[10:21] <brontosaurusrex> unless y4m has no fps info
[10:28] <relaxed> y4m should have the fps
[10:29] <relaxed> oh, -r after the input will set its framerate
[10:30] <relaxed> $(sed q input.y4m) should show you
[11:37] <zap0> brontosaurusrex, y4m is just raw video in'it ?
[11:38] <brontosaurusrex> zap0, mostly
[11:38] <brontosaurusrex> should have a simple header
[11:38] <zap0> brontosaurusrex, maybe it has no fps
[11:38] <brontosaurusrex> it has resolution
[11:39] <brontosaurusrex> well, i got the pipe to work
[11:39] <brontosaurusrex> ffmpeg -r 24 -i big_buck_bunny_%05d.png -pix_fmt yuv420p -r 24 -f yuv4mpegpipe - 2> /dev/null | x265 --crf 20 -o bbb.hevc - --y4m
[11:39] <brontosaurusrex> i think
[11:39] <brontosaurusrex> and it looks like x265 knows the framerate now
[11:40] <brontosaurusrex> "y4m  [info]: 1920x1080 24/1 fps i420, unknown frame count"
[11:40] <zap0> brontosaurusrex http://wiki.multimedia.cx/index.php?title=YUV4MPEG2
[11:40] <brontosaurusrex> can't open the browser for the next few hours i suppose
[11:41] <brontosaurusrex> 100% cpu at 2.3 fps < encoding
[11:41] <zap0> set priority of that to low.
[11:41] <brontosaurusrex> how to do that on nix?
[11:41] <zap0> nice
[11:42] <brontosaurusrex> yes, but i can't change the priority now i guess?
[11:42] <zap0> why not?
[11:43] <active8> OT: anyone know how to fix ++ WARN: [mpeg2enc] Frame height won't split into two equal field pictures... (piped from a jpeg2yuv) ?
[11:44] <zap0> stop using odd height vale
[11:44] <zap0> value/
[11:49] <brontosaurusrex> yeah, htop + f8 did something
[11:50] <active8> zap the conversion came from a gm / imagemagic conversion that specified a width just to keep the aspect ratio. I might need to address the prob that way. maybe crop a line/px
[11:50] <javisepe> Hi! Could you help me with a compiling process?
[11:51] <javisepe> cross-prefix=arm-unknown-linux-gnueabi-
[11:51] <javisepe> arm-unknown-linux-gnueabi-gcc is unable to create an executable file.
[11:51] <javisepe> C compiler test failed.
[11:51] <javisepe> If you think configure made a mistake, make sure you are using the latest
[11:51] <javisepe> version from Git.  If the latest version fails, report the problem to the
[11:51] <javisepe> ffmpeg-user at ffmpeg.org mailing list or IRC #ffmpeg on irc.freenode.net.
[11:51] <javisepe> Include the log file "config.log" produced by configure as this will help
[11:52] <javisepe> solving the problem.
[11:52] <javisepe> cross-prefix=arm-unknown-linux-gnueabi-
[11:52] <javisepe> arm-unknown-linux-gnueabi-gcc is unable to create an executable file.
[11:52] <javisepe> C compiler test failed.
[11:52] <javisepe> If you think configure made a mistake, make sure you are using the latest
[12:11] <active8> hendry: image magic is posed to be less efficient than grahics magic. I think i used im. one or the other has a utility and syntax to specify the width or height only and maintain aspect ratio. right now I'm using gm and it works great but mpec2enc is warning about Frame height won't split into two equal field pictures.
[12:25] <active8> ffmpeg -r 1/10 -i $tmpdir/%03d.jpg -c:v libx264 -r 30 -pix_fmt yuv420p out.mp4
[12:26] <active8> is -r 1/10 just a cool way to say one picture every 10 sec?
[12:32] <brontosaurusrex> active8, i guess, that would be 0.1 fps
[12:33] <active8> that was a cool trick i didn't seen in any of the 1000 tabs open IIRC (or I don't RC)
[12:35] <active8> and I would think the -pix_fmt yuv420p bit should come before -i, but what is that? not in man. is it a way to say "this is what is getting encoded into the mp4?
[12:36] <brontosaurusrex> this is what we will convert into before feeding x264
[12:37] <active8> brontosaurus: -r is fps and it's in Hz which is cycles per second; so yes, you are correct
[12:38] <active8> and x264 is the default mp4 or only. i have to read up on those guys. lots of forums with alphabet soup today, h264 libx faac* vs whatever. heh heh what fun with a kick ass program like ffmpeg
[12:39] <active8> s/only/omly one/
[12:39] <active8> and s/omly.only. oh boy
[12:39] <active8> never mind
[12:54] <njoyard> javisepe: are you sure arm-unknown-linux-gnueabi-gcc is in the PATH and runs ? did you compile a test hello world with it ?
[13:59] <brontosaurusrex> JEEB, how do i mux raw hevc with m4a (aac) audio with l-smash?
[14:01] <JEEBsv> m4a is already muxed so use remuxer for that
[14:01] <JEEBsv> muxer is for raw streams
[14:02] <JEEBsv> (patches welcome to improve it)
[14:02] <Paranoialmaniac> :)
[14:06] <sspiff> I'm having trouble opening a teletext subtitle in my code
[14:08] <sspiff> avcodec_find_decoder(AV_CODEC_ID_DVB_TELETEXT) returns 0 for me
[14:08] <sspiff> and fmt_ctx->streams[i]->codec->codec_id == AV_CODEC_ID_DVB_TELETEXT is true.
[14:09] <sspiff> I recompiled --enable-decoder=libzvbi_teletext explicitly mentioned, but it still won't work
[14:10] <sspiff> anyone here know what I could be doing wrong?
[14:24] <brontosaurusrex> JEEBsv, ok
[14:52] <rjp421> JEEB, ive currently ended up with:  ffmpeg -f dshow -i video="WebcamMax Capture":audio="Microphone (ManyCam Virtual Mic" -c:v libx264 -crf 23 -preset slow -r 24 -profile:v main -maxrate 554k -bufsize 1108k -vf "scale=640:-1,format=yuv420p" -g 48 -threads 0 -c:a libvo_aacenc -b:a 96k -ac 1 -ar 44100 -sn -f flv "rtmp://1.11002.fme.ustream.tv/ustreamVideo/11002/channelStreamKey flashver=FME/3.0\20(compatible;\20FMSc\201.0) live=1"
[14:54] <rjp421> it works but wont go above 554k... that encoding for streaming page said to only count the video bandwidth
[14:55] <rjp421> unless the output is only showing the outbound video bitrate
[16:46] <Demon_Fox> It would seem that celt and opus get nearly the same result
[16:47] <Demon_Fox> though opus uses celt and silk, iirc
[17:01] <Na_Klar> After tons of reading about mapping, I think to understand how it works. But I cannot find out why this command-line fails: http://pastebin.com/DSQrm7ND ... some hint would be appreciated.
[17:09] <Na_Klar> meh, avconv does the job ... nvm
[22:28] <Mista_D> any way to just copy a file as is in the same container (as part of automated process where ffmpeg has to be envoked)?
[22:33] <llogan> Mista_D: i guess you could script it.
[22:34] <Mista_D> llogan: will have to do...
[22:35] <Mista_D> Any way to use "copy" filter?
[22:37] <llogan> http://mywiki.wooledge.org/BashFAQ/073 would be useful for scripting
[22:45] <Mista_D> llogan: Thank you, looked over it...
[22:57] <spaam_> ux is the old ubitux ? :)
[22:57] <ubitux> actually the "new"
[00:00] --- Fri Feb 21 2014