[Ffmpeg-devel-irc] ffmpeg.log.20120618

Tue Jun 19 02:05:01 CEST 2012

[00:22] <josinalvo> Hi. I am looking for a audio and video codec combination that can give me good quality on a low end machine. I'd expect it to use very little compression, and that would be OK
[00:27] <burek> josinalvo, what kind of low end machine
[00:27] <burek> Bolle, you misunderstood ffserver, I guess
[00:28] <burek> although it can take inputs, using File directive in config file
[00:28] <burek> the usual way is to run ffmpeg and feed the ffserver with ffmpeg's output
[00:28] <burek> so, run your ffmpeg, read your mms input and feed the ffserver with it
[00:28] <burek> ffmpeg -i mms://... http://localhost:8090/feed1.ffm
[00:34] <JulioVega> hi, i plan to decode video from a quadcopter, which uses a "modified" h264 codec. From what the SDK guide says (see http://davis.isep.fr/logiciels/ar-drone/SDK18/ARDrone_SDK_1_7_Developer_Guide.pdf page 43), it is a mere subset of the h.264 baseline profile except for the entropy encoding. I'm not really deep into video decoding, so my question is: If I do the entropy decoding myself, can I pass the result to libavcodec for decoding? Or does it
[01:13] <Bolle> No, I wasn't misunderstanding, I already did that, I just wanted to confirm that that was all it could do. So is there a way to run off of a config for ffmpeg?
[01:14] <Bolle> Or being able to run several streams through one terminal/one command
[01:14] <Bolle> possibly ffmpeg -i (input) (output) & ffmpeg -i (input) (output)
[02:02] <josinalvo_> Hi. I want to record a screencast using ffmpeg. My computer is a eeepc, and has not much processor. I'd like to produce a video (with audio) that has a good quality. My plan is to use codecs that use very little processor, and not mind much about the size of the output. What codecs should I use ?
[02:17] <burek> josinalvo, try lossless codec
[02:17] <burek> http://ffmpeg.gusari.org/viewtopic.php?f=25&t=594
[02:18] <burek> pay attention to "General note" section
[02:21] <josinalvo_> thanks!
[02:42] <josinalvo> (i just noticed my ffmpeg is too old for some of these options. I'll come back with the results once this is fixed =P)
[04:15] <josinalvo> burek: the line as described on the topic you mentioned kept not working. However, I got a nice result from https://wiki.archlinux.org/index.php/FFmpeg . Aparently, mkv allows for different rates of compression. Thanks for the help!
[09:49] <Guest13969> hi there
[09:51] <Guest13969> i am trying to cross compile ffmpeg with the android ndk gcc, along with libx264, on ubuntu 10.04. the ffsever executable is built, but i am unable to get the ffmpeg executable. here's the pastebin - http://pastebin.com/Hqr6wurp, could someone take a look?
[12:49] <c10ud> hello, i'm playing with ffserver in order to stream a libvpx/webm live video, but i'm getting an high delay (~10 seconds), i remember i experienced this also with flumotion, and never found an answer. Is there some libvpx setting that i can toy with in order to make the video really live?
[12:51] <Bolle> Hi, just like yesterday, I'm wondering if there is any way of setting ffmpeg to be automated, run off a config, or a command line that lets you create multiple streams?
[13:06] <Bolle> I found something, multiple inputs, use the -map to control them
[13:06] <Bolle> Is there any other way, or is this the best way?
[13:08] <c10ud> oh, maybe i need to pass --rt --cpu-used=4 to libvpx, but how to add that to ffserver.conf?
[13:13] <Bolle> I'm a newb, so don't take from me, but do you possibly have buffer or preroll set in the ffserver.conf?
[13:19] <catalytic> Hi All
[13:20] <catalytic> I am extracting frames from a quicktime.mov
[13:20] <catalytic> and want to maintain the same frame number as that of the quicktime
[13:20] <catalytic> instead of starting from 1 each time. I was wondering if this was possible?
[13:20] <catalytic> I have tried
[13:48] <burek> c10ud, you can submit a ticket at ffmpeg's bug trac
[13:49] <burek> Bolle, what exactly do you need?
[13:50] <burek> catalytic, how do you mean to maintain frame number?
[13:51] <c10ud> burek, i don't think it's a bug, but i cannot find any reference on how to do such thing
[14:25] <Bolle> Umm, can someone give me an example of an input line that takes from two sources to two -f mpegts?
[14:25] <Bolle> I'm trying to get it to work properly, but I keep failing, hehe.
[14:40] <Bolle> Is there any way of setting the same encoder settings for all the -map n?
[14:43] <Bolle> Also, apart from "buffer", is there any way of making the streams better? What options can I use?
[14:43] <Bolle> That is, to make it as fluid as possible.
[14:44] <Bolle> Audio, video, framerate.
[14:46] <burek> ffmpeg -i input1 -i input2 ... -map 1 -f mpegts output1 -map 2 -f mpegts output2
[14:48] <Bolle> dont you mean -map 0?
[14:48] <burek> yes
[14:48] <Bolle> Either way, can you guys suggest options to make the output as fluid, or close to as fluid, as the input
[14:49] <burek> it's always fluid
[14:49] <Bolle> Not really.
[14:49] <mickepaprika> hello..
[14:49] <Bolle> First, audio is taking small pauses, and the video can sometimes also have a frame drop
[14:52] <mickepaprika> i'm looking for a current version of an example that outputs a file using libavformat...
[14:52] <mickepaprika> can someone confirm that this is the latest version:
[14:52] <mickepaprika> http://ffmpeg.org/doxygen/trunk/output-example_8c-source.html
[14:52] <mickepaprika> (it seems to use deprecated functions, and i have found several outer version, so that's why i'm asking)
[15:00] <mickepaprika> ah ok... i found the decoding_encoding.c example, i suppose that is the most current
[15:20] <leandrosansilva> Hello to all. I'm using libavfilter API and I'm having some problems to use it. I tried the libav-users mailing list but nobody answered me. There's no much documentation about it (only an example, which doesn't help much), so I want to know if there's someone here who understand well the libavfilter api
[15:21] <leandrosansilva> I'm trying to convert a image from an input format (jpeg) to another in other format and the half of the original resolution
[15:22] <burek> leandrosansilva, noone has time to write the documentation
[15:22] <burek> especially if the api is changing a lot
[15:22] <burek> so, your best chance is to read the ffmpeg's source code
[15:22] <burek> and to figure out hot it is done properly
[15:22] <burek> -hot +how
[15:24] <leandrosansilva> yes, I'm also reading ffmpeg.c
[15:24] <leandrosansilva> but it's code (~5000 lines in a plain C file) isn't very easy to understand
[15:24] <leandrosansilva> :-(
[15:25] <burek> the code is modular
[15:25] <burek> most probably you don't need to read ffmpeg.c
[15:25] <burek> but some other (more specific) file
[15:25] <burek> try looking at libavcodec/
[15:27] <burek> jpegls*
[15:27] <leandrosansilva> in fact in libavcodec there's nothing about avfilter in libavcodec
[15:27] <burek> well, check libavfilter/ then
[15:27] <burek> it's all there :)
[15:28] <burek> vf_scale.c
[15:28] <burek> will do the scaling, for example
[15:28] <burek> etc
[15:28] <leandrosansilva> Yes, I read it
[15:28] <leandrosansilva> and when I use ffmmpeg of ffplay they work
[15:28] <burek> there you can see how to use libavfilter
[15:28] <leandrosansilva> when I use -vf 'scale=blablabla'
[15:28] <leandrosansilva> the problem is when I try to use it in my code
[15:29] <leandrosansilva> I will post it in pastebin
[15:29] <leandrosansilva> just a moment
[15:31] <leandrosansilva> It's in http://pastebin.com/GDuXpiJr
[15:32] <leandrosansilva> it's  a (non modular, everything is in main()) plain C file which should do what I need (format the image to png and scale to half size)
[15:32] <leandrosansilva> the problem is these two filters aren't being applied
[15:35] <leandrosansilva> when I get the a frame from filtergraph buffer with av_buffersink_get_buffer_ref() and avfilter_fill_frame_from_buffer_ref() after apply the filter with av_vsrc_buffer_add_frame() I get the original frame, that which is the input frame, without filtering
[15:36] <Guest92770> Hi - i've just reported a defect - https://ffmpeg.org/trac/ffmpeg/ticket/1461, can someone look at it if its a genuine defect, if so is there a workaround? (http://pastebin.com/AWnDEKYU)
[15:42] <burek> Guest92770, please post your ffserver.conf file too
[15:45] <Guest92770> http://pastebin.com/7sqe0JSC & updated the defect as well
[16:11] <Bolle> Does the UDP protocol have a buffer in time, rather than size? Or is there an option to set buffer rather? If not what's an appropriate buffer size?
[16:27] <Bolle> Anyone?
[16:28] <Bolle> I tried ?buffer_size=2048, still gives me framedrops, audio drops as well
[17:07] <Mista_D> Can't set "digest mode=on" in mailing list, is it disabled?
[17:09] <Mista_D> Tried with IE and FireFox.
[17:10] <Bolle> Anyone? Are there any options to reduce audio drops and framedrops? Or is the buffer the only solution? Regarding that though, it still drops, regardless of how high I set it.
[17:11] <iive> Mista_D: it is very likely that it is disabled. It is quite horrible when you get replies containing all mails from a single day.
[17:12] <Mista_D> iive: got it.
[17:19] <Bolle> Any suggestions at all how I can get the output as fluid as possible? My command line only consists of ffmpeg -i input -f mpegts -vcodec libx264 -acodec libfaac 'udp://localhost:10000?pkt_size=188&buffer_size=4096' , anyone?
[17:20] <Bolle> framedrops and audiodrops
[17:20] <Bolle> Packet size will be for an another application, since it complains about "incomplete packets"
[17:21] <Bolle> Oh i forgot and -sameq
[18:28] <Bolle> Error while decoding stream #0:1: Operation not permitted, it doesn't state what it is, though the stream starts, also, i was wondering about when you decode wmv3, it gives you an "Extra data: 8 bits left, value: 0"
[18:29] <Bolle> Regardless, is there any way of setting your stream to restart once it fails? Rather than it just stopping completely? Or would you have to write a script?
[18:37] <dougle> Can anyone tell me if there is an easy way to force the PAR and DAR of an image2 encode to h264 mp4?
[18:39] <dougle> i'm creating a small mp4 to cat on the end of a larger mp4 with MP4Box but no joy so far (was working for previous ffmpeg versions), the only difference to the video streams i can see are PAR and DAR
[18:55] <AndyThirtover> Evening all ... could do with some help with a mystery ...
[18:55] <AndyThirtover> ffmpeg is overwriting input files ..
[18:55] <AndyThirtover> So if I have:
[18:55] <AndyThirtover> ffmpeg -i Archive/IMG_*.JPG  ffZlamp.mp4
[18:56] <AndyThirtover> I find that all the files in Archive get overwritten with a compressed copy of the first file in that directory
[18:57] <AndyThirtover> pastebin'd
[18:58] <AndyThirtover> ffmpeg 0.8.3-4:0.8.3-0ubuntu0.12.04.1
[18:58] <AndyThirtover> From latest version of ubuntu 12.04
[18:58] <dougle> you can't glob them: ffmpeg -i Archive/IMG_*.JPG  ffZlamp.mp4  expands to ffmpeg -i Archive/IMG_01.JPG Archive/IMG_02.JPG Archive/IMG_03.JPG  ffZlamp.mp4 (without the necessary -i switch)
[18:59] <iive> AndyThirtover: this is how bash works. it expands the * on its own.
[18:59] <dougle> needs to be ffmpeg -f image2 -i Archive/IMG_%d.JPG  ffZlamp.mp4 (i think)
[19:00] <AndyThirtover> Ok I've seen that in examples
[19:00] <dougle> the %d in place for a decimal which ffmpeg will expand on it's own without shifting the params around
[19:01] <AndyThirtover> K I have 00nn.JPG as numbers ...  I'll research myself since 04%d didn't do as I expected ...
[19:02] <dougle> i found it useful to zero pad the numbers on your index filenames if you can e.g. Archive/IMG_0001.JPG Archive/IMG_0002.JPG etc
[19:08] <Bolle> Question: ffmpeg -i inp -i inp -map0 outp -map1 outp, can I set it to transcode them all with the same codec?
[19:08] <Bolle> I tried before, I either did it wrong, or it can't be done?
[19:17] <AndyThirtover> Just a thought -- will this not work if my images start at IMG_0067.JPG rather than IMG_001.jpg ?
[19:17] <AndyThirtover> I found that IMG_00??.JPG listed the sequence properly ..
[19:46] <AndyThirtover> Well I've make this shellscript : http://pastebin.com/SiddyRpX
[19:46] <AndyThirtover> Am I missing something ..
[19:46] <AndyThirtover> I only get the first frame repeated ..
[19:46] <Bolle> PTS -4294873733, next:93485000 invalid dropping st:1 DTS -4294873693, next:93486000 st:1 invalid dropping  anyone?
[20:01] <llogan> AndyThirtover: cat *.JPG | ffmpeg -f image2pipe -vcodec mjpeg -i - output.mp4
[20:01] <llogan> Bolle: use a pastebin site to show your ffmpeg command and the complete console output
[20:01] <funyun> hi. can anyone help me with compiling ffmpeg with x264 on a mac?
[20:02] <llogan> funyun: there is this: https://ffmpeg.org/trac/ffmpeg/wiki/MacOSXCompilationGuide
[20:02] <llogan> although it doesn't really meet my standards compared to the other guides on the wiki
[20:03] <llogan> (fortunately/unfortunately I am macignorant)
[20:04] <funyun> llogan: cool. thanks :)
[20:08] <AndyThirtover> llogan --- thanks I gave that a try ...
[20:10] <AndyThirtover> [mjpeg @ 0x19f9c80] only 8 bits/component accepted
[20:10] <AndyThirtover> So I guess I have to look at vcodec options
[20:13] <llogan> AndyThirtover: is it actually a jpeg image?
[20:15] <llogan> can you provide one of the images?
[20:22] <AndyThirtover> llogan --- yes I can ... came from a Canon EOS5D
[20:23] <AndyThirtover> Just thinking about where I can put it!
[20:24] <llogan> mediafire, datafilehost, dropbox
[20:24] <llogan> or email
[20:26] <AndyThirtover> email is easy!
[20:29] <AndyThirtover> should be on its way ...
[20:32] <AndyThirtover> Also did datafilehost :  http://www.datafilehost.com/download-16ffc1ff.html
[20:32] <AndyThirtover> That was easy !
[20:41] <AndyThirtover> In fact so easy, I did the whole sequence ..  :   http://www.datafilehost.com/download-778485e6.html
[20:42] <Bolle> Is there a protocol I should be using for rtsp adresses, rather than rtsp://?
[20:43] <Bolle> like mms, mmst, and http, mmsh
[20:45] <llogan> AndyThirtover: got distracted. i'll look at it now.
[20:46] <AndyThirtover> Double cheers
[20:52] <llogan> AndyThirtover: works for me. use a pastebin site to show your ffmpeg command and the complete console output.
[20:54] <AndyThirtover> Here you go:  http://pastebin.com/cdkeRuHj
[20:55] <AndyThirtover> I guess this is the error:   [mjpeg @ 0x1c2cc80] only 8 bits/component accepted
[20:55] <llogan> youre using faux-ffmpeg; not ffmpeg from FFmpeg
[20:56] <burek> AndyThirtover, %04d
[20:56] <Bolle> I still haven't gotten an answer if it's possible to have 2 input streams, 2 output streams and transcode both with the same options, so I won't need to repeat it
[20:56] <llogan> doesn't start with 0001
[20:56] <AndyThirtover> OK ... I have a set that do  ...
[20:57] <llogan> AndyThirtover: https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide
[20:57] Action: llogan needs to finish the "local install" version of the guide
[20:57] <burek> also, that's not ffmpeg, that's libav
[20:57] <AndyThirtover> Does it have to be 0001 or would img1.jpg be fine ..
[20:58] <AndyThirtover> is that because when I install ffmpeg I get libav on 12.04 ?
[20:58] <llogan> if they aren't in sequential order in a format that ffmpeg understands, but still "in order", then cat should still work
[20:58] <burek> that's because of a recent fork of ffmpeg project named libav
[20:58] <AndyThirtover> Ok ... nice fork ?
[20:58] <burek> where one of the guys was debian/ubuntu maintainer
[20:59] <burek> and he did some things on his own, confusing thousand people in return
[20:59] <burek> shortly, you don't  have ffmpeg
[20:59] <burek> you need to compile it, llogan gave you a link for it
[20:59] <llogan> AndyThirtover: when people come here for libav help, and things don't work, swtiching to recent ffmpeg often works.
[20:59] <llogan> or you really want to keep libav you can:
[21:00] <llogan> 1) perform a "local" ffmpeg installation or
[21:00] <llogan> 2) try to get help at #libav or whatever mailing list they use
[21:00] <AndyThirtover> K ... here's the results from ls ... :  http://pastebin.com/2AG6mTrW
[21:01] <burek> AndyThirtover, you could use just %d
[21:01] <burek> ffmpeg -f image2 -i img%d.jpg ...
[21:01] <llogan> as in: ffmpeg -i img%d.jpg output.mp4
[21:02] Action: llogan slow typer
[21:02] <burek> Bolle, rtsp:// is fine
[21:02] <AndyThirtover> Been trying that for some time
[21:02] <Bolle> Yay, an answer :D
[21:03] <burek> but usually you need rtp
[21:03] <Bolle> It works fine in vlc, to view, but wont work with
[21:03] <AndyThirtover> getting @ numbered/img%d.jpg: No such file or directory
[21:03] <burek> AndyThirtover, can you please use pastebin.com, to show your command line and its output?
[21:03] <burek> Bolle, type ffmpeg -protocols
[21:03] <Bolle> Doesn't ffmpeg have rtp already?
[21:03] <burek> if it is supported, it will be shown
[21:03] <AndyThirtover> K ..
[21:03] <Bolle> kay
[21:04] <Bolle> it is
[21:04] <AndyThirtover> http://pastebin.com/GJyiaKvz
[21:04] <AndyThirtover> Thats the results of using %d
[21:05] <burek> AndyThirtover, try using it the way we suggested
[21:05] <burek> without image2pipe, but with image2
[21:05] <Bolle> can you use one option to set all the outgoing streams transcoders though?
[21:05] <burek> Bolle, why would you need that?
[21:05] <llogan> Bolle: i don't think so
[21:06] <burek> you could try with a preset file
[21:06] <AndyThirtover> Thanks .... but back to the first frame problem:  http://pastebin.com/KX13snnr
[21:06] <burek> something like ffmpeg -i input1 -i input2 -i input3 -vpre preset1 out1 -vpre preset1 out2 -vpre preset1 out3
[21:07] <llogan> AndyThirtover: "Changing frame properties on the fly is not supported"
[21:07] <burek> AndyThirtover, all the jpg images need to be of the same characteristics
[21:07] <burek> you can't just mix random images
[21:07] <Bolle> so -vpre preset1, preset1 refers to a file with the options?
[21:07] <burek> they need to have the same frame size, color depth, etc
[21:07] <AndyThirtover> Indeed ...  but as they all came from the same camera I thought they'd been the same!
[21:07] <burek> Bolle, I think so, check the docs
[21:08] <AndyThirtover> Its an animation sequence ...  32 images of a light rotating ...
[21:08] <burek> Bolle, -fpre
[21:08] <AndyThirtover> I'll do a properties check
[21:08] <llogan> AndyThirtover: it worked for me with the images from Archive.zip.
[21:08] <burek> AndyThirtover, are those images created by the same application
[21:08] <burek> well if it works for llogan
[21:08] <AndyThirtover> llogan --- so it is possible !
[21:09] <burek> then your libav is buggy
[21:09] <burek> do yourself a favor, try with static ffmpeg for a quick test
[21:09] <AndyThirtover> So how about I build ffmpeg from scratch ?
[21:09] <burek> if it works, then compile your own ffmpeg
[21:09] <AndyThirtover> On it ..
[21:11] <llogan> burek: ah...pirates.
[21:11] <burek> yeah :) we used to maintain a dc++ hub which was ironically named pirates :)
[21:12] <llogan> is dc++ still used much?
[21:12] <AndyThirtover> Chaps --- Just built a bigger mp4 with that static jobby!
[21:12] <llogan> once again, libav fails.
[21:12] <burek> \o/
[21:13] <burek> just like chinese copies of nike :D
[21:13] <AndyThirtover> Extra Double Extra Monkeys   !
[21:13] <AndyThirtover> I love you guys!
[21:14] <burek> llogan, well it is, especially since piratebay switched to magnet links instead of direct torrent files
[21:14] <burek> I also install dc hubs in companies local networks, so they can chat/share files and stuff :)
[21:14] <Bolle> Is there something I need to enable so that sources can contain domain names rather than IPs? Or does it have to be ips?
[21:15] <burek> Bolle, can you please use pastebin.com, to show your command line and its output?
[21:15] <Bolle> What does that have to do with anything?
[21:16] <burek> Bolle, your question is fuzzy, so it's better when you provide a command line, so that we can all understand what are you talking about
[21:16] <burek> domain names and IP adresses are 2 diff things in nature :)
[21:17] <Bolle> burek, yes, but ffmpeg can only use input sources that are mmsh://ip/.../ for instance, I was wondering if there is any way of having it read the host names
[21:17] <burek> sure, you can use host names instead of ips
[21:18] <burek> if your dns works well
[21:21] <Bolle> Yeah you're right, lol worked on my server but not on my pc.
[21:21] <Bolle> Thanks.
[21:22] <burek> :beer: :)
[21:26] <patheticbliss> hey guys, i have a quick question: i had ffmpeg running, and accidentally alt-tabbed to the console window and started typing, and now it's in this stepping/debugging mode. how can i get it back into the normal mode without restarting it?
[21:27] <burek> what stepping/debugging mode?
[21:28] <patheticbliss> that's what I said. it reminds me of gdb, except it's not documented all
[21:28] <patheticbliss> or as far as i can tell
[21:28] <burek> are you sure you are using ffmpeg at all?
[21:28] <patheticbliss> started through WinFF, yes, i'm pretty sure it's ffmpeg
[21:28] <burek> well ask WinFF maintainers then
[21:29] <burek> this is not WinFF support channel
[21:29] <patheticbliss> ...i am aware of this
[21:29] <patheticbliss> winff just makes batch files
[21:29] <burek> there is no stepping/debugging mode in ffmpeg
[21:29] <burek> (afaik)
[21:30] <Mista_D> getting a report from ffprobe, in csv format with N/A values, any way to find out what are they related to?
[21:31] <patheticbliss> hm... well it's acting like GDB, where it halt after printing out a line about libx264, holding down enter will make it run 'normally'
[21:31] <burek> Mista_D, can you please use pastebin.com, to show your command line and its output?
[21:31] <burek> patheticbliss, then check gdb manual :D
[21:32] <patheticbliss> heh...
[21:32] <Mista_D> http://pastebin.ca/2162644
[21:33] <burek> Mista_D, try running the same command without -print_format csv
[21:33] <burek> and compare results
[21:36] <patheticbliss> typing 'help' into the 'command line' causes a dump of hex data...
[21:41] <patheticbliss> mmk so I think it's actually libx264 that's causing this. god this is annoying.
[21:58] <AndyThirtover> Am I right in thinking that the static ffmpeg wasn't linked to handle PNG ?
[22:05] <llogan> AndyThirtover: what does "ffmpeg -codec | grep png" show?
[22:05] <llogan> *codecs
[22:08] <llogan> works for me
[22:17] <burek> AndyThirtover, it is compiled with png support: DEV D   png             PNG (Portable Network Graphics) image
[22:17] <burek> why do you think so?
[22:40] <Mista_D> burek: runnign without csv, I get none of the N/A ones, just wondering what they are.
[23:00] <AndyThirtover> My ffmpeg gives : http://pastebin.com/wy31jBde
[23:20] <AndyThirtover> Be back tomorrow !
[23:23] <burek> Mista_D, compare the outputs of those two and it will be obvious
[23:26] <Mista_D> burek: there's no mentioing or identifiers for N/A values in XML/JSON formats, while csv has the fields reserved, and values are N/A.
[00:00] --- Tue Jun 19 2012