[Ffmpeg-devel-irc] ffmpeg.log.20120619

burek burek021 at gmail.com
Wed Jun 20 02:05:02 CEST 2012


[00:36] <saste> Mista_D: csv is just a shorthand for a particular set of options of -print_format compact
[00:36] <saste> use compact and see which are the keys
[01:23] <naquad> i need to convert image to 1 second MPEG-TS packed x264 video with 128k stereo mp3 audio. i'm doing it like that: ffmpeg -vcodec libx264  -ab 128k -s 544x436 -loop 1 -i no_signal_720x576.jpg -t 1 out.ts
[01:23] <naquad> but i get error: Unknown decoder 'libx264'
[01:23] <naquad> what decoder? thats a picture :(
[01:32] <saste> naquad: why do you want to specify a decoder for a jpeg?
[01:32] <saste> also libx264 is an encoder, not a decoder
[01:39] <naquad> saste, i don't know :( i've made this line from what i've read in man, maybe i've misunderstood/mixed something
[01:39] <naquad> could you please help me up?
[01:40] <saste> don't specify the codec for the input image, there is no point in it
[01:40] <naquad> but how do i tell ffmpeg that resulting file should be h.264?
[01:42] <llogan> naquad: option placement matters. ffmpeg [input options] -i input [output options] output
[01:43] <naquad> ok, i did ffmpeg -i image ... output.ts
[01:43] <naquad> but that didn't help
[01:43] <llogan> meaning only "-loop 1" should be applied to the input
[01:43] <naquad> ffmpeg -i no_signal_720x576.jpg -vcodec libx264 -acodec libmp3lame -ab 128k  -f mpegts -x264opts frameref=15:fast_pskip=0 out.ts
[01:43] <naquad> ah
[01:43] <naquad> i see
[01:44] <llogan> also, you have no audi oinput
[01:44] <llogan> *audio
[01:44] <naquad> ermmm how do i make a blank audio?
[01:45] <llogan> aevalsrc filter (aevalsrc=0) or something like: ffmpeg -t 3 -ar 48000 -ac 2 -f s16le -i /dev/zero -acodec copy out.wav
[01:46] <saste> naquad: -f lavfi -i "aevalsrc=0" ...
[01:47] <naquad> can i somehow combine it all to one line?
[01:47] <naquad> or i should separately make empty audio file and then encode it somehow with resulting video?
[01:50] <saste> naquad: -map
[01:53] <llogan> naquad: ffmpeg -loop 1 -i input.jpg -f lavfi -i "aevalsrc=0" -vcodec libx264 -vcodec libmp3lame ... output.ts
[01:55] <naquad> ffmpeg -loop 1 -i no_signal_720x576.jpg -f lavfi -i 'aevalsrc=0' -vcodec libx264 -vcodec libmp3lame -ar 128k -f mpegts -t 1  out.ts
[01:55] <naquad> llogan, ^ says Invalid encoder type 'libmp3lame'
[01:56] <naquad> oh
[01:56] <naquad> it was acodec
[01:56] <llogan> oops
[01:57] <naquad> llogan, saste thank you!
[01:57] <llogan> s/-ar 128k/-ab 128k
[02:02] <naquad> as it finds out video is mono :( how do i make it stereo?
[02:03] <funyun> can anyone help me with which -tune preset would be best for MW3 Gameplay? animation or film?
[02:03] <funyun> or maybe another?
[02:04] <naquad> -ac 2 helped
[02:04] <llogan> what's mw3?
[02:04] <funyun> llogan: call of duty
[02:06] <funyun> a video game
[02:09] <llogan> naquad: or aevalsrc=0:0
[02:11] <iive> Vardan: so.
[02:11] <saste> naquad: aevalsrc=0:0
[02:11] <saste> also reads the rest of the documents, sometimes it helps
[02:12] <saste> i mean the aevalsrc documentation
[02:12] <iive> vallor:  as Daemon404 said, you can use avpicture functions.
[02:12] <iive> e.g.
[02:12] <iive> you still alloc the AVFrameRGB the old way.
[02:13] <Vardan> yes
[02:13] <Vardan> I have also tried to use uint8_t *frameData = malloc(1280 * 720 * 3);
[02:13] <Vardan> and gave that to sws_scale
[02:14] <Vardan> in real I have tried to add in buffer the frameData
[02:15] <Vardan> but when I read from buffer and trying to display it's showing black screen
[02:15] <juanmabc> for the record, would'nt it be av_malloc?
[02:17] <iive> i'll just call it avfrgb for short. avfrgb->format=PIX_FMT_RGB24 ; avfrbg->height=720; avfrgb->width=1280;
[02:17] <Vardan> right
[02:19] <iive> just a moment to figure out what the code does.
[02:22] <iive> it's been a while, since i've coded for ffmpeg.
[02:22] <Vardan> here is my latest code http://pastebin.com/qkN3eWty with this code I see only the last frame of my video
[02:24] <iive> avpicture_alloc(avfrgb, avfrgb->format, avfrgb->width, avfrgb->height);
[02:25] <iive> this one would allocate buffer, fill the data and stride
[02:25] <Vardan> instead of avcodec_alloc_frame();?
[02:26] <iive> no, you start with avfrgb = avcodec_alloc_frame();
[02:26] <Vardan> error: 'AVFrame' has no member named 'format'
[02:27] <iive> oh, you need to cast it to (AVPicture*)
[02:27] <funyun> hi. can anyone help me with which -tune preset would be best for MW3 Gameplay? animation or film?
[02:27] <Vardan> ok
[02:27] <iive> as Daemon404 said, it is used this way in many places in ffmpeg, so it should work.
[02:29] <Vardan> so, now I have AVFrame *pFrameRGB = avcodec_alloc_frame(); avpicture_alloc((AVPicture *)pFrameRGB, PIX_FMT_RGB24, pCodecCtx->width, pCodecCtx->height);
[02:29] <Vardan> after which I scale pFrame to pFrameRGB
[02:32] <iive> I guess you could do that too, but you won't have the format and width/heigh stored in the frame.
[02:35] <Vardan> iive: man you are genius :) it's working :)
[02:35] <Vardan> *men
[02:36] <llogan> funyun: how about no tune?
[02:38] <funyun> llogan: the other day someone in here told me playing with tune would give me better quality. so i tried -tune grain and it gave me much better quality. now i'm wondering if some of the other presets would be even better but i am hesitant to try for myself because the 2 minute encode with -tune grain took a little over 3 hours
[02:38] <funyun> so i'm just hoping someone can give their opinion
[02:39] <llogan> let me guess: you're using placebo preset?
[02:40] <funyun> llogan: no. veryslow. but i also have a very slow computer
[02:40] <llogan> veryslow is very slow.
[02:40] <Vardan> does av_fifo_ thread save?
[02:40] <funyun> llogan: veryfast is also veryslow on my computer
[02:42] <funyun> llogan: so do you think any will give better results than grain?
[02:46] <llogan> not if you want to help preserve grain. i don't know if the game add fake grain or whatever
[02:47] <llogan> "x264 --fullhelp" shows what each tune does in terms of options used/modified
[02:50] <funyun> llogan: is animation more for cartoons? or would it work well with realistic animation like a modern day video game?
[02:58] <iive> Vardan: no idea about av_fifo.
[03:23] <burek> funyun, you use veryslow preset when you want your encoder to encode your video slowly
[03:23] <burek> i.e. with higher quality
[03:23] <burek> or better compression
[03:24] <burek> it doesn't mean that it should be used on slow machines :)))
[03:24] <burek> you probably need ultrafast on your machine
[03:24] <funyun> burek: yes but i'm looking for the best quality
[03:24] <burek> well then use -t option to shorten your output
[03:25] <burek> just to see the test results
[03:25] <burek> and for the best quality, use -crf 0 -preset ultraslow
[03:25] <funyun> burek: do you have any ideas about which is better for video game gameplay? -tune animation, grain, film, or something else?
[03:25] <burek> you can't get better than that
[03:25] <burek> it depends how realistic the game is
[03:25] <funyun> burek: modern warfare 3, call of duty
[03:25] <burek> -crf will do lossless anyway, so no need for -tune
[03:26] <funyun> i'm using -crf 19 atm
[03:26] <burek> crf controls your picture quality
[03:26] <burek> preset controls the output compression ratio
[03:26] <burek> if you can wait longer, use -preset slow
[03:27] <burek> or slower..
[03:27] <burek> if not, use fast(er)
[03:27] <burek> but you'll get the bigger output file
[03:27] <funyun> burek: i'm just asking because -tune grain gives me better quality than not using it
[03:27] <burek> then use -tune grain
[03:28] <funyun> burek: but a different tune option may be better. that's what i'm asking. do you think any of the other options would be better than grain?
[03:28] <burek> man.. if you are not sure what do you want
[03:28] <burek> put -preset ultrafast
[03:28] <burek> and -t 10
[03:28] <burek> and just run your tests
[03:29] <Vardan> people does C have something like synchronized methods or members?
[03:29] <burek> until you are satisfied with the image quality
[03:29] <burek> after that set -preset to ultraslow and you are donw
[03:29] <burek> done*
[03:29] <burek> and remove -t 10
[03:29] <burek> Vardan, I think not
[03:29] <burek> you need to implement it manually
[03:31] <funyun> burek: i don't want to sound like a dick but i just don't understand why everybody keeps avoiding my question. i thought i'd come in here and ask, and be gone within 2 minutes. but now it seems i will have to run 24+ hours worth of tests to get the answer. is there any way you can just give me your opinion?
[03:31] <burek> Vardan, http://www.codeproject.com/Articles/12362/A-quot-synchronized-quot-statement-for-C-like-in-J
[03:31] <Vardan> thanks
[03:33] <burek> funyun, because your question is asked without basic understanding
[03:33] <iive> n8 ppl
[03:33] <burek> o/
[03:33] <burek> if you would understand your question completely, you would realize how uncertain it is
[03:33] <burek> we don't know what kind of video you are encoding
[03:34] <burek> it doesn't matter if it's a video game or what.. it does depend on video dynamics
[03:34] <burek> so you need to test it yourself, that's the fact
[03:34] <funyun> burek: can i show you a youtube video? i just assumed people were aware of the biggest video game of all time
[03:35] <burek> did you read what I wrote to you above?
[03:35] <burek> speed up your tests with -preset and -t
[03:35] <burek> and be gone in 10 minutes with results
[03:35] <funyun> so don't use tune at all?
[03:35] <burek> :facepalm:
[03:36] <burek> here is even better suggestion: use 2-pass encoding
[03:36] <burek> of course, first read about it and familiarize yourself with it
[03:37] <burek> then just apply it, without any additional fancy options
[03:37] <funyun> i thought crf was better than a constant bitrate?
[03:38] <burek> well it is
[11:22] <Diogo> hi one question...
[11:23] <Diogo> i need to install libavdevice-dev ...to compile ffmpeg..
[11:23] <Diogo> when i tried to install on debian appears...
[11:23] <Diogo>   libavdevice-dev: Depends: libavdevice54 (= 7:0.11-dmo2) but it is not going to be installed
[11:23] <Diogo>                    Depends: libavformat-dev (>= 7:0.11-dmo2) but it is not going to be installed
[11:23] <Diogo> E: Broken packages
[11:25] <Mavrik_> avdevice is part of ffmpeg, you don't need to install it.
[11:33] <grepper> Diogo: there is an ubuntu compilation guide that works for debian:  https://ffmpeg.org/trac/ffmpeg/wiki/UbuntuCompilationGuide
[11:33] <swkide> Greetings!
[11:33] <swkide> I try to create apple http live streaming segments with ffmpeg. It works, but the audio is not perfect. You hear jumps and rhythmic clicking noises.
[11:33] <swkide> The free segmenter.c is not an option, because it does not compile with ffmpeg 0.11.1.
[11:33] <swkide> Starting video is an HD flv, which I encode to ts and finally segment it with ffmpeg.
[11:33] <swkide> The command and output for encoding to ts is here: http://pastebin.com/QeydjxtU
[11:33] <swkide> The command and output for segmenting is here: http://pastebin.com/FnreDt9M
[11:33] <swkide> You can hear the stream with any iOS device here: http://213.178.91.210/out/test.m3u8
[11:33] <swkide> Any help would be great! Thanks!
[11:55] <Plex> Is there a version of ffmpeg for mpi cluster?
[12:06] <brocatz> hi, my friend just submitted a windows related fix to ffmpeg and i'm wondering who i contact to get a build that includes it
[12:07] <brocatz> he's only built the change against the ubuntu default so it doesn't include a lot of the things i need
[12:28] <burek> brocatz, try mailing lists
[14:11] <Bolle> I will try to keep dumb questions to a minimum today, but I just wanted to know, is there any way of making the transcoding process less CPU intensive? I was wondering if it's easier to write to a file, rather than stream to an adress, in this case localhost.
[14:12] <Bolle> Probably not, but it's worth a shot to ask.
[14:12] <burek> streaming is not cpu intensive at all
[14:12] <burek> neither is writing to a file
[14:12] <Bolle> So it's the same then? Aw well.
[14:13] <burek> cpu usage is directly related to encoding parameters
[14:13] <burek> if you are using x264, try changing -preset
[14:13] <burek> or modify -crf value
[14:13] <Bolle> Hm? I'm not really familiar with those options.
[14:13] <Bolle> What do you suggest I change them to?
[14:14] <Bolle> As you know, I'm a newb to ffmpeg and, well, linux for the most part.
[14:15] <burek> read this first
[14:15] <Bolle> All right, so thank you. I'll check it out.
[14:15] <Bolle> Lol, was just about to.
[14:15] <burek> and then type x264 --help
[14:16] <burek> and see what I am talking about
[14:16] <Bolle> By the way, can I ask a question, sort of ffmpeg related.
[14:16] <Bolle> At least it says it works well with ffmpeg, red5, is it any good?
[14:16] <Bolle> I was gonna read up on it as soon as I was finished with all of this.
[14:18] <Bolle> None the less, thank you guys for standing up with my questions for 2 days now. ^^
[14:19] <brocatz> has anyone built a win32 ffmpeg
[14:19] <juanmabc> zeranoe?
[14:19] <Bolle> yes
[14:19] <brocatz> there seems to be some confusion online as to the appropriate approach
[14:20] <Bolle> Though from what I've seen, it's more buggy and doesn't work as well.
[14:20] <brocatz> i'm assuming ubuntu/debian with the mingw scripts onzeraneo.com
[14:20] <juanmabc> google "zeranoe ffmpeg builds"
[14:20] <brocatz> ^ already read all that
[14:20] <brocatz> guess ideally i'd get the patch we submitted accepted and into a nightly
[14:21] <brocatz> then i wont have to think about it anymore
[14:22] <brocatz> tiny patch https://ffmpeg.org/trac/ffmpeg/ticket/1463
[14:22] <brocatz> just makes it not be a dick in windows
[14:29] <burek> brocatz, were you the one who solved his problem sending 'q' keystroke to ffmpeg on win32?
[14:29] <brocatz> i didn't solve it, i paid someone to patch ffmpeg after explaining the problem
[14:30] <brocatz> and that's the patch
[14:30] <brocatz> i found people asking about it on the web as far back as 2009
[14:30] <brocatz> i figured, may as well just get it fixed
[14:31] <burek> well, try it
[14:31] <brocatz> i'll have to set up a build environment for win32 builds
[14:31] <burek> just send a 'q' keystroke to ffmpeg process
[14:31] <brocatz> i've already tested it against a basic h264
[14:31] <brocatz> that doesn't work
[14:31] <burek> the guy was here a few days ago and solved his problem that way
[14:31] <brocatz> i have a bunch of test programs at work
[14:31] <burek> so I'm pretty sure it works
[14:31] <brocatz> i was that guy
[14:32] <brocatz> it didn't fix it
[14:32] <brocatz> i idle in here 24/7
[14:32] <burek> why didn't it work?
[14:32] <burek> did you redirect the input or something?
[14:32] <brocatz> no idea, there is no way i can attach a debugger to ffmpeg binary on my windows machine
[14:32] <brocatz> to read what the program is doing
[14:32] <brocatz> also on windows it doesn't use stdin
[14:32] <brocatz> it's more complicated
[14:32] <burek> let me test it, I have virtualbox
[14:33] <brocatz> the best / right solution is to transfer a sigbreak into a sigint
[14:33] <brocatz> you'll have to send the key stroke from a popen() call from c or python or similar
[14:33] <brocatz> like you would normally communicate with other processes
[14:33] <brocatz> withs stdio
[14:33] <burek> no, why?
[14:34] <brocatz> because the software is part of an automated system
[14:34] <burek> just a simple send message using ffmpeg window handle
[14:34] <brocatz> if you can do it in an automated way from C or Python then i'd be interested to see it
[14:34] <brocatz> (on windows)
[14:34] <burek> SendMessage(WM_KEYDOWN, ...)
[14:34] <burek> and keyup
[14:34] <brocatz> right
[14:35] <brocatz> the process doesn't have a window fwiw
[14:35] <brocatz> it's part of an automated process
[14:36] <burek> every win32 process has a handle
[14:36] <burek> for the message queue
[14:37] <brocatz> after a CreateProcess call i don't understand how you would find the window handle to send the message to
[14:38] <brocatz> and FindWindow didn't look like it was very exact
[14:39] <brocatz> it seems like a bit of a hack
[14:39] <burek> piProcInfo
[14:39] <burek> hProcess
[14:39] <burek> and hThread
[14:40] <brocatz> also this patch will mean i can use popen to nicely parse the ffmpeg output
[14:40] <burek> http://msdn.microsoft.com/en-us/library/windows/desktop/ms684873(v=vs.85).aspx
[14:40] <burek> (last param of your createprocess method)
[14:41] <brocatz> yeah i have all the values you spoke of
[14:41] <burek> well try sending the process a 'q' keystroke then
[14:41] <burek> or 'ctrl+c'
[14:42] <Bolle> keystroke sounds wrong to me
[14:42] <burek> well he could also send WM_CLOSE
[14:42] <brocatz> this whole approach isn't super good
[14:42] <burek> it's the analogue thing to invoking kill in linux
[14:43] <brocatz> sending a signal and having it handled properly like sigint is, is the best way
[14:43] <burek> man, windows doesn't have signals :D
[14:43] <brocatz> then you can use popen like normal
[14:43] <burek> I hope you knew that much before starting coding win32 apps :)
[14:43] <brocatz> did you even look at the patch description
[14:43] <Bolle> I always just use ctrl+c, though sometimes i want to copy a line fast and by habit, I use ctrl+c...
[14:43] <burek> brocatz, ok
[14:44] <JEEB> brocatz, don't pay too much attention at burek -- he doesn't always fully understand what's going on. If you want to talk with the developers, hop onto #ffmpeg-devel
[14:44] <Bolle> :O JEEB, I thought you died.
[14:44] <brocatz> i'll have to deal with it when i get up anyway, so i'll hassle ffmpeg-devel then
[14:45] <JEEB> yeh
[14:46] <Bolle> By the way, burek, you mean for me to lower the cfr and preset to below 23 and lower than medium, would it make a difference on the output, or will it be barely recognizable if it's slight?
[14:47] <burek> Bolle, can you rephrase your question?
[14:47] <burek> sometimes I have a feeling not even you know what you wanted to ask :)
[14:47] <Bolle>  xD
[14:48] <Bolle> I'm asking if changing -preset and -cfr, to a slightly lower value, slow and 20, for instance(rather than medium and 23), will affect the output stream.
[14:48] <burek> it will
[14:49] <Bolle> That is, would I be trading quality for quantity.
[14:49] <Bolle> What would be a reasonable value for the both of them in that case, would you say?
[14:49] <Bolle> Keep in mind, the sources I use are for the most part around 400 pixels
[14:50] <burek> 400 pixels of what?
[14:50] <Bolle> I was referring to height.
[14:51] <Bolle> The most common res would be 640x420, they differ, one drops down to 320x240.
[14:52] <burek> well, frame size doesn't tell much about the dynamics of your video stream
[14:52] <burek> so I don't know
[14:52] <Bolle> Let's say poor.
[14:52] <burek> let's say you test different values and see what you like? :)
[14:53] <Bolle> Will do. I just like getting an answer if I can, hehe.
[14:53] <Bolle> Thank you.
[14:59] <ozone89> hello!
[15:00] <ozone89> I'm trying to do some hardsubbed videos, but ffmpeg cannot find the fonts, even if they're installed system-wide or they're in the same folder where i launch the command line
[15:01] <ozone89> subs are .ass
[15:43] <censor> hi all
[15:44] <censor> i'm trying to get timecode burn-in to work with 0.8.7, and i'm able to get some text to be burned in, but when i set the timecode parameter, it quits with "[drawtext @ 0x1bc3a00] Key 'timecode' not found."
[15:44] <censor> here's my command line: http://pastebin.com/UuBeGqMn
[15:46] <saste> censor: too old?
[15:46] <censor> in the source i can see references to timecode
[15:52] <censor> okay, i compiled 0.11.1, now i get "Drop frame is only allowed with 30000/1001 FPS"
[15:53] <cbsrobot> censor: show your command and the whole output
[15:54] <censor> there you go: http://pastebin.com/YCbQr1EN
[15:55] <cbsrobot> try: drawtext=fontfile=/usr/share/fonts/truetype/ttf-bitstream-vera/VeraMono.ttf:fontsize=28:fontcolor=white:box=1:boxcolor=0x000000CC:timecode='00\\:00\\:00':rate=25"
[15:56] <censor> you just dropped the frames, right?
[15:56] <censor> then i get "Unable to parse timecode, syntax: hh:mm:ss[:;.]ff"
[15:57] <censor> aaah!
[15:57] <censor> :timecode='00\:00\:00\:00'   <- that works
[15:58] <censor> so it _has_ to be : in front of the frames
[15:58] <cbsrobot> : means pal, . and ; ntsc
[15:59] <censor> uhm, strange, but okay...
[15:59] <ubitux> ':' means non-drop, '.' and ';' means drop
[15:59] <ubitux> and drop is only available in ntsc
[15:59] <censor> hmm, now with the new ffmpeg version, the transcoding is just grey
[15:59] <ubitux> (btw, you can build with freetype to avoid full-path font now)
[16:00] <censor> thanks, i'll build a new one
[16:01] <censor> but do you happen to know why all of a sudden the encoded video is just grey?
[16:01] <cbsrobot> what player are you using ?
[16:01] <censor> vlc
[16:01] <cbsrobot> should work
[16:01] <censor> the previous outputs from 0.8.7 were fine
[16:01] <cbsrobot> but try to add: -pix_fmt yuv420p
[16:01] <cbsrobot> so:
[16:02] <cbsrobot>  /usr/local/bin/ffmpeg -y -threads 4 -i test.m2v -vcodec libx264 -vb 3072k -vf "yadif=1:1, crop=720:576:0:32, scale=640:480, drawtext=fontfile=/usr/share/fonts/truetype/ttf-bitstream-vera/VeraMono.ttf:fontsize=28:fontcolor=white:box=1:boxcolor=0x000000CC:timecode='00\\:00\\:00.00':rate=25" -pix_fmt yuv420p -f mp4 /srv/test.mp4
[16:03] <cbsrobot> ah and change .00 to :00
[16:04] <cbsrobot> ah i messed up
[16:04] <cbsrobot> so again:
[16:04] <cbsrobot>  /usr/local/bin/ffmpeg -i test.m2v -c:v libx264 -b:v 3072k -vf "yadif=1:1,crop=720:576:0:32,scale=640:480,drawtext=fontfile=/usr/share/fonts/truetype/ttf-bitstream-vera/VeraMono.ttf:fontsize=28:fontcolor=white:box=1:boxcolor=0x000000CC:timecode='00\\:00\\:00\\:00':rate=25" -y /srv/test.mp4
[16:08] <censor> yay, the pix_fmt option fixes it, thanks
[16:09] <censor> now to make it perfect, is there a way to add some padding to the box?
[16:28] <burek> vf=pad...
[16:28] <burek> -vf 'pad=...'
[16:35] <censor> burek i think this pads the whole frame, not the box drawn by drawtext
[16:48] <burek> http://ffmpeg.org/ffmpeg.html#drawtext
[16:48] <burek> censor, use x and y
[16:48] <censor> but that only positions the box?
[16:49] <censor> i don't want to set the padding around the box, i want to set the padding inside the box - so that the text doesn't touch the borders of the box
[16:49] <burek> shadowcolor ?
[16:50] <burek> shadowx, shadowy
[16:50] <burek> just read the manual
[16:50] <burek> :)
[16:50] <burek> you can also draw a rectangle using another filter
[16:50] <censor> i did, that's why i'm asking
[16:50] <burek> and then position your text inside that rectangle
[16:51] <censor> the shadow has nothing to do with the box
[16:51] <censor> the box is drawn around the text
[16:51] <burek> draw your own box
[16:53] <censor> that is not very practical, is it?
[16:54] <burek> do you want to be a philosopher or do you want to solve your problem? :)
[16:57] <censor> well, i was just asking if there's a way to add some padding to the box, it's not really a problem and certainly not worth the overhead, our editors will just have to live with looking at an ugly timecode then =)
[16:59] <Bolle> Burek, or any other person that can share their opinion. How well does ffmpeg work with dvb? I've only used MuMuDVB up until now, though that's multicast, so it wastes resources.
[17:00] <Bolle> Also, currently I'm using wowza, I wanted to know your opinions about red5, is it any good? I'm heading to read about it now.
[17:00] <burek> Bolle, use google to find your answers
[17:00] <Bolle> Lol, I'm not looking for those kind of answers.
[17:01] <Bolle> I prefer getting opinions, rather.
[17:01] <censor> what do you want to accomplish exactly?
[17:01] <Bolle> I want to rule the world.
[17:01] <censor> okay, fine, and what did you want to accomplish with the DVB streams?
[17:02] <censor> i've used DVBlast in our offices, for multicast
[17:02] <Bolle> That's just the thing, I'm planning on using unicast.
[17:02] <censor> but that was just no-fuzz access-for-everybody, nothing big
[17:02] <censor> so?
[17:03] <Bolle> I was asking if ffmpeg works well with receiving dvb.
[17:03] <Bolle> I don't have any means of testing it at the moment, so I wanted to ask.
[17:03] <censor> it works, but for something similar is used VLC in the end (yes, VLC uses ffmpeg, i know)
[17:04] <Bolle> yes, and vlc requires more from the system.
[17:04] <Bolle> I prefer ffmpeg, it's better and lighter.
[17:04] <burek> it works well with ffmpeg
[17:04] <censor> if you're willing to spend some time on tinkering, ffmpeg works well with pretty much everything
[17:05] <Bolle> I'm too stubborn on that end, so yes I'm willing.
[17:05] <Bolle> And so I've noticed, as of yet, I love ffmpeg.
[17:05] <censor> who doesn't?
[17:05] <Bolle> Especially since it's open source.
[17:05] <censor> except for the missing padding for boxes drawn by drawntext, of course.
[17:17] <censor> here's another funny thing: i have a clip that's 10s long, at 25fps, and when i draw the timecode on it, the timecode is at 00:00:18:something when the video hits the 10s
[17:18] <censor> it looks like the timecode runs 2x fast
[17:18] <ubitux> oO
[17:18] <ubitux> can i see the cmd line you use, and full ffmpeg output?
[17:19] <censor> sure, sec
[17:19] <censor> here you are: http://pastebin.com/bxRvDGFs
[17:20] <ubitux> mmh
[17:20] <ubitux> i suspect the yadif filter is at fault
[17:20] <censor> okay, let me check without it
[17:20] <ubitux> can you confirm the issue without it?
[17:21] <censor> yes, without it it seems to be fine
[17:22] <ubitux> so you say the frame increment is +2 ?
[17:22] <censor> it seems to be x2
[17:22] <censor> or rather: *2
[17:22] <ubitux> did you check frame per frame?
[17:22] <ubitux> (with '.' for instance in mplayer)
[17:23] <censor> got the shortcut for vlc handy?
[17:23] <ubitux> nop :p
[17:24] <ubitux> i wonder if the yadif filter isn't calling the rest of the filtergraph
[17:24] <ubitux> and then do the dropping
[17:24] <ubitux> saste: can you confirm that kind of behaviour?
[17:24] <ubitux> (sorry i'm still unfamiliar with that stuff..._
[17:24] <ubitux> )
[17:24] <ubitux> s/un/not /
[17:25] <censor> all i can say is that the probably last frame at 10s has the tc 00:00:19:16
[17:25] <censor> so that's pretty much *2, yes
[17:25] <censor> i'll leave yadif out for now
[17:27] <saste> ubitux: don't know what "calling the rest of the filtergraph and then do the dropping" means
[17:27] <saste> maybe a timestamp problem?
[17:27] <saste> in that case showinfo should help debugging the issue
[17:29] <ubitux> saste: no not a ts problem
[17:29] <ubitux> saste: drawtext is incrementing a frame counter in its context
[17:30] <ubitux> and it looks like it is called even if yadif is dropping the frame
[17:30] <ubitux> which looks weird to me :p
[17:31] <ubitux> it's just like we had drawtext=...,yadif
[17:31] <ubitux> while we actually have yadif,drawtext=...
[17:49] <saste> ubitux: the point is, when is drawtext increasing the counter?
[17:49] <saste> *where
[17:50] <ubitux> draw_text()
[17:50] <ubitux> so well& each time it's fed with a frame
[17:50] <ubitux> oh wait
[17:51] <ubitux> it's actually in end_frame()
[17:51] <ubitux> it's not a simple filter...
[17:51] <ubitux> why it isn't one? oO
[17:51] <ubitux> is there a need for prediction or something?
[17:52] <saste> so is yadif sending an end_frame() even when it didn't do start_frame()?
[17:57] <phunyguy> Hello, I asked this a few days ago and never got a response.  I am trying to demux an mkv with mpeg2 video, ac3 audio, subtitles, and chapters, and put it all into a .vob file.  I can't get chapters and subtitles working.  Can anyone help me figure this out?
[17:59] <tonsofpcs> subtitles are a nuanced thing, but I'm not sure how you expect to get chapters out...
[18:02] <phunyguy> Sorry, this network sucks here
[18:02] <phunyguy> mkvextract can pull chapters from the MKV
[18:03] <phunyguy> just not sure how to get them from there to the vob file.
[18:03] <sacarasc> Probably some DVD authoring software.
[18:03] <phunyguy> Hmmm..  as in dvdauthor?
[18:04] <phunyguy> not looking to make a complete DVD structure... just a single vob file.
[18:14] <grepper> phunyguy: that is stored within the DVD directory structure afaik the IFO file
[18:26] <phunyguy> doh...
[18:26] <phunyguy> hmmm any other containers that support chapter markers?
[18:35] <grepper> trouble with a vob is you would be using mpeg2 anyway, which is poor compared to others with regard to filesize/quality
[18:37] <grepper> I don't know a lot about containers that use chapters, but what is wrong with using mkv ?
[18:38] <grepper> ogg and mov apparently: http://en.wikipedia.org/wiki/Comparison_of_container_formats
[18:41] <phunyguy> grepper, my PS3 will play vob natively from a media server without transcoding.  If I feed it an MKV with MPEG2 video via ps3mediaserver, it messes up pretty bad
[18:42] <phunyguy> a vob works perfect, just no subtitles or chapter markers
[18:42] <sacarasc> What format are the subs?
[18:44] <phunyguy> "dvdsub"
[18:46] <phunyguy> what I am trying to accomplish is to rip my dvd/bluray collection to disc for my media server
[18:46] <phunyguy> makemkv works great, but the ps3 hates the MKVs with the DVD video
[18:47] <phunyguy> and I am trying to avoid transcoding because I have a slow processor, but terabytes upon terabytes of HDD space xD
[18:50] <grepper> you could hardcode the subs
[18:51] <phunyguy> well the subs are low on the priority list, I just didn't want to completely rule them out
[18:52] <grepper> it only plays mpeg's ?
[18:52] <phunyguy> I wonder if there is any advanced settings for PS3mediaserver to detect what type of audio/video is in the mkv and handle accordingly, or run a custom ffmpeg/mencoder command
[18:53] <phunyguy> grepper, it assumes that MKVs need to be completely transcoded, which is OK for x264, etc, but with MPEG2 in that container, for some reason it completely mucks it up.
[18:54] <phunyguy> especially interlaced DVDs
[18:54] <phunyguy> So I thought the next best thing would be to put the file format to what is native to DVDs, which the PS3 picks up flawlessly.
[18:55] <phunyguy> and that would include chapter skipping ,etc.
[18:56] <phunyguy> or so I assumed, lol.
[18:56] <mulla> Hello, i have a mkv video with 9gb. I'd like to convert this to a video with only 1-2gb. Do you know an easy way to do this with ffmpeg?
[18:57] <grepper> phunyguy: can it do mp4 ?
[18:57] <phunyguy> mulla that all depends on where you expect to play it, etc
[18:57] <phunyguy> grepper, havent tried, and I did see that it carried over chapters in that container
[18:57] <grepper> apparently you can add chapters with MP4Box
[18:57] <phunyguy> ffmpeg added the chapters just fine
[18:57] <phunyguy> :)
[18:57] <mulla> <phunyguy>: i want to play this with vlc player
[18:57] <grepper> ah, didn't know that was added
[18:58] <grepper> didn't use to
[18:58] <phunyguy> let me see what containers the ps3 supports and what video in each
[18:59] <phunyguy> grepper, mp4 can be played, but only h264/mpeg4 high profile, with AAC
[18:59] <phunyguy> I think this is a lost cause
[19:01] <grepper> why is that a problem ?
[19:02] <phunyguy> because I am not transcoding these movies on my old opteron processor
[19:02] <phunyguy> :P
[19:04] Action: grepper is spoiled with core i5
[19:04] <phunyguy> i know ps3mediaserver has a custom set of rules, so I think I need to just read up on how to use them.  for example: if container is mkv, and video is AVC, encode to MPEG2, if video is MPEG2, just remux to m2ts, etc.
[19:04] <phunyguy> or present as vo...
[19:04] <phunyguy> vob..
[19:05] <phunyguy> but chapter stamps as well as ubtitles are a complete wash.
[19:05] <phunyguy> s
[19:05] <phunyguy> I would like to leave them in there though for the future, in case I use something else that supports it.
[19:06] <grepper> I guess you really want to centralize this to the media server idea, but I guess you could always just burn rw dvds
[19:06] <phunyguy> well I have the DVDs  :)
[19:06] <phunyguy> I just want to leave them on the shelf.
[19:07] <grepper> ah, thought they were BR and ps3 didn't play that or somesuch
[19:07] <grepper> anyway, this is very offtopic, good luck
[19:07] <phunyguy> thanks
[19:38] <necktie> does anyone know what "-strict experimental" does? i can't find anything on it in the documentation
[19:42] <necktie> burek: you around?
[19:45] <burek> necktie, it just makes sure you understand that the codec is not stable
[19:45] <burek> so you dont come here yelling at developers it doesnt work properly :)
[19:45] <burek> i.e. it's beta :)
[19:46] <necktie> ah okay. thanks :)
[20:34] <DX099> hello, I want to downgrade an mkv video with subtitles. But I don't want all of them. How do I specify one specific subtitle ?
[20:35] <DX099> I know the stream id but I can't seem to select it properly
[20:36] <DX099> with the "-scodec:4 copy " won't work for example, the bad sub stream is always selected
[20:50] <DX099> anyone ?
[20:53] <llogan> DX099: see the documentation for the -map option
[21:11] <DX099> llogan, thanks, I figured map out before reading you but thanks. Now another question : how do I know the proper codec syntax now I saw what is available with "-formats" option
[21:12] <llogan> DX099: i don't really understand your question, but there is also "ffmpeg -codecs"
[21:12] <DX099> llogan, thanks
[21:12] <DX099> that was it
[21:14] <DX099> can you advise me about which codec is the best between x264 and h264 ? I always have problem making the difference
[21:16] <hi117> DX099: x264 is a program that encodes the h264 codec....
[21:18] <DX099> ok, it's no codec itself. That definitely fix my mind, thanks
[21:33] <burek> h264 is the decoder and x264 is the encoder
[21:33] <burek> you can't mix them even if you want :)
[00:00] --- Wed Jun 20 2012


More information about the Ffmpeg-devel-irc mailing list