[Ffmpeg-devel-irc] ffmpeg.log.20151117

[burek]

[02:23:09 CET] <darktoad> anyone knows what the ''-g 50'' is for?
[02:28:38 CET] <intracube> darktoad: GOP length?
[02:29:08 CET] <intracube> strange there seems to be no mention of -g in the manual...
[02:29:14 CET] <intracube> (ffmpeg-all)
[02:30:39 CET] <darktoad> it's in: https://trac.ffmpeg.org/wiki/EncodingForStreamingSites
[02:30:58 CET] <darktoad> -bufsize 6000k -pix_fmt yuv420p -g 50 -acodec libmp3lame -b:a 128k -ac 2
[02:31:03 CET] <darktoad> -g 50? wtf
[02:38:16 CET] <intracube> that's a gop length of 50 frames
[03:58:07 CET] <ekleog> Hello there! Is there an equivalent of adelay for video delaying? I tried setpts=PTS-STARTPTS+[nb_seconds]/TB ; but it clearly does not start video at time [nb_seconds]
[04:08:01 CET] <prelude2004c> hey guys.. anyone know why my ffmpeg can't use more than 1.5 CPU's ?
[04:08:11 CET] <prelude2004c> it shows at 150% sending to nvidia card but it will never exceed it
[04:08:20 CET] <prelude2004c> i used threads 0 , 30 , etc etc.. nothing seems to help
[04:15:59 CET] <waressearcher2> ekleog: hallo
[04:22:30 CET] <DHE> prelude2004c: it will depend on what features are used. ffmpeg might not be able to use more than that. in contrast I've had 750%+ with software x264 encoding.
[04:34:09 CET] <prelude2004c> ya GPU i can't do it so i am stuck
[04:34:22 CET] <prelude2004c> trying to overlock CPU from 2.6 to 3.3 right now to see if it will give me the extra room i need
[04:35:00 CET] <prelude2004c> too bad ffmpeg can't use more than a single core when working with GPU encoding
[04:35:01 CET] <prelude2004c> very odd
[04:35:11 CET] <prelude2004c> i know it works with CPU encoding and x264 but with NVENC nada
[04:35:18 CET] <prelude2004c> unless someone knows a param that will do that
[04:47:43 CET] <prelude2004c> anyone .. solutions ?
[04:47:52 CET] <prelude2004c> need a lifeline over here : )
[05:04:06 CET] <waressearcher2> prelude2004c: hallo mein Freund
[05:14:52 CET] <waressearcher2> http://www.mainconcept.com/eu/products/sdks/gpu-acceleration-sdk/cuda-h264avc.html
[08:52:27 CET] <yongyung> How can I check if a .mka file contains a 24-bit audio stream? Because it's kind of weird, if I convert the file to .wav it converts it to 16 bit, but if I convert it to flac it converts it to a 24 bit flac for whatever reason. ffprobe output:  Stream #0:0(eng): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 256 kb/s (default)
[08:54:27 CET] <chungy_> I don't think AAC can be anything but 16-bit
[08:57:53 CET] <yongyung> chungy: That's what I thought as well but why does ffmpeg choose to encode with 24 bit to .flac then? ffmpeg doesn't do that for other files, it's weird
[08:58:24 CET] <chungy> Well, I could easily be wrong. FWIW, converting from a lossy format to a lossless one isn't usually useful :P
[09:00:38 CET] <yongyung> chungy: I'm just testing some things and was converting stuff around, and somehow when I converted to aac -> wav -> flac the file was substantially smaller than aac -> flac, so I checked why that happens and well... somehow ffmpeg prefers to use 24 bit from aac to flac
[09:02:33 CET] <chungy> Hm right, looking up what fltp is, doesn't look like it fits into any of the fixed-size sample sizes. It's probably going to 24-bit just so it can be fully represented
[09:03:18 CET] <chungy> and yeah, I'm just reminding that turning it into a lossless codec doesn't restore the information that was lost in the first place. Some people make that mistake. Still useful at times.
[09:15:38 CET] <yongyung> Well from what I can see fltp would actually correstpond to planar floats, which makes no sense, 32-bit? Oo
[09:15:58 CET] <yongyung> Something's not right here^^
[09:21:29 CET] <durandal_1707> what's problem?
[09:31:31 CET] <fritsch> I also don't see it
[09:31:39 CET] <fritsch> fltp is the intermediate format after decoding
[10:16:02 CET] <yongyung> Does someone here know how samples in 16 bit/44.1khz .wav files are arranged for stereo? Is it just left channel, right channel alternating?
[10:48:32 CET] <durandal_1707> yongyung: yes
[10:49:39 CET] <yongyung> I would've expected left/right samples to be somewhat close to each other for "normal" songs, turns out I was wrong lol
[11:33:24 CET] <florian_he> hello everyone
[11:33:43 CET] <florian_he> is there someone who knows how to compile ffmpeg with decklink support for linux?
[13:05:56 CET] <exploreshaifali> I am trying to change audio frequency of a video, from doc of ffmpeg I can see that it can be done using -ar flag but is there something that ffmpeg support only some particular frequencies? because whenever I try to change audio frequency it give error and says "Specified sample rate 40000 is not supported"
[13:09:00 CET] <furq> it depends on the audio codec
[13:09:26 CET] <relaxed> exploreshaifali: you can get a valid list with "ffmpeg -h encoder=mp3" (replace mp3 with your codec)
[13:10:22 CET] <exploreshaifali> relaxed, I am not changing/using any codec, only want to alter audio frequency of a video
[13:10:43 CET] <exploreshaifali> I like your nick relaxed :D
[13:12:24 CET] <exploreshaifali> is there any documentation where I can read more about it?
[13:40:24 CET] <relaxed> to change the audio frequency you must encode the audio
[13:42:00 CET] <keyser> hello! i have a qustion regarding libavformat. I'm trying to read raw H264 Nals from a mp4 container, using AVFMT_FLAG_NOPARSE and AVFMT_FLAG_NOFILLIN, but av_read_frame still gives me a complete frame, not the individual nal packets. How do I access the individual nal packets?
[13:45:25 CET] <Mavrik> Parse them from the frame?
[13:46:30 CET] <keyser> How does one do that?
[13:47:27 CET] <Mavrik> Look for NAL start code and split the data?
[13:47:35 CET] <Mavrik> Not sure what's confusing :)
[13:50:31 CET] <keyser> I tried that, but it didn't work. I will check that code again. You are certain that NAL start codes are in there?
[13:53:03 CET] <keyser> uhm, ok, i was looking in avframe->data for nal start codes,, but that is clearly wrong. how do I get the backing input data?
[13:59:35 CET] <keyser> i've checked it again, nal start codes are not in avpacket->data.
[14:19:13 CET] <jonga> hello everybody please where can i find FFserver binary ?
[14:20:21 CET] <Fyr> guys, how to split a FLAC/WAV, having its CUE?
[14:22:09 CET] <furq> i don't think ffmpeg deals with cuesheets at all
[14:22:32 CET] <furq> use cuetools
[14:22:55 CET] <furq> ideally the windows cuetools but the linux cuetools + shnsplit will do it too
[14:23:02 CET] <furq> (they're different things)
[14:23:36 CET] <furq> s/shnsplit/shntool/
[14:24:54 CET] <jonga> hi furq what about FFserver there is the documentation but not the file
[14:26:55 CET] <SianaGearz> i wonder. i just found that Michael resigned as a leader of ffmpeg half a year back; so what happened since?
[14:26:59 CET] <SianaGearz> who replaced him?
[14:27:07 CET] <SianaGearz> did libav make any steps to reconcile the projects?
[14:29:14 CET] <jonga> is there no idea about FFserver ?
[14:30:00 CET] Last message repeated 1 time(s).
[14:40:49 CET] <exebook> http://pastebin.com/ZsJJQvgs
[14:41:21 CET] <exebook> why it says "unknown option flashver" and then flashver is in a list of valid options??
[14:43:01 CET] <BtbN> because it expects a - in front of arguments.
[14:52:08 CET] <exebook> it is a subargument to flv in librtmp format
[15:00:08 CET] <ohmy> hello every body
[15:01:02 CET] <ohmy> i'm using a very exotic Operating System (vxworks) and i'd like to be able to extract both video and audio from mpeg ts, i'm completely new to ffmpeg
[15:02:08 CET] <ohmy> still trying to figure out what modules i need to port to achieve my need, if any have some insights please
[15:10:06 CET] <Fyr> if the program is compilable and works with examples - it will work in your case.
[15:12:00 CET] <ohmy> Fyr: me ?
[15:12:12 CET] <Fyr> yeah
[15:13:49 CET] <ohmy> Fyr: There's nothing actually, VxWorks ia a proprietary OS, it cames without autotools, i'd like to know what parts of ffmpeg libraries i need to cross compile for instance. as far as i'm progressing in reading the documentation it looks that i need libavcodec at least
[15:14:38 CET] <ohmy> Fyr: and also livavfilrer are needed with their dependencies
[15:15:32 CET] <Fyr> ohmy, why don't you compile them all?
[15:16:34 CET] <ohmy> Fyr: actually there's no autotool coming with my cross toolchain
[15:16:53 CET] <ohmy> Fyr: and i guess that my OS (wich not linux) will not be recognized
[15:18:20 CET] <ohmy> Fyr: i need to disable lot of features to get a minimalistic set of stuff to build
[15:19:27 CET] <bove> Which sws_flags are used by default for rescaling?
[15:35:27 CET] <ohmy> god
[15:36:27 CET] <Fyr> sure
[15:48:46 CET] <ohmy> ./configure --disable-xlib --disable-yasm --disable-programs --disable-bzlib  --disable-sdl  --disable-doc --cross-prefix='/opt/WindRiver/gnu/4.3.3-vxworks-6.9/x86-linux2/bin/' --cc='ccarm' --as='asarm' --ranlib='ranlibarm' --cxx='c++arm' --ar='ararm' --nm='nmarm' --arch=arm --target-os=linux
[15:48:55 CET] <ohmy> asarm is my assembler
[15:49:24 CET] <ohmy> configure error with : GNU assembler not found, install/update gas-preprocessor
[15:49:51 CET] <ohmy> asarm --version GNU assembler (Wind River VxWorks G++ 4.3-357) 2.19.51.20090709
[15:50:06 CET] <ohmy> i'm i missing something ?
[15:51:17 CET] <ohmy> trying with --disable-asm
[15:51:52 CET] <ohmy> looks working, just a last question please, how to disable perl support ?
[16:22:55 CET] <ParkerR_> I have an ffmpeg command that captures my desktop and saves to an image at 8fps (Downscaling to 128px width) "ffmpeg -f x11grab -s 1920x1080 -r 8 -i :0.0 -vf scale=128:-1 %d.png" I'm wanting to run a command every time a frame is saved (Passing the image path to another program). I know this isn't strictly ffmpeg but I havent come up with a way to script it yet. Anybody know of a way this could be done?
[16:25:54 CET] <furq> ParkerR_: inotifywait
[16:55:08 CET] <ParkerR_> furq, Cheers inotifywait -m -r -q --format '%w%f' Videos/ | while read FILE; do   sudo bin/dpf/dpfsend -i $FILE; done
[18:17:57 CET] <t4nk568> Hey guys - I installed ffmpeg via homebrew and I'm trying to include <ffmpeg/swscale.h>. I'm getting "fatal error: 'ffmpeg/swscale.h' file not found" - any ideas why this isn't working?
[18:36:44 CET] <ParkerR_> t4nk568, My memory may be iffy but I think swscale was depricated
[18:37:36 CET] <ParkerR_> *deprecated
[18:38:07 CET] <ParkerR_> Ok probbaly not http://ffmpeg.org/doxygen/2.7/swscale_8h.html
[18:38:41 CET] <t4nk568> ah ok
[18:39:20 CET] <ParkerR_> Im not the best one to answer this anyways. Just an avid user :) Havent gotten into the code side of it yet
[18:39:34 CET] <t4nk568> cheers anyway :)
[18:53:19 CET] <t4nk523> Hi to everybody!
[18:54:07 CET] <t4nk523> is there the possibility you have the answer to this question? http://superuser.com/questions/767257/concatenation-of-a-2-second-silence-audio-with-a-normal-audio-not-working
[18:54:59 CET] <t4nk523> Sorry, wrong link, I meant this one: http://superuser.com/questions/725432/add-audio-to-a-specfic-time-at-video-while-merging
[18:58:58 CET] <t4nk523> I will be patient because I am really interested in the answer :9
[19:02:14 CET] <relaxed> t4nk523: mux the audio in a container and then try it
[19:03:07 CET] <relaxed> like matroska
[19:03:24 CET] <Fyr> agreed
[19:12:41 CET] <t4nk523> I am looking how to do it in google, this is my first time using this tool (ffmpeg), thanks for the tip
[19:13:44 CET] <relaxed> ffmpeg -i input.wav -c copy audio.mkv
[19:20:53 CET] <t4nk523> I did it but, I can't see the point of doing so, I just got the audio file in another format(?)
[19:25:13 CET] <t4nk523> I tried to add a delay to the command your provided so I could increase the time duration of the audio file but it didn't work, I got the new audio file in another format with the same time duration:
[19:25:21 CET] <prelude2004c> hey guys.. good day.. anyone know why ffmpeg only uses 1 core when working with NVENC ?
[19:25:35 CET] <prelude2004c> doesn't matter how many threads i use it never goes above the 150%
[19:26:18 CET] <waressearcher2> prelude2004c: hallo
[19:26:21 CET] <t4nk523> ffmpeg -itsoffset 5 -i sound.wav -c copy test.mkv
[19:29:50 CET] <prelude2004c> anyone ?
[19:31:35 CET] <waressearcher2> http://www.mainconcept.com/eu/products/sdks/gpu-acceleration-sdk/cuda-h264avc.html
[19:32:29 CET] <prelude2004c> yes i know how to encode with nvenc
[19:32:38 CET] <BtbN> nvenc runs on dedicated hardware on your graphics card.
[19:32:47 CET] <BtbN> There is no concept of using multiple cores on that.
[19:32:50 CET] <prelude2004c> the issue is.. ffmpeg somehow can't take advantage of the threads to handle the communication between the GPU and the system
[19:33:03 CET] <prelude2004c> not cores on the GPU .. cores on the xeon system
[19:33:14 CET] <BtbN> The CPU is not involved with encoding at all.
[19:33:22 CET] <BtbN> It just sends frames to the GPU and gets encoded data back.
[19:33:37 CET] <prelude2004c> right... but somehow ffmpeg is using 150% for each session
[19:33:53 CET] <prelude2004c> let me post the code i have for you
[19:34:11 CET] <BtbN> 150% what?
[19:34:53 CET] <prelude2004c> http://pastebin.com/AsqnUKcV
[19:35:10 CET] <prelude2004c> yup.. like 150% CPU usage for each instance of encoding
[19:35:19 CET] <prelude2004c> so.. i have say 10 channels i want to encode.. into segmenting and hls
[19:35:34 CET] <prelude2004c> i do that.. and for each encoding job there is a process that the CPU has to do , just as you said
[19:35:46 CET] <BtbN> you are converting to nv12, of course that will use some CPU power.
[19:35:47 CET] <prelude2004c> but.... its maxing out one of each of the cores.. and it wont multithread into other cores
[19:35:50 CET] <prelude2004c> so the fps can't maintain
[19:36:04 CET] <prelude2004c> do you mean i should not convert to nv12 ?
[19:36:06 CET] <BtbN> nvenc usualy can't handle more than 2 sessions anyway.
[19:36:15 CET] <prelude2004c> it can on the new cards
[19:36:19 CET] <prelude2004c> i have the M4000
[19:36:35 CET] <prelude2004c> there is no limitation on the card only the total fps
[19:37:08 CET] <prelude2004c> i wish there was some way to use the cosumer cards and transcode 10 channels in 2 sessions somehow :)
[19:37:16 CET] <BtbN> The nvenc encoder is hard-wired. There is a physical limitation
[19:37:27 CET] <prelude2004c> maybe put something in between to handle the sending / receiving to the cuda cores
[19:37:33 CET] <BtbN> CUDA is not involved at all.
[19:37:40 CET] <BtbN> Only to address the device
[19:37:52 CET] <prelude2004c> yup. these M4000 have about 1000fps at the quality i am workin gon
[19:38:01 CET] <prelude2004c> i have another server that is a much faster CPU and i can do 9 channels
[19:38:06 CET] <prelude2004c> before i run out of cpu power
[19:38:23 CET] <BtbN> That will most likely drop radidly once you run more sessions than the hardware actualy supports
[19:38:26 CET] <BtbN> *p
[19:38:32 CET] <prelude2004c> so its not the card... its basically per CPU core.. ffmpeg can't use more than that 150% so the fps drop as it can't keep up
[19:38:35 CET] <BtbN> They must have hacked some weird session-swapping logic
[19:38:50 CET] <prelude2004c> not sure.. its actualy stats by nvidia
[19:39:06 CET] <BtbN> I'd also guess you run into the PCIe bandwidth limit with that many streams
[19:39:34 CET] <prelude2004c> ya but running pcie v 2 and 3
[19:39:38 CET] <prelude2004c> nowhere near the limitation
[19:40:15 CET] <prelude2004c> i checked card bw.. i am nowhere near the limitation.. can do like 4GB/s .. i am hitting maybe 500MB
[19:41:46 CET] <BtbN> Not sure what you expect, you are running into limitations somewhere. The entire stack is closed source, so go find yourself some nvidia engineer to get details.
[19:42:01 CET] <prelude2004c> the nvidia is not the problem.. only ffmpeg and the CPU
[19:42:08 CET] <prelude2004c> so.. ffmpeg starts the job and sends to the GPU as expected.
[19:42:10 CET] <BtbN> If you say so
[19:42:19 CET] <prelude2004c> only problem is.. the ffmpeg session can't multithread
[19:42:27 CET] <prelude2004c> nothing to do with GPU.. its doing the job ok
[19:42:33 CET] <BtbN> You can't multithread hardware encoding...
[19:43:12 CET] <prelude2004c> ic. so you mean i have to have cores that are fast enough to handle the one process to the GPU
[19:43:21 CET] <prelude2004c> i can send 10 process to GPU but they are all individual
[19:43:45 CET] <prelude2004c> meaning if i setup ffmpeg to use many resources it wont handle it unless each single CPU core si fast enough
[19:43:46 CET] <BtbN> It's _very_ unlikely that you are running into CPU limitations with that the ffmpeg side of the nvenc encoder does.
[19:44:11 CET] <BtbN> So you either are doing more than encoding with ffmpeg, or you are on a horribly slow CPU.
[19:44:13 CET] <prelude2004c> its the only thing that makes sense... same card. .on a newer faster CPU i get 9 channels without fps drop
[19:44:41 CET] <prelude2004c> with this sytsem .. the cpu cores are only 2.6Ghz.. and hitting 140% - 150% which i read was the limitation... so it doesn't work on 1 even
[19:44:57 CET] <prelude2004c> maybe some setting in ffmpeg
[19:45:04 CET] <BtbN> you are definitely not hitting 150% on a single core, that makes no sense.
[19:45:04 CET] <prelude2004c> something that is using up more CPU power than i need to
[19:46:21 CET] <prelude2004c> sent you a png snapshot
[19:47:20 CET] <prelude2004c> http://67.55.1.151/snapshot.png
[19:47:23 CET] <prelude2004c> check out that url
[19:47:38 CET] <prelude2004c> its lower now that i removed the pixel format of nv12
[19:47:47 CET] <BtbN> so there is one core at 25%, and one at 39
[19:48:03 CET] <BtbN> That's far from beeing cpu capped.
[19:48:29 CET] <prelude2004c> yes i know... that is the weird part...
[19:48:46 CET] <prelude2004c> but its something to do with ffmpeg using 1.5 CPUs only maximum
[19:49:04 CET] <BtbN> Because it has nothing else to do that would use CPU?
[19:49:05 CET] <prelude2004c> i was readon about that.. now that i dropped the nv12 for example.. its running only total at 113% which now the fps is lower
[19:51:05 CET] <prelude2004c> ok i  removed the nv12 pixel change.. which lowered computing power per instance. and now i have 5 running and no fps drops
[19:51:10 CET] <prelude2004c> so clearly its a per core issue
[19:57:42 CET] <t4nk523> Could you please help me out with a task I'm trying to do once you finish the current discussion? (I think it should be easy for you due it is only a command)
[20:04:02 CET] <llogan> don't ask to ask; just ask
[20:08:52 CET] <t4nk523> http://pastebin.com/9J9sLqiy
[20:10:21 CET] <t4nk523> what I am trying to do is to merge an audio file into a video at a specific time. I want to keep the original audio from the video so I found that command which works but
[20:10:41 CET] <t4nk523> The audio is always starting at second 0 from the new video
[20:11:58 CET] <t4nk523> I tried to use -itsoffset offset after the audio input to specify a delay but it did not work, I read -itsoffset only works for video containers(?)
[20:13:02 CET] <t4nk523> so the question is... how can I specify the audio should start at the second 7 of the video? (the audio has a time duration of 1 second)
[20:15:01 CET] <c_14> itsoffset is an input option
[20:18:50 CET] <llogan> t4nk523: if itsoffset doesn't do what you want then see adelay filter
[20:20:06 CET] <t4nk523> Ok, I am reading its documentation
[20:28:51 CET] <t4nk523> Great, adelay worked as expected! thanks for your help
[20:33:22 CET] <shincode> if I install a bin from a command line where bin creates a folder... who would be responsible for making the .pc files
[20:33:58 CET] <llogan> what you say !!!
[20:34:02 CET] <shincode> and why dont pkgfiles understand environment variables?
[20:45:06 CET] <furq> what does this have to do with ffmpeg
[00:00:00 CET] --- Wed Nov 18 2015