[Ffmpeg-devel-irc] ffmpeg.log.20160909

burek burek021 at gmail.com
Sat Sep 10 03:05:01 EEST 2016


[00:01:09 CEST] <haagch> interesting. I tried pacat --record -d alsa_output.pci-0000_00_1f.3.analog-stereo.monitor --format=s16be | ffmpeg -f s16be -i - test.opus
[00:01:21 CEST] <haagch> and the resulting file is pretty much half the speed of the original
[00:09:53 CEST] <kepstin> that happens when you interpret stereo audio as mono, yeah. Add "-channels 2" before the -i on ffmpeg.
[00:10:23 CEST] <kepstin> could be a sample rate issue too, you'll want to explicitly specify sample rate on both ends
[00:10:45 CEST] <kepstin> (it'll default to 44.1KHz in ffmpeg, while your pulseaudio command might be different)
[00:11:11 CEST] <kepstin> hmm, looks like pacat defaults to 44.1 as well
[00:11:28 CEST] <haagch> it was stereo
[00:11:29 CEST] <haagch> thanks a lot
[00:12:34 CEST] <kepstin> also, if you're running on an intel or arm processor box, 's16be' will require a format conversion on both ends, you probably want to use 's16le' instead.
[00:14:04 CEST] <haagch> oh it means endianess
[00:14:19 CEST] Action: haagch makes an epiphany sound
[00:28:32 CEST] <Enverex> kepstin: Still seems to happen with Compton unfortunately
[00:34:02 CEST] <infinito> how can i deinterlace a video in a nvenc workflow?
[00:35:53 CEST] <infinito> it is posible use yadif over nvenc?
[01:07:04 CEST] <DHE> infinito: as long as the frame hits the main memory, yes. some people have tried to do pure-GPU conversion (decoding, scaling, encoding) in which case that wouldn't work
[01:25:32 CEST] <Victorma> hi! anyone knows what does it mean return -12 in av_read_frame, besides of meaning that i am having any kind of error?
[01:26:01 CEST] <Victorma> or, when can I have valuable info about that code?
[01:26:05 CEST] <Victorma> where*
[01:26:44 CEST] <DHE> many codes are errno values, but negative. on linux -12 is Out of memory
[01:27:00 CEST] <Victorma> I've googled but no references at all
[01:27:16 CEST] <Victorma> it's not windows, is an arm device, but could be
[01:27:21 CEST] <Victorma> sorry, not linux*
[01:28:53 CEST] <Victorma> so, the code is generated by the core supporting ffmpeg
[01:29:11 CEST] <Victorma> ?
[01:29:44 CEST] <DHE> yes, errno is a posix standard
[01:29:59 CEST] <DHE> and most arm systems will be running linux, at least if you can compile and install software on it
[01:30:24 CEST] <Victorma> this is not running linux, it's a 3ds hehe, in fact is compiled as "none" os
[01:30:53 CEST] <DHE> oh...
[01:31:09 CEST] <DHE> uhhh.. well, that might explain tight memory if the error codes are the same
[01:31:48 CEST] <Victorma> could be, but i've had memory problems before and all of them where resulting in a stack overflow
[01:32:02 CEST] <Victorma> and this is not
[01:32:09 CEST] <Victorma> i'll try to increase the heap
[01:32:22 CEST] <Victorma> sorry, not stack overflow, out of stack
[01:32:28 CEST] <Victorma> out of heaf
[01:32:33 CEST] <Victorma> heap* god
[03:38:24 CEST] <ferdna> guys i need help... i need to get a video source to a webm format...
[03:38:45 CEST] <ferdna> ive tried different settings but cant get my conversion video to look good and real-time
[06:16:58 CEST] <jpmh> I am using: ffmpeg -r 1 -i x_%d.jpg  test3.mp4, I have 149 images that meet the spec, I get a 15 second video made of the first 1 images.  I clearly don't understand the parameters, how do I make the 149 second video I expect?
[06:19:01 CEST] <relaxed> how are the images named?
[06:21:31 CEST] <jpmh> relaxed, they are named x_1.pg, x_2.jpg... theough x_149.jpg
[06:23:11 CEST] <relaxed> try, ffmpeg -framerate 1 -i x_%d.jpg -r 25 -tune stillimage -pix_fmt yuv420p out.mp4
[06:27:28 CEST] <jpmh> relaxed, what does the -tune parameter do?
[06:31:08 CEST] <relaxed> it's an x264 encoder option to optimize for encoding still images
[06:32:03 CEST] <jpmh> testing now, stand by
[06:33:03 CEST] <jpmh> no, now I get a file with just the first image, and that's it
[06:43:08 CEST] <jpmh> fflogger - neither of those urls work for me
[06:43:37 CEST] <relaxed> fflogger is a bot
[06:44:17 CEST] <relaxed> http://pastebin.com/ works for me
[06:47:11 CEST] <jpmh> relaxed, correct, he gave me www. - pasting now
[06:48:41 CEST] <jpmh> pasted - shouldn't it give me a link to share with you?
[06:51:45 CEST] <jpmh> http://pastebin.com/rDDxddYV
[06:57:50 CEST] <relaxed> did you omit some of the console output? I don't see the version.
[07:01:48 CEST] <relaxed> pastebin the command I gave you
[07:03:35 CEST] <jpmh> relaxed, not intentionally, let me check - I re-directed the output to a file and copied the file so generated to pastebin
[07:04:25 CEST] <jpmh> ah, ok - will repeat with the command you gave me too - stand by, I was assuming you wanted to see what I had been doing and would help me with what I was doing wrong
[07:08:34 CEST] <jpmh> the paste of the output from your command is at: http://pastebin.com/7sNJhQmr
[07:08:55 CEST] <jpmh> interestingly, I see that my sysadmin has aliased ffmpeg to avconv, is that the issue?
[07:09:09 CEST] <jpmh> I had no noted that before
[07:09:41 CEST] <jpmh> I have alsways used ffmpeg and I just did that without thinking/checking, since it appears to work, and for eveything else I have done, has worked?
[07:11:39 CEST] <relaxed> yeah, that's from the fork libav.org. I recommend you compile a recent ffmpeg or use my static build, https://www.johnvansickle.com/ffmpeg/
[07:14:28 CEST] <jpmh> relaxed, the other thing I do note, about your command is that it is makign a file that appears to be 10 times bigger, is it making multipleframes from each iput image whereas mine is making 1
[07:15:52 CEST] <relaxed> yes, it's duplicated each frame to 25
[07:16:56 CEST] <relaxed> you can remove -r 25
[07:19:50 CEST] <jpmh> relaxed, I really don't want to alienate the sysadmin, so would prefer not to install anything new, is the fork that bad?
[07:20:59 CEST] <relaxed> No, but for support you need to ask in #libav
[07:21:52 CEST] <jpmh> AND THE REALY GOOD NEWS.  Now I have removed the -r 25 YOUR code works like a champ - thank you SO MUCH.  I really appreciate your patience
[07:22:09 CEST] <jpmh> now I just have to go and read the man pages to understand what it is doing
[07:23:08 CEST] <relaxed> \o/
[07:24:59 CEST] <jpmh> yes, I realize that now I see that my ffmpeg has been aliased to avconv.  I had not noticed that - but as I say, thanks to your help it it all working as I hoped.  Thank you so much
[07:25:15 CEST] <relaxed> you're welcome
[07:35:58 CEST] <jpmh> relaxed, so I did a little testing, it is really the change from -r to -framerate that was required.  The other parameters you gave did not really seem to be needed.  The docs are VERY unclear as to what -r and -framerate mean, at least to me.  Can you explain to a novie like me
[07:37:34 CEST] <relaxed> -framerate 1 tells the demuxer the input fps is 1 fps
[07:37:52 CEST] <jpmh> in fact, the man pages I can find for ffmpeg do not even seem to have -framerate as an option
[07:38:14 CEST] <relaxed> it's in the man page
[07:39:25 CEST] <jpmh> I have the man page up right now, on a different system and a search for -framerate does not find anything.  The man page for avconvert does, and it just says "set the frame rate"
[07:39:51 CEST] <jpmh> doesn't -r, as I was doing, also tell it one per second
[07:40:27 CEST] <relaxed> https://ffmpeg.org/ffmpeg.html  look at "-r[:stream_specifier]"
[07:40:29 CEST] <jpmh> indeed, it did generate the first 15 seconds of vidoe exactly as does your version, it just stopped at frame 15 whereas yours goes to frame 149, as it should
[07:41:49 CEST] <jpmh> wow - your reference there is GREAT, thanks, reading it now
[07:43:03 CEST] <jpmh> however, I still do not see why using -r 1 caused it to generate only a 15 second file whereas -framerate 1 generates the full expected length.  Can you explain that?
[07:44:53 CEST] <relaxed> look at your output, http://pastebin.com/rDDxddYV   Notice "drop=134" in the output
[07:46:08 CEST] <jpmh> tep, I see a bunch of "drops" there, what are they?
[07:56:46 CEST] <relaxed> they're your missing frames, becuase you were using the wrong input option
[11:44:14 CEST] <zoeurk> hello, is it possible to encode (from a webcam) at the same time audio and video?
[11:44:48 CEST] <BtbN> sure. Would be kinda pointless if you could only encode audio or video.
[11:46:17 CEST] <zoeurk> how I can do that please
[11:46:45 CEST] <BtbN> do what exactly?
[11:47:08 CEST] <zoeurk> encode from my webcam audio and video
[11:47:30 CEST] <bizzeh> hey, i have 2 videos, both encoded in the exact same way, using the same settings. when i run them through ffprobe. the audio for one of them comes out as this "aac (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 164 kb/s (default)" and the other comes out as "aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 163 kb/s (default)". why would one of them show "(LC)" and the other not?
[11:47:30 CEST] <BtbN> https://trac.ffmpeg.org/wiki/Capture/Webcam
[11:47:35 CEST] <BtbN> first google hit for ffmpeg + Webcam.
[11:47:36 CEST] <zoeurk> thanks
[12:01:11 CEST] <dehun> hi guys. I am looking for a way to disable CODEC_FLAG_LOW_DELAY from command line for ffmpeg.
[12:04:26 CEST] <relaxed> dehun: I think -flags +low_delay
[12:05:34 CEST] <dehun> ffmpeg is 2.7.2. The reason I would like to do that is because I am getting a message "Delayed frames seen reenabling low delay requires a codec flush" from time to time during decoding process to rawvideo.
[12:07:08 CEST] <BtbN> you should update your ffmpeg.
[12:08:03 CEST] <dehun> @relaxed: it looks like +low_delay will do right the opposite - it will enable CODEC_FLAG_LOW_DELAY thing =)
[12:09:05 CEST] <relaxed> sorry, I just glanced at your question. Are you sure it's on by default?
[12:09:56 CEST] <BtbN> Isn't that an encoder flag anyway?
[12:10:08 CEST] <relaxed> decoder too?
[12:10:23 CEST] <BtbN> What would an encoder do diffrently when told to have low delay?
[12:10:29 CEST] <BtbN> It decodes as fast as it can anyway
[12:10:33 CEST] <BtbN> *decoder
[12:11:23 CEST] <dehun> it looks like it is on by default in my case.
[12:12:07 CEST] <BtbN> The message you pasted does not exist in current ffmpeg btw.
[12:12:14 CEST] <BtbN> And 2.7.2 is not even the most recent 2.7
[12:12:17 CEST] <BtbN> you really should update.
[12:14:15 CEST] <relaxed> maybe -flags -low_delay
[12:15:46 CEST] <dehun> @BtbN: well, decoder can drop some video frames. Thas is what happens in our case. We got flaky tests in case when we run multiple ffmpeg instances decoding stuff to rawvideo.
[12:16:18 CEST] <dehun> Thanks guys. I will try to specify -flags -low_delay. And then will see is it possible to update
[14:51:43 CEST] <dehun> tried both options. Updated to 3.0.2 and still get framedrops during cpu limited time spans. Any ideas how to force ffmpeg to not drop frames at all? Output parameters looks like this '-vcodec', 'rawvideo', '-f', 'rawvideo', '-pix_fmt', 'rgb24'?
[15:02:15 CEST] <BtbN> ffmpeg will slow down when your CPU is too slow, not much you can do about it
[15:02:24 CEST] <BtbN> reduce the quality to speed up the encoding
[15:02:34 CEST] <BtbN> Or get a faster CPU
[15:04:53 CEST] <dehun> if it will just slow down writing to output file that is fine with me.What I dont want is frame dropping to speed-up stuff
[15:07:26 CEST] <BtbN> it does not drop frames to stay at real time. It will happily slow down to 1fps and slower.
[15:08:17 CEST] <BtbN> Unless you are transcoding from a live source. Then it will eventually run out of buffer, and new data will be discarded until there's room again.
[15:08:44 CEST] <BtbN> In that case there's nothing you can really do, except for making encoding fast enough to keep up at real time.
[15:18:34 CEST] <iive> dehun: try to output into /dev/null, to check if your bottlenech is (not) in the disk writing.
[15:19:42 CEST] <iive> also, avoid color format conversion if possible, if your input is yuv420 then conversion to rgb24 not only slows you down, it also uses twice more storage.
[15:58:53 CEST] <ses1984> hey, i'm not that knowledgeable about all the correct terminology so i was hoping a human could help me figure out what to search for... i was wondering if ffmpeg would work for my use case
[16:00:02 CEST] <ses1984> there are two live video feeds i want to consume, one is over rtsp and the other is something like... asf over http? not sure ... i was wondering if i could get the stream, apply some filters, and then pipe it to something else like a gui video player
[16:04:20 CEST] <ses1984> i'm not sure if ffmpeg is the right tool for the job, or there is some other tool i need to use instead of, or in addition to, ffmpeg
[18:16:18 CEST] <persina> I have files in a directory labeled test.1990.png, test.1991.png, test.2015.png, etc. how do I target these with ffmpeg?
[18:19:47 CEST] <jkqxz> ffmpeg -f image2 -start_number 1990 -i 'test.%04d.png' ...
[18:23:26 CEST] <persina> jkqxz: "Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height" what's this mean?
[18:25:51 CEST] <jkqxz> That you should look in the lines above for what the actual error was.
[18:30:23 CEST] <persina> jkqxz: "[libx264 @ 00000000027c08a0] width not divisible by 2 (2457x1589)" the png sizes have to be divisable by 2?
[18:33:58 CEST] <DHE> x264 requires it to be that way, yes
[18:34:19 CEST] <DHE> you can stretch it to meet that requirement in ffmpeg if you want
[18:35:05 CEST] <persina> DHE: How do I do that?
[18:35:58 CEST] <DHE> before the output filename put "-s 2456x1588
[18:37:46 CEST] <persina> DHE: Does that change the resolution of the mp4?
[18:40:28 CEST] <DHE> ... what mp4?
[18:45:12 CEST] <persina> the mp4 ffmpeg creates from the png's
[19:12:52 CEST] <persina> what's the difference between the -r and -framerate paramaters?
[19:14:47 CEST] <Lirk> hi all
[19:14:57 CEST] <relaxed> persina: -framerate declares the frame rate of an input stream and -r changes it
[19:15:37 CEST] <Lirk> how i can to remove metadata such as service_provider and service_name from mpegts muxer?
[19:16:00 CEST] <Lirk> map_metadata set doesn't work
[19:20:14 CEST] <persina> relaxed: What param is used to slow down the fps?
[19:24:38 CEST] <llogan> persina: see setpts or framerate filters depending on what you actually want to do
[19:24:59 CEST] <DHE> well I was going to answer lirk...
[19:28:21 CEST] <llogan> new to IRC i suppose
[19:29:33 CEST] <JEEB> t
[19:46:58 CEST] <lindylex> I want to build MPD ver 0.19.17  from source and it utilizes ffmpeg version less than 3.1 currently installed on my Debian testing desktop.  How can I install and older version of ffmpeg?
[19:49:01 CEST] <llogan> why not install mpd 0.19.19 instead?
[19:50:11 CEST] <lindylex> It has a bug.  When I leave it playing for over 3 hours it starts to repeat small sections of a song.
[19:50:33 CEST] <llogan> did you report it upstream?
[19:50:51 CEST] <llogan> what about mpd from git?
[19:57:19 CEST] <lindylex> One second I have to reboot.  I guess I will try that.
[19:57:26 CEST] <furq> llogan: apt-cache showpkg ffmpeg
[00:00:00 CEST] --- Sat Sep 10 2016


More information about the Ffmpeg-devel-irc mailing list