[Ffmpeg-devel-irc] ffmpeg.log.20151009

burek burek021 at gmail.com
Sat Oct 10 02:05:01 CEST 2015

[00:24:19 CEST] <PlanC> I'm converting a bunch of old video and M4A files
[00:24:50 CEST] <PlanC> to MP3 format
[00:25:17 CEST] <PlanC> does FFMpeg automatically set the bitrate to the audio bitrate of the video?
[00:25:35 CEST] <PlanC> or do I have to put in an additional parameter to have it do that?
[00:26:34 CEST] <grublet> PlanC: -c:v copy -c:a copy
[00:26:45 CEST] <grublet> oh wait
[00:26:49 CEST] <grublet> you're transcoding
[00:26:49 CEST] <grublet> nvm
[00:27:15 CEST] <Max-P> PlanC: It has defaults, but if you want to be sure just specify them to your taste with the -b:a option
[00:28:20 CEST] <Max-P> It can't really guess settings, especially if you're transcoding because the bitrate doesn't have the same effect on all codes. For example, AAC requires less bits than MP3 for similar quality
[00:34:00 CEST] <PlanC> but the thing is that each video and M4A has a different bitrate
[00:34:28 CEST] <PlanC> how can I find the optimal bitrate for each?
[01:37:35 CEST] <alesan> so I am trying to use the output of "convert" (imagemagick) to feed ffmpeg
[01:37:41 CEST] <alesan> so far I had very little success
[01:38:16 CEST] <alesan> I have convert output to a pipe, that is then fed by bash to ffmpeg
[01:38:53 CEST] <alesan> I have tried to use jpeg output in convert, and I specify -vcodec mjpeg in ffmpeg it scales it down to 160x120 ... ???
[02:12:38 CEST] <alesan> I updated ffmpeg, now it seems OK
[04:02:44 CEST] <alesan_> hello
[04:03:21 CEST] <alesan_> pipe:: Cannot allocate memory I get this error message and then ffmpeg hangs
[04:04:16 CEST] <alesan_> http://pastebin.ca/3186422 this is the invocation
[04:04:27 CEST] <alesan_> I am trying to pass raw YUV data from convert to ffmpeg
[04:04:46 CEST] <alesan_> if I use jpeg, it works with pretty much the same script
[04:07:09 CEST] <relaxed> alesan_: you need to use -f rawvideo and specify each of these demuxer options "ffmpeg -h demuxer=rawvideo"
[04:08:20 CEST] <alesan_> -video_size<image_size> is this in the format 1920x1080?
[04:08:39 CEST] <relaxed> yes
[04:08:48 CEST] <alesan_> let's see what it does
[04:09:35 CEST] <alesan_> it does not throw an error at least :)
[04:09:37 CEST] <alesan_> yes it'
[04:09:54 CEST] <alesan_> s encoding something that looks probably correct
[04:10:24 CEST] <alesan_> is it an impression or... it's slower than if I use JPEG? Is it possible???
[04:11:23 CEST] <relaxed> can ffmpeg's filters not cover what you're doing with convert?
[04:37:08 CEST] <alesan_> well
[04:47:25 CEST] <alesan_> I need to perform some benchmark
[05:20:18 CEST] <alesan_> so
[05:20:41 CEST] <alesan_> if I use YUV the script takes real    14m4.462s - user    10m39.036s - sys     20m33.324s
[05:20:54 CEST] <alesan_> if I use JPEG real    2m9.359s - user    6m6.440s - sys     0m38.124s
[06:26:06 CEST] <alesan_> soooo... if I completely remove the ffmpeg (and only leave the yuv or jpeg generation) I get:
[06:26:21 CEST] <alesan_> yuv: real    5m7.452s - user    8m58.684s - sys     3m23.196s
[06:26:35 CEST] <alesan_> jpeg: real    2m7.369s - user    5m47.088s - sys     0m36.784s
[06:27:16 CEST] <alesan_> so, for whatever reason, yuv takes more than double, when I would expect it takes quite a bit less, meaning that to encode jpeg it has to get to the same yuv anyway, and then proceed with compression
[06:28:12 CEST] <alesan_> thing is, in case of jpeg the whole video encoding speed is limited by the generation of the image
[06:28:40 CEST] <alesan_> real 2m7s with no ffmpeg and 2m9s with video encoding
[06:28:53 CEST] <alesan_> but if I use yuv, both generation and encoding take so much more time
[06:29:35 CEST] <alesan_> 5m7s yuv generation only and 14m11s with ffmpeg
[06:30:13 CEST] <alesan_> what could possibly be wrong, I mean there is more data to be passed on the pipe, but I'm not piping through a dialup modem ;)
[06:39:52 CEST] <relaxed> alesan_: well, jpg is lossy
[06:40:42 CEST] <alesan_> relaxed, jpeg quantization filtering etc should take much longer to process than spitting out the yuv as it is already
[06:40:55 CEST] <alesan_> http://pastebin.ca/3186564
[06:41:17 CEST] <alesan_> see the two scripts, I made sure they generate the same type of subsample
[06:41:26 CEST] <alesan_> 4:2:0 (the opriginal jpegs are 4:2:2)
[06:42:39 CEST] <relaxed> again, can you not achieve what you want with ffmpeg filters?
[06:46:12 CEST] <alesan_> that convert command resizes from 7Mpixel to "full HD" (also changing the ratio) and adds some text on each frame (date and time extracted from exif)
[06:46:33 CEST] <alesan_> I am not aways of ffmpeg doing compositing or drawings on frames?
[06:47:25 CEST] <alesan_> relaxed, ^^^
[06:47:36 CEST] <relaxed> there's a scale and drawtext filter
[06:48:01 CEST] <alesan_> cool, I did not know.
[06:48:03 CEST] <relaxed> https://trac.ffmpeg.org/wiki/FilteringGuide
[06:48:20 CEST] <alesan_> but will I have to rename all the images for ffmpeg to like them as input?
[06:48:59 CEST] <relaxed> you can use globs
[06:49:25 CEST] <relaxed> ffmpeg -h demuxer=image2
[06:51:57 CEST] <alesan_> fcrom the drawtyext documentation (hey it's a great filter) I am not sure how to run an expression to determine the text on each input frame
[06:52:08 CEST] <alesan_> it might be possible, or not, I am still not sure
[10:34:32 CEST] <DarthDepa> Hi guys! :) I have a problem
[10:35:02 CEST] <DarthDepa> I'm compiling "ffmpeg" (from master branch of GIT repo) on my Raspberry PI (Model B)
[10:36:40 CEST] <DarthDepa> I get this error message
[10:37:14 CEST] <DarthDepa> http://pastebin.com/SqgCZ0Xa
[10:37:31 CEST] <DarthDepa> (I'm going to share with you the config.h too, just a moment)
[10:39:10 CEST] <DarthDepa> Here the config: http://pastebin.com/2g2WprAt
[10:39:13 CEST] <DarthDepa> why???? :'(
[10:40:11 CEST] <c_14> config.h isn't particularly interesting. What's in config.log? first and last 50 lines or so are usually enough
[10:40:42 CEST] <DarthDepa> ok, just a moment
[10:40:43 CEST] <DarthDepa> wait
[10:42:28 CEST] <blue112> Hi here.
[10:42:46 CEST] <blue112> Is there a way to have an hardware acceleration when using swscale to convert from yuv420p to argb ?
[10:44:21 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721783/
[10:44:25 CEST] <DarthDepa> thanks
[10:45:56 CEST] <DarthDepa> I'm going crazy lol
[10:47:06 CEST] <c_14> Hmm, do you have more of the compilation output?
[10:47:16 CEST] <DarthDepa> yeah, wait
[10:47:16 CEST] <c_14> The lines preceding those you pasted.
[10:47:17 CEST] <DarthDepa> lol
[10:48:18 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721802/
[10:48:20 CEST] <DarthDepa> thanks
[10:48:30 CEST] <DarthDepa> this is the complete output
[10:50:22 CEST] <DarthDepa> (I want to use the compiled version for UMS lol)
[10:51:54 CEST] <BtbN> blue112, if you didn't explicitly disable it, it will use whatever extensions your CPU has available, provided someone wrote code for that.
[10:52:43 CEST] <blue112> BtbN, I didn't disabled anything, but I have the warning: [swscaler @ 0x7f31f80008e0] No accelerated colorspace conversion found from yuv420p to argb.
[10:52:52 CEST] <blue112> When trying to use swscale()
[10:52:59 CEST] <blue112> Still, it works, but its kinda slow.
[10:53:13 CEST] <BtbN> Then nobody wrote code for that conversions, and it's using a fallback slow path.
[10:54:02 CEST] <c_14> DarthDepa: can you try again with make V=1 ?
[10:54:21 CEST] <DarthDepa> c_14: yeah :) which is the difference?
[10:54:27 CEST] <DarthDepa> c_14: I'm trying...
[10:54:39 CEST] <c_14> It outputs the commands it uses to compile.
[10:54:43 CEST] <c_14> So it's easier to debug.
[10:54:51 CEST] <c_14> Or in this case to link.
[10:55:08 CEST] <c_14> The linker path is probably not set correctly.
[10:56:07 CEST] <DarthDepa> ok, thanks
[10:56:28 CEST] <DarthDepa> Ok, I got the new output
[10:56:31 CEST] <DarthDepa> please, wait
[10:57:01 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721829/
[11:04:21 CEST] <c_14> What version are you trying to compile?
[11:04:50 CEST] <DarthDepa> c_14: 2.8.git
[11:04:59 CEST] <DarthDepa> From file RELEASE
[11:05:39 CEST] <DarthDepa> and if I run version.sh I get
[11:05:44 CEST] <DarthDepa> N-75841-g5911eeb
[11:07:34 CEST] <c_14> Can you run `objdump -t libavfilter/libavfilter.a' and look for the section headed transform.o ?
[11:08:53 CEST] <DarthDepa> yeah
[11:08:55 CEST] <DarthDepa> just a moment
[11:11:10 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721894/
[11:11:18 CEST] <DarthDepa> c_14: I was not able to find that head
[11:11:19 CEST] <DarthDepa> :(
[11:12:05 CEST] <c_14> Do you have a libavfilter/transform.o ?
[11:12:32 CEST] <DarthDepa> Yep :)
[11:13:25 CEST] <DarthDepa> but is empty
[11:13:30 CEST] <c_14> eeeh
[11:13:32 CEST] <c_14> That's not good.
[11:13:40 CEST] <DarthDepa> lol
[11:15:15 CEST] <c_14> Can you rm libavfilter/libavfilter.a and then run `make V=1 libavfilter/libavfilter.a' ?
[11:17:10 CEST] <DarthDepa> Yeah, just a moment
[11:18:17 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721927/
[11:19:10 CEST] <c_14> libavfilter/transform.o is still empty?
[11:19:27 CEST] <DarthDepa> yep :(
[11:19:49 CEST] <DarthDepa> I have to run again "make V=1"?
[11:19:51 CEST] <c_14> rm libavfilter/transform.o; make V=1 libavfilter/transform.o
[11:19:59 CEST] <DarthDepa> ok, wait
[11:20:45 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721936/
[11:21:13 CEST] <DarthDepa> And now is not empty
[11:21:14 CEST] <DarthDepa> :)
[11:21:33 CEST] <DarthDepa> Now is "ELF 32-bit LSB relocatable, ARM, EABI5 version 1 (SYSV), not stripped"
[11:23:32 CEST] <c_14> make libavfilter/libavfilter.a; make
[11:23:35 CEST] <c_14> maybe with V=1 just in case
[11:23:41 CEST] <DarthDepa> ok, wait
[11:23:42 CEST] <DarthDepa> lol
[11:26:21 CEST] <DarthDepa> It seems not working... I'm posting the output :(
[11:26:40 CEST] <DarthDepa> c_14: http://paste.ubuntu.com/12721961/
[11:27:39 CEST] <c_14> is libavfilter/split.o perchance also empty?
[11:28:00 CEST] <DarthDepa> Yep, is empty
[11:28:45 CEST] <c_14> rm libavfilter/*.o libavfilter/libavfilter.a; make V=1 libavfilter/libavfilter.a
[11:30:13 CEST] <DarthDepa> Yeah, wait
[11:31:46 CEST] <DarthDepa> It seems long...
[11:31:50 CEST] <DarthDepa> lol please, wait
[11:32:19 CEST] <c_14> Probably don't need the output for that (as long as it works. Just check that transform.o and split.o aren't empty again.
[11:33:54 CEST] <DarthDepa> c_14: Still compiling
[11:38:56 CEST] <waressearcher2> DarthDepa: what linux distributive you use ?
[11:39:35 CEST] <DarthDepa> waressearcher2: Raspbian (Debian 8 based), for my Raspberry Pi Model B+
[11:40:14 CEST] <DarthDepa> waressearcher2: Why? :)
[11:50:00 CEST] <DarthDepa> c_14: Done! :) they are not empty
[11:50:02 CEST] <DarthDepa> now?
[11:50:15 CEST] <DarthDepa> I can run the "big make" again?
[11:51:53 CEST] <c_14> yep
[11:52:10 CEST] <DarthDepa> ok :)
[11:55:04 CEST] <DarthDepa> c_14: Seems that works well now..
[11:56:13 CEST] <DarthDepa> It is on "ffserver.o" at the moment, so I think that it's ok ;)
[11:56:20 CEST] <DarthDepa> c_14: thanks so much, dude
[11:57:25 CEST] <c_14> Oh, just as a question, but why are you compiling with --enable-nonfree when you don't need it?
[11:57:32 CEST] <DarthDepa> If I get more errors I come back
[11:59:53 CEST] <DarthDepa> c_14: because I copied the options from a blog of a guy that make it works for Raspbian
[11:59:57 CEST] <DarthDepa> lol
[12:04:11 CEST] <waressearcher2> DarthDepa: I mean if you use ubuntu why compile anything
[12:30:06 CEST] <DarthDepa> waressearcher2: Yeah lol but on Raspbian there isn't a package in the repo :'(
[12:34:32 CEST] <djems54> Hi, I wish to broadcast/re-stream a stream rtsp (h264, aac) with ffserver (output stream) to another ffserver (feed input), but I have this error message : "Invalid data found when processing input" when I use the "rtp" format. What do I stream was using ?
[12:47:39 CEST] <DarthDepa> c_14: Ok, its finished but there is a little problem
[12:47:57 CEST] <DarthDepa> c_14: When I use it, from UMS, I got this error message
[12:48:01 CEST] <DarthDepa> "[AVFilterGraph @ 0x1dcd9b0] No such filter: 'ass'"
[12:48:11 CEST] <DarthDepa> "Error opening filters!"
[12:48:14 CEST] <c_14> You need to build with libass support
[12:48:31 CEST] <DarthDepa> oh, nice lol
[12:48:45 CEST] <c_14> and I think libfreetype and probably fontconfig
[12:50:27 CEST] <DarthDepa> always via che configure, right?
[12:50:31 CEST] <benbro> when cutting and concatenating mp4 videos, can I do it without re-encoding?
[12:50:32 CEST] <DarthDepa> --enable-xxxx
[12:52:16 CEST] <c_14> yep
[12:52:25 CEST] <xerox> benbro: yes
[12:52:44 CEST] <xerox> benbro: https://trac.ffmpeg.org/wiki/Concatenate
[12:54:10 CEST] <DarthDepa> c_14: Ok lol I will try... simply "make clean" and new run of "configure"?
[12:54:35 CEST] <benbro> xerox: nice. thanks
[12:59:25 CEST] <c_14> DarthDepa: yep
[13:00:33 CEST] <DarthDepa> c_14: Pray for me ahahah
[15:47:36 CEST] <bove> Would there be any way of applying this function with lutrgb? (t > 0.149658) ? exp((t - 0.385537) / 0.107353233) : (t / 0.9661776 - 0.04378604);
[15:47:47 CEST] <bove> (t is the input value)
[17:55:52 CEST] <wingless> Hello everyone. I'm trying to split a video to 5s parts to no avail.
[17:56:26 CEST] <wingless> $ ffmpeg -i bunny.mp4 -vcodec copy -acodec copy -map 0 -segment_time 5 -f segment output%03d.mp4
[18:00:34 CEST] <wingless> only the first segment is playable. other segments can only be played in mplayer, but not in vlc or any other player I tried.
[18:44:12 CEST] <Emmett> Good morning folks.  I am a unix sysadmin who has inherited a system that utilizes ffmpeg and suddenly, the system has started going crazy.  The load will skyrocket to 200-300 and peg the CPU.  The only odd thing we seem to be able to find is small (5k) videos that come through that result in a muxing overhead of 23.338% !  Unfortunately, I know very little (i.e. zip) about ffmpeg, and this is a production system that I can't play with.  With all of the loggin
[18:45:21 CEST] <JEEB> and now you hoåefully learn to enable linebreaking of long lines in your client :)
[18:49:29 CEST] <ln-> to put it in another way, your message was cut at "With all of the loggin" ...
[18:50:22 CEST] <Emmett> Thank you.  It's been probably 10 years since I've been on IRC&I forgot that there was a line limit.
[18:50:42 CEST] <Emmett> I was trying to keep from sending 5 separate messages.
[18:50:58 CEST] <Emmett> & With all of the logging that ffmpeg produces, I am hoping there is an indicator in there of the problem, but I don't seem to be able to find it.\
[18:58:06 CEST] <ChocolateArmpits> Emmett: If it's not a secret, what is the system intended to do and are those small videos part of the process ?
[19:00:29 CEST] <Emmett> They are basically video voicemails.  Users connect to the website (which runs apache) and ffmpeg ends up getting called to convert the video to stream it out to them.
[19:01:49 CEST] <Emmett> The system runs fine then all of a sudden the ffmpeg processes start stacking up (they all complete, but they take forever)&and the only thing that really jumps out at us is these small ones with the really large muxing overhead.
[19:04:46 CEST] <ChocolateArmpits> So it's using ffmpeg through command line ?
[19:05:17 CEST] <ChocolateArmpits> Or starts it through some application ?
[19:05:24 CEST] <Guest79016> Hi everyone. can anyone help me in merging GIF files where each GIF plays for 5 seconds?
[19:07:24 CEST] <qroft> at this moment i am having this: ffmpeg -f concat -i files.txt -c:v libx264 test.mp4
[19:08:12 CEST] <Emmett> Apache kicks off a shell script which runs ffmpeg (when it's done, appache sends a 302 redirect to the converted file)
[19:09:04 CEST] <ChocolateArmpits> Did you try investigating the script ?
[19:10:53 CEST] <qroft> ChocolateArmpits: you mean me?
[19:11:05 CEST] <ChocolateArmpits> qroft: Sorry, writing to Emett
[19:11:21 CEST] <qroft> ChocolateArmpits: np :)
[19:11:53 CEST] <ChocolateArmpits> qroft: If you have some issues I can try helping you too, just joined the channel so if you have written something before that last message of yours, please paste it again or something
[19:12:25 CEST] <Emmett> The "script" is just a one liner to set some variables and call ffmpeg
[19:13:03 CEST] <ChocolateArmpits> Those variables I guess are output file and the like, right ?
[19:13:57 CEST] <Emmett> Yep.
[19:14:11 CEST] <qroft> ChocolateArmpits: Thanks a lot. My aim is to merge GIF files to a mp4. I want each gif to play for 5 seconds (even if they are shorter) looping and then the next one.
[19:14:44 CEST] <ChocolateArmpits> Emmett: Those "small (5k) videos" are created once per every voicemail call ?
[19:15:10 CEST] <Emmett> When the user connects to view the call&yes
[19:16:02 CEST] <Emmett> Like I said&I am not a ffmpeg guy&so I don't "grok" the output well to know if there is a clear indicator of the problem or not.  Can I post a few snippets from the logs to see if anything stands out to you guys?
[19:16:28 CEST] <ChocolateArmpits> Well that won't hurt
[19:17:36 CEST] <Emmett> Just don't want to offend anyone. :-)
[19:17:39 CEST] <Emmett> [Thu Oct 08 05:53:46 2015] [error] [client] Input #0, avi, from 'pipe:':
[19:17:39 CEST] <Emmett> [Thu Oct 08 05:53:46 2015] [error] [client]   Metadata:
[19:17:39 CEST] <Emmett> [Thu Oct 08 05:53:46 2015] [error] [client]     encoder         : MEncoder SVN-r36450-snapshot-4.4.6
[19:17:39 CEST] <Emmett> [Thu Oct 08 05:53:46 2015] [error] [client]   Duration: N/A, start: 0.000000, bitrate: N/A
[19:17:39 CEST] <Emmett> [Thu Oct 08 05:53:46 2015] [error] [client]     Stream #0:0: Video: flv1 (FLV1 / 0x31564C46), yuv420p, 352x288, 29.92 tbr, 29.92 tbn, 29.92 tbc
[19:17:50 CEST] <ChocolateArmpits> Emmett: wait there, use pastebin
[19:18:02 CEST] <ChocolateArmpits> And paste everything relevant to ffmpeg there
[19:18:22 CEST] <ChocolateArmpits> Unless this is all your log output there is
[19:18:23 CEST] <Emmett> Ok&sorry
[19:18:39 CEST] <Emmett> Just a little bit more&but let me do it the right way
[19:28:34 CEST] <Emmett> Ugh&my laptop hung&let me try it now.
[19:31:19 CEST] <Emmett> http://pastebin.com/4VzKfB98
[19:32:25 CEST] <Emmett> Crap&I supposedly cleaned that up some. Sorry.
[19:38:32 CEST] <DHE> the input is a pipe, making me suspect ffmpeg is being invoked directly from the CGI that's launched from the web server. My concern there is some kind of attack could end up spawning a lot of ffmpeg instances and then you just fall over from that alone.
[19:39:51 CEST] <Emmett> I've been watching the connections very closely&we definitely aren't getting DOS attacked.
[19:43:19 CEST] <ChocolateArmpits> If I guess right the output videos are unplayable or are they black ?
[19:46:34 CEST] <Emmett> Probably&the problem is that they do not stick around for long once streamed back to the customer.  Plus there is a bit of a privacy issue. :-)
[19:50:13 CEST] <ChocolateArmpits> the seemingly huge overhead is due to the not much stream content getting written
[19:50:43 CEST] <ChocolateArmpits> It's also receiving pipe input from Memcoder
[19:51:04 CEST] <Emmett> What might cause that?  We don't see any i/o issues.
[19:52:15 CEST] <ChocolateArmpits> Well, I'm not familiar with H263 to know if it can encode a suposedly black video at that little space
[19:53:10 CEST] <ChocolateArmpits> I could try setting up similar settings and encode a blank signal myself to see if it's any close
[19:54:03 CEST] <Emmett> That would be awesome!
[19:54:39 CEST] <ChocolateArmpits> Well it can only act as a clue, but not a definit answer. Best would be if you could get a hold of those generated videos
[19:59:31 CEST] <ChocolateArmpits> Emmett: Is there anything specified in the command line call for ffmpeg similar to " -g " command ?
[20:00:43 CEST] <oreWizard> hi
[20:00:46 CEST] <oreWizard> i am getting error
[20:00:50 CEST] <oreWizard> on using this transcode_aac_8c-example
[20:00:56 CEST] <oreWizard> can someone help mew
[20:01:01 CEST] <Emmett> ffmpeg -i - -s cif -passlogfile ffmpeg_p1_1444413625 -pass 1 -an -b:v 384k -vprofile baseline -r 30 -strict -2 filename
[20:02:40 CEST] <oreWizard> someone knows how to write rgb frame with transcode_aac_8c-example
[20:05:31 CEST] <ChocolateArmpits> Emmett: Well that's strange, "cif" implies a resolution of 352x288, while in the log the output res is 176x144, the rate is set at 30 while the log has the output at 15. These settings match the input, but certainly no the output that the command line should be describing
[20:05:59 CEST] <ChocolateArmpits> Is the log output of this command ?
[20:07:24 CEST] <ChocolateArmpits> You can get a report of a particular ffmpeg process using "-report" parameter
[20:08:43 CEST] <ChocolateArmpits> It's a single switch command that outputs a log file to the current directory
[20:09:16 CEST] <ChocolateArmpits> Place it anywhere between process call and output, just not directly between a param and its argument
[20:10:22 CEST] <Emmett> My problem is that this is a production system and I cannot change it without about 3 days worth of approvals.  And that is the only environment that is exhibiting this problem.
[20:16:05 CEST] <ChocolateArmpits> Emmett: Until you can find a way to do testing with those exact binaries and the scripts that use them, my best guess the problem is with what's being fed via pipe to ffmpeg
[20:16:43 CEST] <ChocolateArmpits> Your log is also full of "[error]" flags that happen all through it and before ffmpeg starts any encoding
[20:17:47 CEST] <Emmett> Are those really "errors"?  They looked like version information and other debug data.
[20:19:24 CEST] <ChocolateArmpits> Emmett: Well I've never seen those included with ffmpeg logs, so I can only guess it's part of your system
[20:20:39 CEST] <Emmett> Ok&thanks&I'll keep trying to catch one of those videos on the system.
[20:21:13 CEST] <ChocolateArmpits> unless the source code to ffmpeg was modified to include those messages in the console before getting compiled
[20:21:43 CEST] <ChocolateArmpits> I can't judge, the log is strictly ffmpeg's
[20:21:49 CEST] <oreWizard> somone know how to use this transcode_aac_8c-example?
[20:21:54 CEST] <oreWizard> https://www.ffmpeg.org/doxygen/trunk/transcode_aac_8c-example.html#a90
[20:27:36 CEST] <alesan> so I am checking a performance issue in my script, where I use imagemagick to ouput yuv data to ffmpeg for encoding. it works - but if I use yuv from convert->ffmpeg it takes MUCH more time than if I use jpeg encoding
[20:28:02 CEST] <alesan> I was expecting something like a (small?) reduction in time needed for yuv processing, not a huge increase
[20:28:06 CEST] <Emmett> Thanks for your help CA.
[20:28:47 CEST] <alesan> it might have to do with the yuv format, like both convert and ffmpeg have to re-order the YUV for proper processing, while jpeg decoding already has it "right"
[20:28:52 CEST] <ChocolateArmpits> Emmett: np
[20:29:08 CEST] <alesan> so, when encoding with libx264, what is the best yuv format to be used
[20:29:36 CEST] <ChocolateArmpits> alesan: best depends on what you need it for
[20:29:42 CEST] <alesan> as source so there is no re-order needed? assuming my theory makes sense, but I have some evidence it might be correct
[20:30:24 CEST] <alesan> ChocolateArmpits, well I am sure libx264 wants some 4:2:0 data in a given format to proceed with encoding, right?
[20:30:51 CEST] <alesan> maybe it wants the planes non interleaved, or maybe it works best with luma and then interleaved chroma
[20:31:02 CEST] <alesan> I know some hardware encoders want this latter option
[20:31:19 CEST] <iive> alesan: try `perf top`
[20:31:42 CEST] <alesan> iive, oh OK what is that? please give me more context
[20:32:30 CEST] <alesan> won't I need imagemagick and ffmpeg compiled with debug symbols for such tool to be effective?
[20:32:34 CEST] <iive> perf is tool that comes with linux kernel. It interrupts cpu execution at fixed intervals and checks the execution point, creating statistics.
[20:32:55 CEST] <iive> if one function is taking a lot of cpu, it would be probed frequently.
[20:33:12 CEST] <alesan> installing...
[20:33:23 CEST] <iive> `perf top` lists the functions in format like the tool `top` aka one line per function.
[20:33:32 CEST] <iive> so... less guessing.
[20:33:42 CEST] <alesan> oh OK I need to... reboot
[20:33:52 CEST] <iive> why?
[20:34:00 CEST] <alesan> I had a kernel update recently and I have not rebooted yet and this versoin of perftop requires a specific kernel
[20:34:02 CEST] <alesan> or vice-versa
[20:34:11 CEST] <iive> aha... ok.
[20:34:12 CEST] <__julian> how does mpeg2video encoder parallelize? intra-frame based or actually encoding multiple frames in parallel?
[20:34:32 CEST] <alesan> BRB
[20:37:36 CEST] <alesan> re
[20:38:25 CEST] <alesan> OK now perf "works"
[20:53:16 CEST] <alesan> iive, it seems the vast majority of time is spent in __memcpy_sse2_unaligned _IO_putc and __ieee754_pow_sse2 __exp1
[20:53:52 CEST] <iive> IO_putc, are you using pipe?
[20:54:29 CEST] <alesan> yes
[20:54:36 CEST] <alesan> convert pipes to ffmpeg
[20:54:40 CEST] <iive> i'm not sure where the pow function comes from, that's power
[20:54:59 CEST] <iive> aka x^p, maybe convert uses gamma correction?
[20:55:26 CEST] <alesan> by the way, the perf top results are somewhat similar if I use jpg or yuv as exchange format
[20:55:37 CEST] <alesan> well... isn't that unlikely?
[20:55:49 CEST] <alesan> I mean... well maybe, what can I say, but I do not expect it to
[20:56:19 CEST] <iive> yeh, especially with the difference you notice.
[20:56:48 CEST] <iive> btw, why not do color conversion inside ffmpeg?
[20:57:36 CEST] <alesan> why do I have to perform color conversion at all
[20:57:54 CEST] <alesan> I start from a 4:2:2 jpeg and encode to 4:2:0 H.264
[20:58:06 CEST] <alesan> I should not need any color conversion
[20:58:20 CEST] <alesan> unless you refer to the jpeg and H.264 differences in colorspace coefficient
[20:58:44 CEST] <alesan> I always forget the name of those... the "analog TV" one used in video formats and the "full range" one used in photo?
[20:59:11 CEST] <ChocolateArmpits> the unorthodox ones are "tv levels" and "pc levels"
[21:02:16 CEST] <alesan> ye
[21:04:07 CEST] <Dexterrp> Hey there, anyone there managed to get quicksync going in FFMPEG?
[22:59:31 CEST] <thejamespinto> hey guys, sorry for the dumb question but is there a way I can set <Stream STREAM_NAME> to some kind of "index page" ?
[23:04:48 CEST] <thejamespinto> oh cool, I can literally call it index.html thx
[23:17:01 CEST] <wallbroken> hi
[23:17:11 CEST] <wallbroken> how to split audio track with ffmpeg?
[23:21:02 CEST] <llogan> wallbroken: ffmpeg -i input -map 0:a -c copy -ss 30 -t 20 output
[23:21:31 CEST] <wallbroken> llogan, what are 30 and 20 ?
[23:21:55 CEST] <llogan> number of seconds
[23:22:13 CEST] <wallbroken> no wait
[23:22:21 CEST] <wallbroken> maybe you mistaken what i need to do
[23:22:29 CEST] <llogan> you did not explain much so i made a guess
[23:22:33 CEST] <wallbroken> i have muxed audio and video file
[23:22:36 CEST] <wallbroken> yes
[23:22:39 CEST] <wallbroken> so
[23:22:54 CEST] <wallbroken> during the playback, audio becomes slower than video
[23:23:17 CEST] <wallbroken> and at some point i see audio completely desync
[23:23:30 CEST] <wallbroken> somebody else muxed it and works perfectly
[23:23:43 CEST] <wallbroken> i checked with mediainfo the info of audio
[23:25:08 CEST] <wallbroken> http://pastebin.com/VAuajbwK
[23:25:10 CEST] <wallbroken> this is the right
[23:25:35 CEST] <wallbroken> http://pastebin.com/wbwRKVqf
[23:25:43 CEST] <wallbroken> this is the broken
[23:25:49 CEST] <wallbroken> (created by mine)
[23:29:26 CEST] <wallbroken> llogan *
[23:35:18 CEST] <llogan> that's from mediainfo. i thought you were using ffmpeg
[23:39:42 CEST] <wallbroken> yes i was trying to get info
[23:39:57 CEST] <wallbroken> on how audio was muxed
[23:40:16 CEST] <wallbroken> ffmpeg.exe -i erfge.avi -i "AudioMUX ITA 2x01.mp3" -c copy -shortest out.avi
[23:40:23 CEST] <wallbroken> this is what i used to mux
[23:44:36 CEST] <llogan> you never provided the console output, but now i have to go
[23:45:02 CEST] <thebombzen> does anyone here know how to add motion blur with an FFmpeg filter? That is, basically slur frames into each other?
[23:45:41 CEST] <thebombzen> (I have a fan blowing highspeed in a 60 fps video. when I drop the rate to 10 the fan slows down. I'm trying to make that not happen.)
[00:00:00 CEST] --- Sat Oct 10 2015

More information about the Ffmpeg-devel-irc mailing list