[FFmpeg-devel] One pass volume normalization (ebur128)

Reimar Döffinger Reimar.Doeffinger at gmx.de
Tue Jul 16 21:30:04 CEST 2013


On Tue, Jul 16, 2013 at 08:16:39PM +0200, Jan Ehrhardt wrote:
> Nicolas George in gmane.comp.video.ffmpeg.devel (Tue, 16 Jul 2013
> 17:28:55 +0200):
> >L'octidi 28 messidor, an CCXXI, Jan Ehrhardt a écrit :
> >> It might be a crappy OS (Windows), which cannot be configured better.
> >> But it perfectly explains the speed decrease (from 17 to 27 seconds)
> >> when the only thing you do is turn FFMpeg's logging on.
> >
> >No, it does not, since the amount of ffmpeg's log messages is negligible.
> >Please see the rest of my previous message for things that matter.
> 
> Two command lines:
> 
> ffmpeg.exe \
>     -i U:\MP_ROOT\100PNV01\SD00004.MP4 \
>     -i U:\MP_ROOT\100PNV01\SD00005.MP4 \
>     -filter_complex " \
>     [0:a]asetpts=PTS-STARTPTS[a0]; \
>     [1:a]asetpts=PTS-STARTPTS[a1]; \
>     [a0][a1]concat=n=2:v=0:a=1[a]; \
>     [a]volumedetect[am]" -map [am] \
>     -f null -y /dev/null
> 
> Execution time: 16.765 seconds.
> 
> ffmpeg.exe \
>     -i U:\MP_ROOT\100PNV01\SD00004.MP4 \
>     -i U:\MP_ROOT\100PNV01\SD00005.MP4 \
>     -filter_complex " \
>     [0:a]asetpts=PTS-STARTPTS[a0]; \
>     [1:a]asetpts=PTS-STARTPTS[a1]; \
>     [a0][a1]concat=n=2:v=0:a=1[a]; \
>     [a]volumedetect[am]" -map [am] \
>     -report -f null -y /dev/null
> 
> Execution time: 27.154 seconds
> 
> Please tell me where I go wrong. And/or guide me to a quicker
> volumedetect an a concatenated audio track.

You probably have write caching or something similar off
(on Windows, it is off by default for anything it believes
you might be able to detach).
If you use log/temporary files you really should be
using a disk suitable for it.
Since pipes are unfortunately a rather unusable pain,
a RAM disk would be the best performing alternative.


More information about the ffmpeg-devel mailing list