[FFmpeg-user] Concat limited to certain file types?

[Jan Ehrhardt]

Nicolas George in gmane.comp.video.ffmpeg.user (Wed, 17 Oct 2012
17:23:06 +0200):
>Le sextidi 26 vendémiaire, an CCXXI, Jan Ehrhardt a écrit :
>> This one comes quite close, I think:
>> 
>> -filter_complex " \
>> 	[0:v]  setpts=PTS-STARTPTS [v0]; \
>> 	[0:a] asetpts=PTS-STARTPTS [a0]; \
>> 	[1:v]  setpts=PTS-STARTPTS [v1]; \
>> 	[1:a] asetpts=PTS-STARTPTS [a1]; \
>> 	[v0] [a0] [v1] [a1] concat=n=2:v=1:a=1 [v] [a]; \
>> 	[v] scale=384:288 [vs]" -map [vs] -map [a]
>
>It looks quite right, assuming all files start with audio and video in sync.

I hope they are: it are files produced by all kinds of camcorders. If
they are not JVC, Sony, Canon, Panasonic or Zoom has a real problem ;-)

I tested it on a testsuite with 26 input files: 25 times a short one
with different lengths of the video and audio and then a longer one.
Video and audio stay in sync.

BTW: the commandline was over 4000 characters long...

I have to test it yet on Canon's 50fps interlaced MTS files, which have
always been the most troublesome files.

>I do not know the status of volnorm, I do not like it. You can do a two pass
>normalization, first detecting the volume and then adjusting, using the
>volumedetect or ebur128 filters (the first one is much faster, the second
>one is way more accurate).

I am doing something like that right now: the second pass is a MEncoder
pass with -af volnorm, which goes through the encoding with 1000fps,
which leads to 2 minutes extra transcoding time for our typical files
(1-2 hours).

Our videos are recordings of therapist-client meetings. Video is not
really important, audio is voice only but sometimes of low volume.
MEncoder's volnorm filter has been adequate for increasing the volume
over the last 8000 videos.

I will look into volumedetect.

Thanks for the help.

Jan