[FFmpeg-user] volumedetect audio filter result

[Nicolas George]

L'octidi 18 nivôse, an CCXXI, jacky a écrit :
> video:0kB audio:2158kB subtitle:0 global headers:0kB muxing overhead
> -100.000000%
> [Parsed_volumedetect_1 @ 023df480] n_samples: 1104896
> [Parsed_volumedetect_1 @ 023df480] mean_volume: -30.4 dB
> [Parsed_volumedetect_1 @ 023df480] max_volume: -10.0 dB
> [Parsed_volumedetect_1 @ 023df480] histogram_9db: 1
> [Parsed_volumedetect_1 @ 023df480] histogram_10db: 9
> [Parsed_volumedetect_1 @ 023df480] histogram_11db: 19
> [Parsed_volumedetect_1 @ 023df480] histogram_12db: 85
> [Parsed_volumedetect_1 @ 023df480] histogram_13db: 175
> [Parsed_volumedetect_1 @ 023df480] histogram_14db: 436
> [Parsed_volumedetect_1 @ 023df480] histogram_15db: 810
> 
> questions
> 1- is there a way to find out where these peaks are located (time) ?

Not currently.

> 2- there are different volume references between lame (replygain
> value) and volumedetect what is the reference level  24dB or 27dB other
> one value ?

I am not sure what your question is. volumedetect is very stupid: the
reference value is 0 dB when all samples are constant at 1, which does not
make sense since a constant level is silent; the loudest audible sound would
be a sine with amplitude 1, it is measured as -3.01 dB.

> 3- is this reference level fonction of the -tune option fixed during
> encode ?

The only -tune option I know is for x264.

If you are looking for a better volume measurement filter, look at ebur128:
it is much more accurate than volumedetect (it takes the frequency into
account) and gives more detailed output. volumedetect is just much faster.

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-user/attachments/20130116/7f9e1087/attachment.asc>