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

[Paul B Mahol]

On 7/17/13, Jan Ehrhardt <phpdev at ehrhardt.nl> wrote:
> Jan Ehrhardt in gmane.comp.video.ffmpeg.devel (Mon, 15 Jul 2013
> 06:47:48 +0200):
>>Would it be possible to insert the momentary value for volumedetect in
>>the metadata and use that as input in af_volume.c? One pass
>>normalization based on volumedetect should be faster than what we have
>>now.
>
> To answer my own question: Yes, that is possible. Since Nicolas
> problably does not want to tamper with volumedetect, I created a clone
> called af_volumeinject.c. See below.
>
> Just like I now use volume=metadata=lavfi.r128.I this volumeinject
> allows me to use volume=metadata=lavfi.mean_volume. It is only a tiny
> bit slower than something like volume=-3dB, but faster than a two
> pass volumedetect + volume=-xx.xdB.
>
> One thing came while testing volumeinject. I am really a novice in
> loudness and would have expected that volumedetect and ebur128
> would report more or less the same value. They do not. They were
> 3-4 dB apart in my testfiles, with ebur128 reporting a louder value.

Because they are doing different things ...