[FFmpeg-user] audio normalization - nth attempt

[Paul B Mahol]

On Tue, Apr 13, 2021 at 7:42 AM Emanuel Berg <moasenwood at zoho.eu> wrote:

> Paul B Mahol wrote:
>
> >>> One problem with just af=dynaudnorm is that when you
> >>> backtrack to play a sequence again the volume drops so
> >>> much you don't hear anything, almost, until it picks up
> >>> much later.
> >>
> >> This helps for that
> >>
> >>   af=dynaudnorm=g=7
> >
> > That enables very agressive compression.
>
> Please visit Uppsala, Sweden, we'll watch 4K Ultra HD
> "Mission Impossible 6: Fallout" together with and without it.
> With it, works great. Without it, you barely hear what the
> say (you don't hear what they say) yet 5s later, firefight
> BOOM BOOM BOOM with dramatic music and the building wakes up.
>
> > Have you ever read documentation of dynaudnorm filter, the
> > ffmpeg one or the original one?
>
> I don't know, I have read ffmpeg-filters(1) lines 2885-2899
> and 2918-2935.
>

Try alt boundary option, b=1, that will prevent fades after seek/start.

>
> --
> underground experts united
> https://dataswamp.org/~incal
>