Sound Finder / Silence Finder improvements

[steve]

@Gale Andrews

If this was a “classical concert” then the “short loud noises to be ignored” setting would be very useful, but so would the “allow silences between sounds” setting which I would set quite long. It would let me digest the last movement while listening to the low level audience noise (the “atmosphere”) before the next movement, without being lifted out of my seat by someone moving the piano or chair along next to the mic. Pauses between movements can be important for this case, witness often heard complaints if a conductor runs “attacca” into the next movement when this is not in the score.

We generally recommend that people do snap crackle and pop removal before running sound/silence finder. For the same reason it would be better to deal with issues such as chair clatter (with, for example, “Pop Mute”) before running sound/silence finder.

Having gone through dozens of tests, I think that the only practical way to implement “ignoring short sounds” (without producing bad side effects) is to have:
“Ignore Isolated Short Sounds Less Than: … (seconds)”

An isolated short sound being a sound, shorter than the specified length, that is between two recognised silences.
A recognised silence is audio below the silence threshold that is longer duration than “allowed silence”.

With every other algorithm that I can think of, there are real world situations that cause the algorithm to do “bad things”.
With this algorithm, there are some situations that cannot be dealt with as one might like, but I don’t think it does anything “bad”.

@steve

Silence detection based on peak level is a lot easier to understand and the noise floor can be easily measured using the Amplify effect. I think this should be the default.

RMS or weighted RMS should be much better for vinyl transfers but requires that the silence level is set substantially lower to avoid marking non-silence as silence. If this option has as much benefit as I expect then I think it would be a shame to not include it, but judging the correct threshold level may be a bit difficult for novices.

We’ve already got stats.ny that can analyze the rms (and A-weighted) level, so like I’ve included a tool in the noise gate plug-in for measuring the “silence” level, it would be easy enough to have a “measure noise level” option in Sound Finder.