<<This is how I understand it. But I think the writer's objection is
that this process doesn't track changing -noise- levels. For example,
a passing airplane will be above the thresholds set for the ambient
noise. The fact is, these algorithms have no intelligence, they don't
distinguish the sound you want from the sound you don't want except by
level-in-frequency-band. I think Cedar has some automatically adapting
algorithms, but I doubt they'd be effective on nature sounds.>>
Right, the adaptive-ness of the algorithm adapts to changing desired
signal levels, not changing levels of residual noise. However, BIAS
claims the new version of SoundSoap Pro, version 2, does in fact adapt
to changing noise levels. We'll have to see about that. There are
certainly ways around the former method, though. I've occasionally
done several passes through SoundSoap Pro with differing thresholds,
then crossfaded amongst the resulting files. Bit of a PITA but
workable if not too many different noise spectra are presented. For
the example of a passing airplane I tend to use the Spectral Repair
mode of iZotope RX, isolating the offending frequency spectra. Lots of
manual work & not really different than rolling in some tight filter
bands, although you can get extremely frequency selective with RX,
without the ringing of a high Q filter. With a nature recording I
think one could fairly easily do a pass through SoundSoap Pro tuned
for the aircraft noise, then simply cut that file into place in the
original, although I'd probably just cut the entire offending section
of time & discard it.
Scott Fraser
|