Hmm, this sort of brings up an interesting question: Does anyone
make a device that does all this via software? If it happens
you're going to record everything on a computer anyway, you
could do it all through DSP, and presumably you could make
a system that autodetects everything it needs to autodetect.

How would the software know who was talking? That isn't obvious. You might just
as well use a compressor, which would probably use audible side-effects.