I sent this thread to my friend Tony Lauck, who has asked me to post the
following. (Tony is a highly experienced scientist and engineer in the
computer field, an audio recordist, and a classical music lover.)
On the Izotope RX Advanced:
I have been using RX Advanced for a few weeks and it has done a great job.
In particular, it is possible to get rid of lots of background noise
automatically without introducing "musical artifacts". On the best setting
it is a little slow, in that it can take several times slower than real time
to do the noise reduction with a 2.9 GHZ PIV processor. However, usually
you can take one 20 second section to explore the parameter space, so each
iteration only takes a minute or so. Then when one has it right, turn RX
loose on an entire track and go for coffee if necessary.
I found in many cases that the RX noise reduction took care of hum,
generator noise, monitor whine, etc. all automatically. In some really bad
cases, slight improvements were had by using the hum reduction module.
There are a variety of ways to patch individual defects and these are well
integrated with the spectrographic display. It is very easy to spot
individual defects like mike pops by a speaker or singer, audience coughs,
birds singing outside an open window, etc., not to mention digital ticks,
vinyl noise. One needs a lot of discipline and a strategy to determine which
"defects" are part of the performance and which are unwanted artifacts of
On one voice recording there was a loud truck for 20 seconds that remained
after the general noise reduction. A second pass with a noise print on the
truck removed almost all of the truck noise, but left a complete absence of
ambiance. This was easily cured by mixing in room noise from another
section of the recording. The result was described as almost magical by my
friend who will be using this sound track in a video. Of course speech is
easier than music, because of the frequent pauses.
I have yet to do the due diligence as to high end sound quality when a file
is passed through RX. I expect some degradation on portions of the file
that are being repaired or cleaned up. But it would be nice to know that
other parts of a file are not subtly degraded just by being passed over by
the software. So I am going to do some tests. I expect some loss of sound
quality, consistent with introduction of dither noise, but since RX operates
at 32 bits floating point or 24 bits fixed point I expect that these tests
will find a minimal impact.
I am still learning about tradeoffs with various parameters, such as how to
set up equalizers, when to use minimum phase or linear phase, etc., but I
can't go too far on these issues until I get better monitoring capability
and improve the acoustics of my workspace.
The $1200 I spent on this software appeared to be excessive at the time, but
in retrospect I consider it a bargain.