Izotope RX3 denoiser instructions/ tutorials for music NR?

[Audiogeek]

I can't seem to find much of anything on music noise reduction tutorials
using RX3. There's some on NR of dialogue but not so much for music.
I'd like to find good tutorials on both manual and adaptive NR.
Suggestions?

[FL]

In my experience, the techniques used for music differ substantially from those required for dialog. Often, settings which result in excessive pumping when applied to dialog are perfectly acceptable for music, whereas the various processes specifically designed for dialog (particularly the dialog noise reduction and de-reverberation) are all but useless for music.

In both cases, however, the bulk of your time will be spent experimenting with various settings, and varying the order in which you apply the different processes. It takes much more time determining what will work, than the time spent actually applying the "winners". If you are doing this work for hire, it can be difficult to determine the sweet spot between being paid for all the work put in vs. the client's valuation of the result.

For music, if you're looking for a "standard technique", I'm afraid I haven't found one - every project is different. I will generally do the long term corrections first: hum/buzz/whine removal, de-crackling (if I'm doing transcriptions from records) and broadband noise reduction (air handling, electronic hiss, etc.). Then I will go after individual short term noises with the Spectral Repair. That said, I have found some projects where it was best to do the long-term stuff last. Nothing for it but trial and error. A very useful tool is the capability in several modules that will play a preview of what's being removed (as opposed to the cleaned material) - if you listen to that and hear too much of the musical performance, then you know you should back off. Some projects benefit from doing several passes of gentle reduction; others turn out better if you do it all in one go.

Personally, I have always found it much more productive to use the stand-alone program than to use the various plug-ins within an active DAW project - particularly for problems where you have to identify specific regions of sound to either be removed, or to be used to "train" the de-noiser. In a DAW, you are often limited to using only those processes which can be accomplished in real time, whereas results with much fewer processing artifacts are only attained when you allow the advanced algorithms time to do their stuff. Even if you do the "slower-than-realtime" operations as a Render in your DAW, you do not have the flexibility available when running RX as stand-alone. Yes, this means you can only process one or two tracks at a time.

Whatever method you wind up using, one fact is constant: for every layer of the onion you peel off, another is revealed (with the accompanying tears). The trick is knowing when to stop. It's very easy to go too far, resulting in something which may be quiet, but has had all the subtlety sucked out of it. Luckily Izotope has a great "Undo" set-up, and if you quit the program with a file still loaded and running, the next time you start up, it's all still there, including all the Undo levels. This alone is one of the most brilliant design features of the program.

I keep a quote from Bob Olhsson (made right here in rec.audio.pro many years ago) framed on my wall: "Signal processing is always a trade-off between degradation and a subjective improvement."