I've had this app for a month now but just haven't had the time to really fool around with it till today. Well, I think a spent about an hour with it last week but that was more of just reading the manual. I do audience recording of jazz for the most part so chatter is a real problem. To minimize this, I use the card-cap on my AT 853's. I record to a Hi Minidisc so the recordings are uncompressed 16/44. An annual festival is one of the main venues I record and in the main arena, the seats are assigned so moving around isn't an option.
Enter iZotope RX. Billed as restoration application, it has several different modules, dehummer, de-clicker, de-noiser , de-clipper and spectral repair. Each module has plenty of adjustments and sub-modules to fit the offending sound/s. There are plenty of ways to eliminate steady state problems like compressors (from ice cream trucks or air brush artists enhancing the features of virginal ,.. well, you get the idea) or unruly ground loops. The module that makes this a tapers dream come true is the Spectral Repair module. There are 4 sub modules, attenuate, replace, pattern and damned if I can remember the name of the last one. The process goes as follows. Just like any other audio program, your looking at the waveform of your project. Only, with RX, the waveform is sitting on top of a spectrogram. For the most part, I dimmed the wave view so all I could see was the spectrogram. With a wave, your looking at the peaks of the sound. With a spectrogram, your seeing all the elements that are within the waveform. So first you have to find the problem. Let's say it is a cough from the person sitting in front of me. I would make as many passes as it takes to find the exact spot that the cough begins zooming in as I go. Then you look for an event that's different from the surround beats. The louder the sound, the lighter the event appears within the spectrogram. Once you've found a likely candidate, you can either select the entire area (all frequencies) or just drag a box around that light spot. Now comes the fun stuff. You punch up spectral repair and need to choose which of the sub modules your going to try. The Pattern setting is what I'll explain briefly and will give you an idea of the power of this program. With Pattern set you hit apply and depending on the parameters you set, RX looks on either side for a pattern similar to the one surrounding your selection. This happens in about 1/4 of a second and the light spot is replaced by what RX thinks should be there. It adjusts the surroundings so there's no abrupt change. The Replace setting just replaces the selected spot. I spent about 3 hours with it and about half of what I learned was to recognize the noise from the music. The other half was trying to figure out which technique worked best with each noise. In some cases, selecting the entire area (all frequencies) was best. With other ones, selecting just the loudest part of the noise (box) worked better. And sometimes, it was best to remove the loudest part and then remove the harmonics of the noise later.
The project I was practicing on had a range of problems. From faint chatter a few seats from me, to "Me" coughing once, and saying goodbye to someone I wouldn't see for another year. Whatever the origin, I was able to seamlessly remove all noises that I wanted to. Even when I was the offending source, with the mikes attached to my glasses just inches from my drooling gob. Whatever I wanted to remove, I could. No artifacts! I might be able to find those spots if I looked for them but not all of them. I was at the point where I had to decide to leave some low level stuff to make it sound live!
By the time I finished for the day ( I had some very important NFL games to watch) I knew that I no longer felt a need to use card.s anymore. So, DPA or Sanken? RX has freed me from worrying about audience noise.
http://www.izotope.com/products/audio/rx/