Over the years, I have noticed my ears have become very (overly?) sensitive to these narrow frequency peaks that can occur in a lot of situations - for example with ambient pads that have some serious frequency processing, or when loud vocals are compressed & processed heavily. Often single tracks suffer from these peaks, or it can happen when two or more more tracks are summed.
Now, I have noticed that every time my ear catches these annoying peaks, it's very easy to locate them visually with a frequency analyzer (the peak or peaks clearly stand higher than the surrounding frequencies), and then I can just use a narrow-band EQ to tame them. But the problems start when the vocalist or instrument moves to another note (as they often do, those bastards), which means the problem frequencies shift to another place.
I have sometimes used a dynamic EQ and pre-programmed the most problematic frequencies, but it seems that no matter what I try, it never really does what I want it to do (it cuts some even when it's not supposed to, and does not cut enough when my ears hear a very apparent spike). Not to mention that depending on the audio content, there can be a dozen different problem frequencies at different times. In the end I always have to use static EQs and a lot of automation, which is extremely frustrating and time-consuming.
The problematic peaks I'm talking about are very narrow (often single harmonics), and sometimes very close to each other, and require surgical precision. For example, on a vocal track, there might be problem peak at 350Hz, then a second later, at 380Hz. To cut those properly, the cut needs to accurately follow the peak, one fixed setting doesn't quite do it. The peaks usually occur somewhere between 300-1000Hz, and they are often one of the first harmonics above the fundamental (at least in vocals).
So. What I'm after is an effect that intelligently monitors the whole frequency range, and notices any narrow frequencies that are about to rise too high compared to the rest of the audio (usually there are only 1-2 of those at a time), and makes them softer. It's also important that the imaginary effect would always compare the level of the peak to the overall audio level, instead of just monitoring if it exceeds some fixed threshold.
My limited common sense tells me that this shouldn't be impossible, since I can easily spot those frequencies with both my ears and eyes, although it probably requires some processing power. I know there are "feedback eliminator" devices which probably work similarly to what I'm after? Note that I'm not after a traditional multiband compressor - unless it has at least 1000 bands and an intelligent dynamic threshold for each band.
Also tried some Googling, but I really don't know what to look for... if such an effect exists, I have no idea if the effect type has an established name (like "compressor" or "reverb"). It might be possible to build something like this in Reaktor, I already tried some simple tests, but if the effect already exists, I better not waste my time on something that probably doesn't work anyway...I can also post some audio examples if you don't quite know what I'm talking about. Any ideas are welcome, thanks!
