Noise reduction effect

Question

I am trying to reduce the background noises in a voice recording from  video shot at a Gas Station with  loud vehicle noise in Audition. I do not want to eliminate the sounds altogether but lower them quite a bit. I watched the Tutorial on the website by Mr. Levine but with the variance and number of controls I am finding it difficult to get the same results as he did in his example. I have multiple clips of this kind and would like to streamline the process.
Is there a sweet spot for the Spectral Decay Rate, Precision Factor, Smoothing and Transition Width in the advanced settings in the Noise Reduction Effect to reduce loud background noise from  a voice recording without producing the tunnel like echo in the voice after the effect is applied?

Answer 1

Werlien wrote:
Is there a sweet spot for the Spectral Decay Rate, Precision Factor, Smoothing and Transition Width in the advanced settings in the Noise Reduction Effect to reduce loud background noise from  a voice recording without producing the tunnel like echo in the voice after the effect is applied?
Firstly, no there's no sweet spot. All noise is different, and comes in different quantities. Secondly, variable noise, of the kind you get on a garage forecourt, is very difficult to treat anyway. The basic NR in Audition is designed to deal most effectively with fixed noise sources and not variable ones. That said, if you have a fixed level of background noise as well as some variable noise, you can generally reduce the fixed noise to a degree, even if it doesn't touch the variable stuff.
In general, for variable noise reduction, you need to use the Adaptive settings, but you have to be careful with these, and the results are never going to be that wonderful, quite frankly. You may need to play with the settings quite a bit to get any sort of an acceptable result at all.
If you use the basic NR system, the best way to use it is to do more than one pass, at different FFT settings - at least one low setting, and at least one at the highest setting there is. And at each pass, don't try to take off too much. The results you'll get like this will be much better than any attempt to optimise a single pass will ever be. Please note that for each different FFT setting, you will need a new noise sample. By and large, leave the rest of the controls at their defaults, although you may wish to reduce the spectral decay rate back to zero for multiple passes - it can make the transitions sound even weirder when you go from wanted signals to the background sound. By using multiple passes with less reduction on each one, you will pretty much eliminate the need tor spectral decay.
I suppose that the ultimate bad news about NR is that unless you're incredibly lucky, it never works particularly well until you've made quite a bit of effort with it. It's not that unusual to go back to the start of a NR session and start over, using completely different settings before you get anywhere at all.
And yes, Jason chose his noisy material carefully....!