The one thing you don't want is a portion of audio!
What you need to select is a portion of silence where there isn't any wanted audio - in other words, it's the noise before or after speaking that you need to select. You make the selection in the normal way you make one; position the cursor at the start of your background noise, hold the left mouse button down and swipe the mouse forward to the end of the noise selection. When you let go of the mouse, the highlighted selection is what you use as the noise print.
The correct way to use NR isn't to make the noise selection first, incidentally. The best method is to open the plugin first, and it will say 'noise print undefined'. Don't worry about that. Go then to Advanced, and select the FFT size. Don't leave it at the default, but increase it to the maximum value. Now, whilst keeping the plugin open, go back to your waveform and make the selection, as above. When you subsequently click on 'Capture Noise Print' that selection is what is captured, and that FFT size is the one it uses.
The next important bit is how much to use. To minimise artifacts, take off very little noise at a single pass. So at a high FFT setting, don't take off more than about 3-4 dB. This means setting the Noise Reduction value so that the green band is slightly under the yellow band, (usually around 70-80%) and the Reduce by slider set to 3-4dB. Apply that. At that point, you need to repeat the process, only with the FFT set to a mid-range value. Once again, only take 3-4dB off to avoid artifacts. And then repeat the process once more, using the lowest value of FFT, still taking only 3-4dB.
Because you have to resample each time, these steps are cumulative - you are taking a fresh noise print with what's actually there, rather than what you started with, and you will end up with a decent amount of artifact-free NR - as long as you don't overdo it. That's the key to the whole process.
3
Replies
AdChoices