Ideally, audio that you want to perform noise reduction on will have a noise-only portion at the beginning or end of the recording that DINR can analyze and learn. Unfortunately this is not always the case, and in many recordings some amount of signal is always mixed with the noise. Obviously, analyzing such audio will produce a noise signature that is based partially on signal. Luckily, DINR has provisions for cases such as this, and this is where the
Auto Fit feature comes in.
If your audio file lacks a noise-only portion for DINR to analyze, you can still obtain reasonable results by selecting and learning a segment of audio that has a relatively
low amount of signal and a
high amount of noise (as in a quiet passage). By then selecting a frequency range of the noise signature and using the
Auto Fit function to generate a generic noise curve, you can recompute the Contour Line based on this selection.
Some editing of the newly generated Contour Line will probably be necessary to yield optimum results, since it is not based entirely on noise from your audio file.
See Editing the Contour Line.
5 In BNR’s Spectral Graph, Control-drag (Windows) or Command-drag (Mac) to make a selection. Select points where the high-frequency noise components are most evident. In general, the flatter areas of the Spectral Graph, are better, since they represent quieter areas where there is probably less signal and more noise.