What is the primary difference between FFT-based noise reduction and spectral subtraction noise reduction in vocal editing?

The primary difference between FFT-based noise reduction and spectral subtraction noise reduction lies in how they identify and remove unwanted noise from an audio signal. Spectral subtraction, the older method, operates by first analyzing a section of the audio that contains only the noise (a 'noise profile'). It then subtracts this noise profile from the entire audio signal's spectrum. It assumes that any consistent energy present in the noise profile is also present throughout the recording and therefore subtracts it. FFT-based noise reduction, using the Fast Fourier Transform, provides a much higher resolution spectral analysis. This allows for a more detailed and frequency-specific identification of noise components. FFT-based systems often allow for manual drawing or shaping of the noise profile, giving the user more control. Furthermore, many FFT-based systems do not simply subtract the noise but use more sophisticated algorithms to estimate and attenuate the noise components while preserving the desired signal. The key difference is in the level of precision and control. Spectral subtraction tends to be less precise and can often result in a 'watery' or 'phas