Soft clipping starts to reduce the peaks a little below the threshold level and progressively increases its effect as the input level increases such that the threshold is never exceeded.

Soft Clipping works in much the same way as 'Hard Clipping', but is less fierce in that it 'squashes' the peaks rather than cutting them Hard clipping may also be useful for producing synthetic signals for scientific purposes. For heavier use of distortion the 'Soft Clipping' option may be preferable. Excessive use of hard clipping creates a harsh distortion that is usually unpleasant. Hard clipping may be useful for purposefully introducing distortion on high peaks, for example to add high harmonics to percussive sounds. It just chops off the peaks at the 'Limit to' threshold. Hard Clipping is the simplest method for reducing peaks. Where the input level (after applying optional input gain) exceeds the threshold, an equal amount of negative gain is applied so that the peaks never exceed the threshold. Hard Limit makes no change to the audio until the peaks reach the 'Limit to (dB)' threshold. Soft Limit (default) progressively reduces the gain as the amplitude of the waveform approaches the 'Limit to (dB)' threshold and prevents the waveform from exceeding that level.

Audacity effects for vocals