rsdio writes: This is a strange one. It claims to prevent clipping, but still appears to introduce distortion. Perhaps the idea is that soft clipping is preferred to hard clipping? If there were a voltage-controlled gain input, then true peak limiting would be possible, but this seems to trade one form of distortion for another. In any event, if you are interested in audio distortion circuits, then this article should certainly be interesting.
To enable an audio amplifier to accommodate both loud and soft-spoken voices without clipping the loud waveforms, you can implement a nonlinear transfer function by adding a few components to a standard application circuit.