Provided are systems, methods and techniques for compressing the dynamic range of an audio signal. In one implementation: an input audio signal is obtained, and a time-varying gain signal is provided based on it and a desired output range. The time-varying preliminary gain signal is then applied to the input audio signal to provide an output audio signal. Provision of the time-varying gain signal involves lowpass filtering a signal that is based on the input audio signal, using an attack gain-response time and a release gain-response time as filtering parameters, with the attack gain-response time decreased and the release gain-response time increased in response to a determination that a transient is occurring in the input audio signal.