Would it be possible to implement some sort of adaptive track gain? Something that uses a timespan to calculate the gain to apply on a sound fragment. For example use a 3 second frame to calculate the gain needed. This way the gain applied varies from time to time so the soft sections are amplified more than loud sections...