A feature of multitimbral synthesizers where a voice always is made available to sound new notes when all the synth’s polyphonic voices are in use. In the most common scheme, the most recently played note steals the voice from the oldest note currently sounding, or sometimes the lowest amplitude sound. The usual alternative to dynamic allocation is to assign an inflexible, predetermined number of voices per sound.
See voice stealing.