A "voice" is the complete signal path (oscillators, filter, envelopes) needed to produce one note; polyphony is simply how many voices a synth has available to run at once.
Voice stealing
When you try to play more notes than available voices, the synth has to "steal" a voice from a currently sounding (often still-decaying) note to free it up for the new one - typically cutting that older note off abruptly rather than letting it finish naturally. This is completely normal behavior on every polyphonic synth once its limit is reached, not a malfunction.
How many voices are actually enough
- Playing simple melodic lines or basslines rarely needs more than a handful of voices.
- Full chords, especially with a long release time so old notes are still ringing when new ones are played, use up voices much faster than the number of keys held down at any instant suggests.
- As a practical rule of thumb: for chord-based playing with sustain, doubling the number of simultaneously held notes gives useful headroom so previous chords can ring out without being cut off by new ones.
Many classic vintage analog polysynths offered only a handful of voices (commonly in the single digits) by modern standards, which is worth knowing when comparing a vintage instrument's specs to a modern synth with much higher polyphony as standard.
