Each voice is a monophonic subtractive synthesizer, combined into a common low-pass filter in parallel.