What Is Voice Allocation?

Between your fingers and the sound sits a quiet bookkeeper: the allocator, deciding which of a synth’s voices takes each key you press. Its policies shape how the instrument feels.

The bookkeeper between your hands and the sound

Every time you press a key, something must answer a small question: which voice plays this note? The component that answers is the voice allocator. On one-voice and unlimited instruments the answer is trivial; everywhere between, allocation policy is a real part of how the instrument feels.

The policies

A polyphonic allocator keeps a list of free voices, assigns the next free one to each key press, and reclaims voices as notes end. When none are free it applies its stealing rule — on the instruments modeled here, last-note priority: the oldest sounding note is sacrificed (the voice-stealing guide walks through it). A paraphonic allocator tracks pitches but routes them into one shared envelope. A monophonic allocator has one voice and a distinctive extra: when a new key overlaps a held one, it can glide — slide the pitch from the old note to the new one rather than retriggering. That portamento annotation is why legato mono lines sound the way they do.

Glide, in milliseconds

On this site’s mono synth the legato glide is a real, inspectable number: 60 ms of portamento whenever a new key overlaps a held one. Hold a low C, tap the G above it without releasing, and the pitch slides for sixty milliseconds instead of retriggering — the annotation the allocator attaches (glide from C) is the same fact the sound acts on. Detach the notes and the glide disappears, because the allocator only glides overlapping presses.

Why velocity sometimes doesn’t matter

Allocation also normalizes what a key press even is. The organ and the paraphonic model here ignore how hard you strike — every note enters at the same level, by architecture, not accident — while the pianos and the poly/mono synths respond to touch. The site’s allocator applies the same rule its hardware inspirations do, so what you hear on a chart matches the architecture named in its title.

One allocator, audited

Keys Warrior’s playback runs on a single, deliberately simple allocator: last-note stealing for the polysynths, one shared envelope for the paraphonic pages, single-voice glide for the mono synth. It is pure, deterministic code — the CI suite replays chords through it and asserts the steal, the glide, and the shared envelope land exactly where these guides say they do.