What Is MIDI Polyphony and Voice Stealing?

MIDI polyphony refers to the capacity of a digital instrument or synthesizer to sound multiple notes at the same time, constrained by the hardware or software's available voice count. When a performance requires more simultaneous voices than the system can process—often due to dense chords or sustained release tails—the instrument must implement "voice stealing" to manage its resource ceiling. This article explores how polyphony functions in MIDI environments, why limits exist, and the specific allocation algorithms synthesizers use to drop active notes to make room for new ones.

Understanding MIDI Polyphony

In electronic music and MIDI (Musical Instrument Digital Interface), polyphony defines the maximum number of distinct audio voices an instrument can generate concurrently. Unlike acoustic instruments like a piano, which vibrate naturally based on physical strings, digital and analog synthesizers rely on dedicated voice architectures.

A "voice" is not just a single pitch; it represents a complete signal path that includes oscillators, filters, envelope generators, and modulation sources. An instrument with 8-voice polyphony can produce eight individual signal paths simultaneously. If a patch uses complex layering—such as two detuned oscillators per note—some synths allocate two voices per key, effectively halving the playable polyphony.

How Polyphony Limits Are Reached

Polyphony limits are strictly bounded by processing power (CPU/DSP) in digital instruments, or by physical circuit components in analog synthesizers. A performer can easily exhaust an instrument's polyphony through common musical practices:

• Sustain Pedal Usage: Holding the sustain pedal keeps the envelope release phases open, preventing voices from closing even after keys are released.
• Long Release Times: Pads and ambient presets with lengthy decay or release tails keep voice channels occupied long after a chord change.
• Dense Arrangements and Unison Modes: Playing fast arpeggios, complex chords, or activating multi-voice unison stacking rapidly multiplies the number of voices required.

Once the total active voice count matches the maximum voice limit of the synthesizer, any incoming MIDI Note On message triggers the voice stealing process.

How Voice Stealing Occurs

Voice stealing is an automated resource management process designed to prevent the system from ignoring new notes or crashing due to overload. Instead of dropping incoming notes, the synthesizer terminates an existing, active voice to reassign its resources to the new note.

To make this transition as imperceptible as possible, synthesizers rely on predetermined voice allocation algorithms:

• Oldest Note Priority (FIFO): The voice assigned to the note that has been ringing the longest is cut off first. This is common, but it can disrupt sustained bass notes while playing melodies on top.
• Lowest Amplitude / Quietest Note: The algorithm scans for the voice currently generating the least volume—typically a note far into its decay or release phase—and reassigns it, minimizing audible clicks or sudden cutoffs.
• Released Note Priority: Notes that the player has physically released, but are lingering due to sustain or envelope release stages, are stolen before any note that is still actively being held down.
• Pitch Priority (Lowest or Highest Note Preservation): Some synthesizers prioritize specific registers. For example, lowest-note priority ensures that root bass notes are never stolen, sacrificing higher harmonic notes instead.
• Same-Pitch Reassignment: If a new note shares the pitch of a voice that is already decaying, the synthesizer interrupts that specific voice rather than cutting off a different pitch.

When voice stealing occurs abruptly, it can cause noticeable artifacts, such as audible clicks or unnatural dropouts in the stereo field. Modern synthesizers smooth out these artifacts by applying micro-fades (extremely fast volume envelopes) to the stolen voice just before it is terminated and re-triggered.