MIDI input check

Why does audio to MIDI choose the wrong octave?

In this controlled six-note check, the high C6/E6 source and a C4/E4 source with a simultaneous C5 competitor each produced a median of three octave-down outputs. Removing the C4/E4 fundamental did not.

High C6/E6 and a competing C5 condition each show three octave-down MIDI outputs in a six-note fixture.
Only the tested high range and simultaneous competitor produced octave-down events.
Short answer: check range and competing pitches before blaming a weak fundamental. In this fixture, high notes and a second simultaneous pitch produced octave-down results; removing the fundamental did not. Best with one exposed melody, not chords or a full mix.

Convert one clear melody

Controlled wrong-octave check

ConditionMedian across 10 fixturesOctave result
C4/E4 normal harmonic profile6 correct; 0 other0 up; 0 down
C2/E2 normal harmonic profile0 events; 6 without a matching eventNo octave event
C6/E6 normal harmonic profile3 correct; 3 extra3 down
C4/E4 without fundamental6 correct; 0 other0 up; 0 down
C4/E4 plus simultaneous C53 correct; 3 extra3 down

Read the result as a diagnosis, not a guarantee

The source alternated two target pitches for six 160 ms notes. The high range and competitor conditions each caused a median of three events one octave below the declared target. The low C2/E2 test did not produce a usable event, which is a different failure from an octave shift. This test does not show that every high note or every added pitch will fail.

Try these checks before editing

  1. Shorten the source: leave one exposed melodic line rather than two pitches at the same time.
  2. Check the note range: compare a high phrase against a nearby lower phrase before moving every MIDI event by an octave.
  3. Keep the source beside the piano roll: correct only notes that you can confirm by listening to the matching moment.

Fixture conditions and limits

Each condition used 10 mono 44.1 kHz synthetic PCM fixtures, six alternating notes at 160 ms each, and 10 initial phases. The normal target had fundamental through fourth-harmonic gains of 1.00, 0.42, 0.20, and 0.08. The missing-fundamental condition used 0, 0.62, 0.36, and 0.20. The competitor condition added a simultaneous C5 at gain 0.65. Melody / solo instrument mode, Balanced sensitivity, a 2048-sample analysis window, and a 1024-sample hop stayed fixed. An octave-down result was an overlapping output exactly 12 semitones below the declared target. Detector SHA-256: DDA185AC8FB6A179A4FCFA73FD4197ACC95B8B6F59B233F5D3E330F596309F9F.

These are controlled source checks, not a universal octave, instrument, song, or accuracy promise.

Related: missing fundamental · high notes · competing pitches