Why does vibrato create extra MIDI notes?

A single-line pitch detector can read changing pitch inside one held source note as separate events. In this controlled fixture, shallow plus-or-minus 0.5-semitone vibrato stayed one event at 5, 6, and 7 Hz, while plus-or-minus 1.0-semitone vibrato created 20, 24, and 28 events respectively.

Should I delete extra MIDI notes from vibrato?

Delete or merge short events only after replaying the matching source moment and confirming there was no separate re-attack. Keep one stable event when the source is one held note with vibrato; retry with a clearer phrase when the broader melody is wrong.

Is this a general vibrato accuracy score?

No. These are controlled synthetic source checks for one detector version and one set of settings. They are not a general voice, instrument, or song-transcription accuracy promise.

MIDI troubleshooting guide

Why vibrato creates extra MIDI notes

Vibrato can turn one held source note into a row of MIDI events. In this controlled single-line check, shallow vibrato stayed one event; at the deeper tested movement of plus or minus one semitone, 5, 6, and 7 Hz produced 20, 24, and 28 events.

Shallow vibrato remains one MIDI event while deeper five, six, and seven hertz vibrato creates 20, 24, and 28 events. — Depth decided whether the fixture split; at the same depth, rate decided how many extra events appeared.

Short answer: do not assume every short MIDI event is a separate note. Replay the matching source moment. If you hear one held note with vibrato and no re-attack, keep one stable MIDI event or merge the fragments. These fixture figures describe one controlled setup, not a general accuracy promise.

Convert one clear melody

What the controlled check measured

Each fixture was a two-second C4 source with a declared moving pitch. The purpose was not to score a singer or instrument. It was to separate two questions that are often mixed together: does vibrato start splitting a result, and once it does, does speed change how densely it fragments?

Declared source	Median result across 10 fixtures	What the result says
C4, 2 seconds, 5 / 6 / 7 Hz, plus or minus 0.5 semitones	1 retained event; 0 extra events at each rate	At this shallower tested depth, the one-line result did not fragment.
C4, 2 seconds, 5 Hz, plus or minus 1.0 semitone	20 events; 20 extra events	Deeper movement crossed the splitting boundary.
C4, 2 seconds, 6 Hz, plus or minus 1.0 semitone	24 events; 24 extra events	Same depth, more pitch cycles, more fragments.
C4, 2 seconds, 7 Hz, plus or minus 1.0 semitone	28 events; 28 extra events	Same depth, still more fragments.

The 5 / 6 / 7 Hz sequence contains 10 / 12 / 14 pitch cycles in two seconds. The matching 20 / 24 / 28 event counts show two detected events per cycle in this fixture. That internal pattern is why the result is more useful than an isolated anecdote.

Use the source before editing

Locate one held note. Work on a short exposed phrase where you can hear whether there is a new attack, syllable, pluck, or bow change.
Replay the same moment. If the source is one held note that wavers in pitch, a busy row of small MIDI blocks is not automatically a sequence of intended notes.
Choose the smallest repair. Merge or remove fragments only when the source supports one sustained note. Preserve separate events when you hear a real re-attack.
Retry instead of redrawing another melody. When the main contour follows accompaniment or another voice, trim an exposed phrase and convert again.

Do not confuse drift with vibrato

Slow pitch drift and vibrato are different axes. In the same controlled work, a C4 source moving plus or minus 1.5 semitones at 0.10 Hz gave one event; 0.125 Hz was the first tested drift rate to add an extra event. That is slow movement, not ordinary vocal vibrato. Keep the names with their conditions instead of borrowing one number to describe the other.

Fixture conditions and limits

Every parameter point used 10 generated mono WAV fixtures with the same declared pitch, duration, amplitude envelope, 44.1 kHz sample rate, and detector settings. The fixtures varied documented initial phase so frame alignment was exercised without changing the tested condition. Matching used the declared pitch at an output event midpoint, at least 50% time overlap, and a 50-cent tolerance. An extra event was unmatched, or a second output event matched to the same input note.

Detector settings were Balanced Melody / solo instrument mode, a 2048-sample analysis window, and a 1024-sample hop. Detector source SHA-256: DDA185AC8FB6A179A4FCFA73FD4197ACC95B8B6F59B233F5D3E330F596309F9F.

These are controlled source checks, not a general voice, instrument, complete-song, stem-separation, chord-transcription, or universal accuracy claim. Quote the conditions with the figures; without them, the figures become an unsupported promise.

Questions about vibrato and MIDI events

Why does my vocal vibrato look like many MIDI notes?

The detector follows changing pitch over time. When the movement is large enough in relation to the analysis, it can mark separate events inside one held vocal note. Check the source for a real re-attack before treating each event as intentional.

Should I quantize vibrato fragments?

Do not quantize first. Decide whether the fragments represent real separate notes or one held source note. Merge a source-backed artifact before applying timing changes.

Can this convert a full vocal performance into finished MIDI?

No. Start with one exposed melody line. This tool does not separate stems, transcribe a full arrangement, or reconstruct every vocal ornament as notation.