Pitch correction (Auto-Tune / Melodyne class).
| Package | What |
|---|---|
@audio/tune-snap |
scale/key correction: pYIN pitch → the notes meant on the scale → each centred on its note, slow drift flattened → one curve through tune-curve |
@audio/tune-midi |
melody correction — YIN → per-guide-note median → PSOLA retune to a MIDI/note guide |
@audio/tune-curve |
a drawn pitch curve: the voice's own glottal cycles re-spaced (TD-PSOLA): formants and consonants kept, streaming |
import { snap, tuneMidi } from '@audio/tune'
snap(data, { scale: 'major', root: 0 }) // to C major, vibrato kept
snap(data, { speed: 0 }) // the hard-tuned effect
tuneMidi(data, { guide: parsedSmf.notes }) // retune to a reference melody
tuneCurve(data, { semitones: { t: [0, 1, 1.5], v: [0, 2, 0] } }) // along a curve, as drawn (Melodyne class)tune-snap builds one pitch curve over the whole take and retunes along it with tune-curve, so notes are never cut out and crossfaded back over the dry voice. tune-midi still shifts each guide note apart with @audio/shift-psola and crossfades it back.