Audio to Sheet Music

An automatic-transcription draft you can open in any notation editor. Spotify's Basic Pitch model turns the audio into note events; onsets and durations are snapped to a sixteenth- or eighth-note grid at the detected or supplied tempo. The key signature comes from Krumhansl–Schmuckler correlation: a duration-weighted pitch-class histogram is matched against all 24 rotated Krumhansl–Kessler profiles, and the best Pearson correlation picks the key (with confidence). The engraver then writes real notation: divisions of 24 per quarter, barlines from your time signature, chords when simultaneous notes share a span, dotted durations, ties across barlines, accidentals spelled against the key signature (sharps or flats), and an F clef for bass transcriptions. You get MusicXML 4.0 — the format MuseScore, Sibelius, Dorico and Finale open directly — plus an ABC text score, and optionally a quantized MIDI file. Transcribe the full mix, or isolate vocals / bass / other instruments first with the local HT-Demucs model for a cleaner single line. Automatic transcription is a starting point: expect rhythm simplifications on expressive performances, and clean up ornaments and articulations in your editor.