How to transcribe audio to sheet music
Updated July 2026
Transcribing audio to sheet music sounds like a pitch problem. It is mostly a rhythm problem. Hearing which notes were played is the easy half — deciding which note carries the beat, where bar one falls, and how to write a phrase that leaned against the tempo is what makes a page readable or useless.
First, work out what is in the file
Two cases, and they do not get the same treatment. A recording of one instrument playing alone — a solo piano take, an unaccompanied horn, a single guitar overdub — can go straight to a transcriber. A finished stereo mix cannot. In a mix the instruments share the same frequency space, and a transcriber pointed at the whole thing spends most of its effort on whatever is loudest while the inner parts smear together.
Audio is routed per instrument once it is known what is being read: a dedicated polyphonic model for piano, register-conditioned pitch tracking for solo winds, brass, strings and bass, and a general polyphonic model for guitar and ensemble material. Name the instrument or let it be inferred.
For a mix, separate before you transcribe
Splitting the song into stems — vocals, drums, bass, guitar, keys — and transcribing each stem on its own is the single biggest accuracy win in the whole process. It is not a preprocessing nicety. A bass line read off an isolated bass stem, and the same line dug out from under a full band, are different results, and the difference shows on the page.
The order is: separate, transcribe each stem, then assemble the parts into one score. Dense mixes, heavy distortion and heavy reverb are harder either way — separation improves them, it does not make them easy.
The grid is decided before the notes
This is where most audio-to-notation output falls apart. Pitch detection is good enough for a lot of material; metrical interpretation is what decides whether the page is readable. Is that a dotted eighth or the first note of a triplet? Does the phrase start on the downbeat or a beat ahead of it? Where is bar one?
So tempo, downbeat and meter are detected first and the notes are fitted to that grid, rather than the grid being reverse-engineered from a pile of note events afterwards. The tempo map is allowed to move: a ritardando into the last chorus stays a ritardando instead of being flattened onto a click. Chords are detected in the same pass and come back as a chord track.
What turns notes into sheet music
A score is more than pitches stacked on a staff. Multiple parts are bracketed into one system with the bar lines connected and aligned, so bar nine of the bass sits directly under bar nine of the horns. Each part gets the clef that fits its register — a bass line on bass clef, not three ledger lines under a treble staff.
The key is detected and printed as a key signature, with accidentals only where a note departs from it, and the spelling follows the key: the same black key is D♭ in A♭ major and C♯ in D major. Transposing instruments are written the way their players read them — B♭ trumpet and clarinet a major second up, E♭ alto sax a major sixth, F horn a fifth, B♭ tenor sax a major ninth, guitar and bass an octave — with the key signature transposed alongside the notes. A concert-pitch toggle puts everything back at sounding pitch.
Correct it, then export it
No transcription of a real recording is perfect, and a tool that pretends otherwise is just hiding the errors somewhere you will find them later. The result opens as an editable multitrack MIDI arrangement, one lane per part. Fix a wrong note there and the score re-engraves from the same edit; the MIDI you export is what you corrected, not what the model first guessed.
Out the other side: MIDI per part, or a Logic or Ableton session with the stems on named tracks, per-part MIDI, the tempo map and a chord guide. What you do not get is publication-ready engraving — final spacing, page turns and articulation polish belong in a notation editor.
Transcribe a recording to sheet music
Drop an audio file and get a score, per-part MIDI and a chord track back. Runs in the browser, no install, with free runs daily.
Open the transcriber →FAQ
Can I transcribe audio to sheet music without installing anything?
Yes. It runs in the browser — you upload a file and get a score, per-part MIDI and a chord track back. Free runs are included daily with an account.
Do I need to say which instrument I recorded?
You can, and it helps. Audio is routed per instrument: a dedicated polyphonic model for piano, register-conditioned pitch tracking for solo winds, brass, strings and bass, and a general polyphonic model for guitar and ensemble material. If you say nothing, the instrument is inferred.
Will the sheet music be ready to print and hand to a player?
It will be readable and correctly transposed, with a real key signature and a clef chosen per part. It is not publication-ready engraving — final spacing, page turns and articulation polish belong in a notation editor, and the export is built to hand off to one.
What happens to the chords?
Chords are detected in the same pass as the notes, including over a moving tempo, and come back as a chord track, so a lead sheet carries symbols above the staff rather than note heads alone.