How to transcribe a saxophone solo
Updated July 2026
A saxophone plays one note at a time, so the hard part is not polyphony — it is that the horn barely stops. Slurred runs with no re-articulation, ghost notes, scoops into pitch: the pitches are the easy half, and the onsets are where a transcription of a solo lives or dies.
One line, and still not easy
A pianist gives a transcriber ten notes and ten clean attacks. A saxophonist gives it one note and, across a four-bar phrase, maybe three attacks — the rest is a continuous tone bending between pitches. Where one note ends and the next begins is a decision, not a measurement.
That is also what makes a good solo transcription valuable and a mediocre one useless: the phrasing is the information. Notes without the right articulation boundaries read as scale exercises.
Get the rhythm section off it first
How you transcribe a saxophone solo depends on what you have. If the solo was recorded on its own, go straight to the transcriber. If it is inside a track, separate the mix into stems — vocals, drums, bass, guitar, keys — before transcribing anything. Pulling the drums and the comping out from under the horn is the single biggest accuracy win available, and each stem is then transcribed on its own rather than fighting the others for the same frequency band.
Honest limit: a horn inside a dense, heavily reverberant mix is one of the harder cases, and separation improves it rather than solving it.
Register is the setting that matters
Solo winds get register-conditioned pitch tracking rather than the general polyphonic model, so telling it the horn is an alto and not a tenor narrows the range it searches. The two places octave errors turn up are the extremes: a subtone low B♭ where the fundamental is weak and the second harmonic dominates, and altissimo, where the tone is thin and unstable.
Check both ends of the range first. The middle of a saxophone is the well-behaved part.
Written pitch, not concert pitch
Transcription hears concert pitch; a saxophonist does not read it. An E♭ alto part is written a major sixth above what it sounds, a B♭ tenor part a major ninth — and the key signature moves with the notes, so a concert C minor blues is written in A major on the alto part.
Get this wrong and the part is unplayable by the person you hand it to. Each instrument carries its own transposition, and a concert-pitch toggle collapses everything back to sounding pitch when you want to compare against the record.
Swing does not sit on the grid
Tempo, downbeat and meter are detected before the notes are fitted, and the tempo map is allowed to move, so a solo over a band that pushes is not flattened onto a click. Notation still has to commit to a written value for each note — that is what notation is — but the MIDI keeps the placement that was actually played.
So read the score for the line and the MIDI for the feel. Chords are detected in the same pass and come back as a chord track, which is what makes the solo analysable rather than just readable.
Then correct it by ear
No transcription of a real recording is perfect. The result opens as an editable multitrack MIDI arrangement, one lane per part; fix a note and the score re-engraves from that edit, and the exported MIDI reflects it. Export MIDI per part, or a Logic or Ableton session with stems on named tracks, per-part MIDI, the tempo map and a chord guide. Final engraving polish belongs in a notation editor.
Transcribe the solo
Upload the take or the whole track, name the horn, and get a written part plus editable MIDI back. Browser only, free runs daily.
Open the transcriber →FAQ
Which saxophone should I tell it I am transcribing?
The one you played: soprano, alto, tenor or baritone. Solo winds use register-conditioned pitch tracking, so naming the horn narrows the range being searched and cuts octave errors. If you say nothing, the instrument is inferred.
Does it write the part at concert pitch or at written pitch?
Both are available. A transposing part is written the way the player reads it — an E♭ alto part a major sixth above sounding, a B♭ tenor part a major ninth — with the key signature transposed to match. A concert-pitch toggle puts every part back at sounding pitch.
Can it pull a solo out of a full band recording?
Better than transcribing the mixdown directly, yes. Separate the song into stems first and transcribe from there — getting the drums and bass off the horn is the single biggest accuracy win. A horn buried in a dense, heavily reverberant mix is still one of the harder cases.
Will it catch ghost notes, scoops and falls?
Not reliably. Those are expressive gestures with no single correct notated form, and a model that guessed at all of them would be inventing detail. The result comes back as editable MIDI alongside the score, so you decide what gets written and what stays feel.