Audio to MIDI: From a Hum to Editable Notes

An idea you hum into your phone is audio: a fixed recording of one performance. MIDI is different — it's a list of notes, each with a pitch, a start time, a length and a velocity. Turn audio into MIDI and you can change the sound, fix a wrong note, transpose, quantise or double the part with another instrument, all without re-recording.

Audio-to-MIDI (also called transcription) listens to a recording and writes those notes for you.

What it's good at

• Single-note melodies. A sung or hummed line, a lead synth, a bass line, a solo guitar riff. One note at a time is the easiest case and gives the cleanest results.
• Lightly layered chords. Strummed guitar or held keys can be transcribed, but expect to tidy up extra or missing notes.
• Drums, with a dedicated drum transcriber that listens for hits rather than pitches and maps them to drum lanes.

It writes notes, not sound. You won't get the original guitar tone back; you'll get its pitches and timing, played by whatever instrument you choose.

Three ways to get notes in SoundGen

1. Hum straight into the piano roll. Open a MIDI clip in the piano roll, press Cmd+I (Ctrl+I on Windows), hum or sing, and press it again to stop. The notes appear at the next bar.
2. Convert a clip on the timeline. Right-click an audio clip, open the AI menu and choose Convert to MIDI. The part becomes a new MIDI track lined up under the clip. Save the project first so the clip is uploaded.
3. Use the free browser tool. The audio to MIDI converter turns a melody, bass line or riff into a MIDI file you can download, without signing up, and your audio stays on your computer.

For drums, ask the assistant — it listens for the hits and writes them as drum notes.

Getting a cleaner transcription

The quality of the result depends mostly on the recording:

• One instrument at a time. A solo part beats a full mix every time. If the part you want is inside a finished song, separate the stems first and convert just the one you need.
• Keep it dry. Reverb and delay smear notes into each other, so the transcriber hears extra notes or long held ones.
• Hum with consonants. "Doo" or "da" on each note gives sharp starts the transcriber can find; a smooth "ooo" blurs note boundaries.
• Stay in tune and steady. Big slides and wide vibrato get written as runs of short notes.
• Record at a decent level without clipping and with little background noise.

Cleaning up the result

Expect to spend a few minutes in the piano roll. A good order:

1. Delete the junk. Very short notes (ghost notes from breaths, slides or bleed) usually stand out — remove them or merge them into neighbours.
2. Fix octave errors. Transcribers sometimes put a note an octave too high or low, especially on bass and low voices.
3. Check the key. Detect the key of the clip or song (the key, BPM & chord finder does it in the browser) and look for notes a semitone outside it. Not every out-of-key note is wrong, but they're the first ones to check.
4. Quantise gently. Snap note starts to the grid if you want a tight part, but consider a partial strength (50–70%) to keep some human feel.
5. Even out velocities if the instrument you've chosen exaggerates dynamics.

What to do with the notes

Once you have clean MIDI:

• Re-voice it. Play a hummed melody back on a synth lead, a piano or strings.
• Double it. Put the same line on a second instrument an octave up for a thicker lead.
• Harmonise it. Copy the part and move it up a third or a sixth within the key.
• Build around it. Use the bass line's root notes to write a chord progression, or the melody to set the song's key and tempo.

Audio-to-MIDI turns the first, fragile version of an idea into something you can build a production on. The faster you capture an idea as notes, the more of it survives.