Hum the tune in your head, then turn it into music you can edit

You can hear the melody, and you can hum it, but putting the notes into music software feels like a different skill. Basic Pitch offers a useful bridge. Spotify’s free browser tool accepts a sung or instrumental recording and returns a MIDI file that can be opened and edited in a digital audio workstation.[1]

The rewarding part comes afterwards: the phrase can become a piano line, a different rhythm or the starting point for your own short piece. This is a documented workflow, not a transcription experiment conducted by this publication. Basic Pitch was introduced in 2022; the opportunity here is to use that existing capability creatively rather than mistake it for a new release.[2]

Capture one musical thought

Begin with a short phrase you made up yourself—perhaps eight to fifteen seconds, sung clearly in a quiet room. That length is our suggested starting exercise, not a product limit. Leave a little silence at either end and keep the original recording. A melody with a recognisable beginning and ending is easier to compare than a minute of improvisation.

The browser demo lets you record into a computer or select an audio file, then download its MIDI interpretation. A computer browser is the documented starting route. You can transfer the resulting MIDI to a phone for editing; this guide does not assume that microphone recording works identically in every mobile browser.[1]

Basic Pitch supports more than one note at a time, but its maintainers recommend one instrument at a time. A solo voice or guitar is therefore a better first input than a full band. Polyphonic transcription and separating instruments from an existing mix are different jobs.[3]

What the AI has given you

MIDI represents notes and performance instructions rather than preserving the original recorded voice. Spotify’s model estimates notes, their beginnings and pitch bends. The music editor can use those instructions to play a software instrument. That is why a hummed phrase can emerge as something you can move around on a piano-roll display.[2]

This makes an initially imperfect result useful. You do not have to accept a whole generated song or start again. One wrong note can be corrected; a note that lasts too long can be shortened. The recording remains your reference for what you intended, while the MIDI becomes material you can develop.

Bring the phrase into a music editor

Use any editor that can import MIDI and expose its notes. GarageBand on iPhone provides one concrete route. Save the MIDI in Files, set the song section to Automatic, then open Tracks view, the Loop Browser and Files. Choose Browse items from the Files app, locate the MIDI and drag it into Tracks view at the intended starting position. The Automatic setting prevents the current section length from cutting off the import.[4]

Apple documents MIDI import into Keyboard or Drums tracks. For a sung melody, start with a Keyboard track and a clear piano sound. That sound choice is our recommendation: it makes the note sequence easier to hear before a rich instrument or effect changes its character.[4]

Open the green instrument region by double-tapping it and choosing Edit. The notes appear as bars. Apple’s editor lets you move them vertically to change pitch, horizontally to change timing, and adjust their length and velocity. You can also work without snapping everything to the grid.[5]

Play the original, then the MIDI. Correct the phrase one small section at a time. Listen for an unexpected leap, a note missing from a repeated pattern or a sustained note broken into several tiny ones. These are suggested checks, not errors observed in our own trial. If the transcription barely resembles the idea, try a cleaner, simpler recording before spending a long time repairing it.

Make a variation you could not easily play

Once the melody is recognisable, duplicate the working phrase and change one property. Move the ending higher. Delay the last note. Shorten the opening notes while keeping the long final one. Compare each variation with the first version. Limiting the edit to one choice makes it easier to hear why you prefer it.

A particularly satisfying experiment is to keep the first half and write a different response in the second half. You now have something close to a musical question and answer. The AI supplied a route from sound to notes; your listening decides which relationship between those notes deserves to remain.

Pitch-bend information can preserve aspects of a sung slide, but how it sounds depends on the receiving software and instrument. Treat that expression as something to audition rather than expecting every voice nuance to survive conversion. The model’s documented support gives you another editable possibility.[3]

The finish line for a first session can be modest and complete: one recognisable melody, one deliberate variation and a project you can reopen. You have taken an idea that existed only in your voice and made it available for composition. That is a useful kind of creative independence, even before there is a finished song.

Report an error or suggest a correction