Skip to content
12notez

Audio to MIDI Converter

Free · No sign-up · Files never leave your device

Made for one voice or one instrument playing one line at a time. A full mix, or several instruments at once, gives messy results.

What is an audio to MIDI converter?

An audio to MIDI converter listens to a recording and writes down the notes it hears as a MIDI file, which holds each note's pitch, start and length but not the sound. This one runs Basic Pitch, a neural network from Spotify, in your browser. It is made for one voice or one instrument playing one line at a time. A full mix gives messy results.

How it works

  1. Step 1

    Your browser reads the file

    The file is decoded by your browser's own audio engine, mixed down to mono and resampled to 22,050 Hz, the rate the model expects. Nothing is uploaded.

  2. Step 2

    Basic Pitch finds the notes

    After you pick a file, a background worker loads the note model (about 225 KB) and the ONNX Runtime engine, then runs the model on about two seconds of audio at a time. Note sensitivity, minimum note length and pitch range turn its output into notes, and you can change them afterward.

  3. Step 3

    Check, then download

    See the notes on the piano roll, play the original and then the notes on a plain synth, set the tempo (120 BPM to start) and download a Standard MIDI File.

Questions

Is my audio uploaded anywhere?

No. Your file is decoded, analyzed and turned into a MIDI file inside this browser tab. After you pick a file, the page loads the note model (about 225 KB) and the ONNX Runtime engine from 12notez.com. Nothing about your file is sent to any server.

What kind of recording should I use?

A clear recording of one voice or one instrument playing one line at a time, such as a sung melody, a whistled tune, a flute, a bass line or a guitar solo. The model can find several notes at once, so chords from one instrument are possible, but a full mix with drums and many instruments gives messy results. To work on one part of a song, split it first with the vocal remover.

Which files work, and how long can they be?

Any audio file your browser can play, usually MP3, WAV, M4A, FLAC and OGG. Files up to 10 minutes are accepted. If your browser cannot read a file, the page says so; converting it to WAV or MP3 usually fixes that.

Will the MIDI match my recording exactly?

No. The model predicts notes from the sound, so it can miss quiet notes, split one long note in two, or add notes from overtones and noise. Raise Note sensitivity to find more notes, raise Minimum note length to drop short ones, and narrow the pitch range to your instrument. Then play the notes next to the original and fix the rest in your music software.

What does the tempo setting do?

It sets the tempo stored in the MIDI file, 120 BPM to start and from 20 to 300. Each note keeps its time in seconds, so a different tempo changes the bar and beat grid your music software draws, not when the notes sound. Set it to the song's tempo if you want the bars to line up with the music.

Privacy: what runs where

  • Your audio is decoded, analyzed and written to a MIDI file in this browser tab. It is never uploaded.
  • The Basic Pitch model (about 225 KB) and ONNX Runtime load from 12notez.com after you pick a file. Nothing about your file is stored on our servers.
  • Basic Pitch is copyright Spotify AB and used under the Apache License 2.0; see the notices.
  • Only use audio you have the rights to. A MIDI file of someone else's song is still their composition.

Related tools