From MP3 to Playable Piano Score — A Step-by-Step Tutorial

Step 1: Choose the right MP3

The converter on this site uses autocorrelation pitch detection — a method that finds one fundamental frequency per analysis window. This means it detects one pitch at a time. A solo melody line — a single voice, a single flute, a single synthesizer lead — converts cleanly. A full mix with drums, bass, chords, and vocals all playing at once does not. The detector locks onto whatever pitch is loudest in each window, which may not be the melody you want.

For a piano score, the ideal input is:

• A recording of a single melodic instrument or voice with no accompaniment
• A monophonic line — one note at a time, not chords
• A recording with minimal reverb and no distortion
• A clean recording where the instrument is the loudest thing in the audio

If your MP3 is a full song, you will not get a clean piano score from it. You will get fragments of the loudest pitch at each moment, with gaps where the detector found silence or noise. If you want to extract a melody from a mixed recording, you would need to isolate the melody track first using a stem separation tool, then feed the isolated vocal or instrument track into the converter.

Step 2: Convert the MP3 to MIDI

Go to the converter page and either drag your MP3 into the drop zone or click to browse. The tool accepts MP3, WAV, OGG, M4A, AAC, FLAC, and Opus files. When the file loads, the browser decodes it locally:

audioBuffer = await ctx.decodeAudioData(buf);

This call happens in your browser's AudioContext. The file data never leaves your device. There is no upload to a server, no API call, and no login requirement.

After decoding, you see a waveform preview and a Convert button. Click Convert. The tool runs the pitch detection pipeline, generates a Standard MIDI File, and gives you a download link for the .mid file. You also see a detected-note list showing each note's name (like C4, D#4, F4), start time, and duration.

Step 3: Open the MIDI as a piano score

Once you have the .mid file, you can open it in any of these programs to see it as notation or a piano roll:

MuseScore (free, desktop)

Open MuseScore, go to File > Open, and select your .mid file. MuseScore imports the MIDI as standard notation. You will see each note on a staff with its pitch and duration. The MIDI file's tempo (BPM=120) is imported as the score tempo. Each note appears with the pitch the detector found and the duration it measured.

Because every note in the MIDI file has velocity 90, MuseScore will not show dynamic markings — the velocity value is uniform. You can add dynamics (crescendos, accents, staccato markings) manually in the score editor after import.

Flat.io (free, browser)

Go to flat.io, create a free account, and import the MIDI file. Flat renders it as sheet music in your browser. You can edit notes, adjust durations, and add expression markings directly in the browser interface.

Your DAW's piano roll

If you are producing music rather than reading notation, open the MIDI file in your DAW — Ableton Live, FL Studio, Logic Pro, Cubase, Studio One, Reaper, or Bitwig Studio all import SMF format 0 files. The notes appear in the piano roll, where you can drag them, change their lengths, adjust velocity, and reassign the instrument.

For a detailed walkthrough of importing MIDI into specific DAWs, see MP3 to MIDI Into Your DAW.

Step 4: Clean up the score

The converted MIDI is a starting point, not a finished score. The autocorrelation detector produces pitch and timing data that is accurate enough to recognize the melody, but it has known behaviors you need to correct:

Fix octave errors

The autocorrelation method can lock onto a harmonic of the fundamental frequency instead of the fundamental itself. This means a note might appear one or more octaves higher than what you hear in the original recording. If a passage looks right melodically but sits too high or too low on the staff, select those notes and transpose them by one or more octaves in your notation software.

Fix note durations

The detector steps through audio in hops of 0.023 * sampleRate samples with a window of 0.046 * sampleRate samples. This means note boundaries are detected at discrete hop positions. A note's reported start time is the hop boundary, not the exact onset moment. Additionally, notes shorter than 0.05 seconds are filtered out entirely — they do not appear in the MIDI file.

In practice, this means note durations in the score will not match the original performance precisely. Some notes may be too long (the detector held them across a hop boundary) or too short (the detector caught only part of the note). Adjust durations in the notation editor until the rhythm looks right.

Remove false detections

The detector uses an RMS threshold of 0.008 to reject silence, and an amplitude threshold of 0.2 to trim the edges of each analysis window. In noisy recordings or quiet passages, the detector may produce spurious notes — pitches that do not correspond to anything you hear in the original. Compare the score against the original recording and delete any notes that do not belong.

Add the key signature and time signature

The MIDI file does not contain a key signature or time signature in musical notation terms. MuseScore and Flat will attempt to detect these on import, but you should verify them. If you know the key of the original recording, set it manually. The tempo in the MIDI file is BPM=120 — if the original recording was at a different tempo, change the project tempo in your notation software or DAW before working with the score.

Step 5: Play and verify

After cleaning up, play back the score in your notation software or DAW. Compare it to the original MP3. If a passage sounds wrong, go back to the detected-note list on the converter page — it shows each note's name, start time in seconds, and duration in seconds. This list is the raw data behind the MIDI file, and it can help you identify which notes the detector got right and which ones need correction.

You can also use the MIDI to WAV converter on this site to render the cleaned MIDI back to audio. This lets you hear the result without opening a DAW, and it gives you a quick check on whether your edits improved the score.

What kind of melodies convert best

Based on how the detector works — autocorrelation with a frequency range of 0 to 2000 Hz (if(f > 0 && f < 2000)) — melodies that convert most cleanly to a piano score are:

• Single-instrument recordings with no competing sounds
• Melodies in the mid-range — the frequency-to-MIDI formula Math.round(69 + 12 * Math.log2(f / 440)) maps frequencies around 440 Hz to MIDI notes in the 60–80 range, which corresponds to the middle of the piano keyboard where notation is most readable
• Melodies with clear gaps between notes — the detector's ±1 semitone merge rule (if(Math.abs(midi - active.midi) <= 1)) means that if two consecutive notes are adjacent semitones and the gap between them is smaller than one hop, they may merge into one note
• Melodies without rapid runs — notes shorter than 0.05 seconds are filtered out, so very fast passages may have missing notes

What this tutorial does not cover

This tutorial covers converting a single melodic line to a piano score. It does not cover:

• Chord extraction — the detector is monophonic. For chords, read why MP3 to MIDI cannot hear chords.
• Multi-instrument transcription — for separating instruments from a mix, see converting vocal, bass, and guitar MP3 to MIDI.
• Polyphonic transcription — tools that use neural networks can detect multiple simultaneous notes. They require server-side processing and are not free. For the difference, see free vs paid MP3 to MIDI.
• Full sheet music with expression — the MIDI file contains pitch and timing only. For what MIDI cannot capture, read MIDI format limitations.

Summary: the full path