Convert a recording of piano, guitar, singing or whistling into MIDI notes. The Basic Pitch AI model (by Spotify, running in your browser with TensorFlow.js) detects chords and melodies; a built-in tracker handles single-note lines instantly. Fine-tune the result with the sliders, check it on the piano roll and play it back, then download a standard .mid file.
Last updated:
Showing a demo melody.
Fewer notesMore notes
Longer notesShorter notes
20 ms300 ms
0:00.0
Advertisement
Advertisement
How to use the Audio to MIDI Converter
Choose a recording: drop an MP3, WAV or other audio file. A demo melody is converted when the page opens.
Pick the engine: Basic Pitch AI for piano, guitar and anything with chords; the built-in tracker for singing, humming or whistling one note at a time. Press Convert to MIDI.
Adjust the sliders: raise Note sensitivity if notes are missing, lower it if you see stray notes. Raise Minimum note length to remove short blips.
Check the piano roll and press play with MIDI notes, Original or Both to compare.
Set tempo and quantize if you want notes snapped to a grid, then Download .mid file and open it in your DAW or notation app.
Getting a clean transcription
Recording
Engine
Suggested settings
Solo piano
Basic Pitch AI
Sensitivity 45–55 %, sustain 50 %, min length 60–100 ms
Acoustic or clean electric guitar
Basic Pitch AI
Sensitivity 40–50 %, lowest note E2, min length 80 ms
Singing or humming
Built-in tracker
Sensitivity 50 %, min length 100–150 ms
Whistling, flute, recorder
Built-in tracker
Sensitivity 50–60 %, lowest note C4
Bass line
Built-in tracker (or AI)
Highest note C4, min length 80 ms
Full song mix
Basic Pitch AI
Sensitivity 30–40 %, then clean up; isolate the part first if you can
Notes, MIDI numbers and frequencies
MIDI numbers each key from 0 to 127; middle C (C4) is 60 and A4 (440 Hz) is 69. A frequency f converts to a MIDI note with n = 69 + 12 × log2(f ÷ 440), rounded to the nearest whole number.
Note
MIDI
Frequency
Typical source
A0
21
27.5 Hz
Lowest piano key
E2
40
82.4 Hz
Guitar low E string
C4
60
261.6 Hz
Middle C
A4
69
440.0 Hz
Tuning reference
C6
84
1046.5 Hz
High soprano, whistling
C8
108
4186.0 Hz
Highest piano key
Tips
Record close to the instrument in a quiet room; reverb and background noise turn into extra notes.
Sustain pedal and ringing strings make notes overlap. Raise the sustain threshold if the MIDI sounds smeared.
The detected notes come from pitch, not from music theory: check accidentals and octaves before printing sheet music.
Find the song's tempo with a BPM finder and enter it before quantizing, so the MIDI lines up with your project grid.
Frequently asked questions
How does the audio to MIDI conversion work?
With the AI engine, the audio is resampled to 22,050 Hz mono and fed to Basic Pitch, an open-source neural network from Spotify's Audio Intelligence Lab. It outputs, about 86 times per second, the probability that each of 88 piano keys is sounding and that a note starts. Notes are then extracted with the sensitivity (onset) and sustain (frame) thresholds you set. The built-in tracker instead measures a single pitch with the YIN algorithm and splits notes where the pitch or loudness changes.
Can it convert chords and full songs?
Basic Pitch is polyphonic, so piano chords, strummed guitar and several instruments at once are detected. A full mix with drums, bass, vocals and effects produces many extra notes, though: results are best with one instrument recorded cleanly. For a song, isolate the part first (for example with a vocal remover) and convert that.
Which engine should I use for singing or humming?
Use the built-in tracker for a single voice, whistling, flute or any instrument that plays one note at a time. It ignores harmonics and breath better than a polyphonic model, works instantly and needs no download. Basic Pitch also works on voice but may add octave or harmony notes.
Why are there extra or missing notes?
Lower Note sensitivity to remove ghost notes from harmonics, room echo or noise; raise it to catch soft notes. Raise Sustain threshold if notes run on too long, and increase Minimum note length to drop very short blips. Limiting the lowest and highest note also removes rumble and squeaks. Expect to tidy up the MIDI in your editor: no automatic transcription is perfect.
Does quantize change the timing?
With Quantize off, notes keep their exact timing in seconds and the MIDI file just uses your tempo to convert seconds into beats. With quantize on, every note start and end snaps to the nearest grid step (for example a 1/16 note at your BPM). Set the tempo to the song's real BPM first, or the grid won't line up with the beat.
What is in the .mid file?
A standard MIDI file (format 0) with one track: a tempo event, a 4/4 time signature, a program change for the instrument you picked, and note-on/note-off events with velocities taken from the loudness of each note. It opens in GarageBand, FL Studio, Ableton Live, Logic, MuseScore and any other MIDI app.
Is my audio uploaded?
No. Decoding, the AI model and the MIDI writing all run in your browser. The first time you use the AI engine, the model files and TensorFlow.js are downloaded from a public CDN, but your recording never leaves your device.
Are all the tools on FreeToolCabin tested for quality and accuracy?
Yes! Every tool on our site undergoes automated Q/A testing in real browser environments to ensure high precision, correct output, and zero errors. If you happen to spot a bug or have a suggestion, please contact us at [email protected] and our team will review it right away.