A sung or played note is not one frequency. It is a stack of them: the fundamental, which is the pitch you name, plus overtones at two, three, four and more times that frequency. The overtones give a voice or a guitar its colour. The pitch you hear is set by how often the whole pattern repeats, which is the fundamental.
That is why a simple “find the loudest frequency” approach often fails. In many voices and instruments an overtone is louder than the fundamental, and small speakers and microphones can lose the fundamental almost entirely, yet you still hear the right note. A pitch detector has to find the repetition, not the loudest peak.
This one uses autocorrelation, which is easier to picture than it sounds. Take a short slice of the audio, about a twentieth of a second, slide a copy of it along itself, and measure how well the two line up at each delay. When the delay equals one cycle of the wave, the copy matches the original almost perfectly. That delay is the period, and one divided by the period is the frequency.
The specific method is the McLeod pitch method. It normalises the comparison so loud and quiet notes are treated alike, and it picks the first strong match rather than the strongest, which avoids reading a note an octave too low. It also returns a clarity score. Breaths, consonants and noise score low and are ignored, and a short median filter removes one-off jumps before the note is shown.
From frequency to note name, octave and cents
Once the frequency is known, the note follows from a simple rule: every doubling of frequency is one octave, and each octave is split into 12 equal semitones. With A4 at 440 Hz, A5 is 880 Hz and A3 is 220 Hz. The detector rounds to the nearest semitone for the note name and reports the remainder in cents, where 100 cents make one semitone.
Octave numbers follow scientific pitch notation, where middle C on the piano is C4 and each octave number starts on C. So the B just below middle C is B3, not B4. Black-key notes have two names, and the detector shows both: C♯4 is the same key as D♭4.
The MIDI number is the same note as a single number, with middle C at 60 and A4 at 69. It is handy if you are going to type the note into a DAW or a piano roll.
Singing practice with a pitch detector
Start with long tones. Sing one comfortable note on “ah” for four or five seconds and watch the cents. A steady line on the trail means steady support; a line that sags at the end usually means the breath ran out before the note did.
Next, try sirens: glide slowly from low to high and back. The trail shows where your voice jumps or thins out, often around the break between registers. Practising that area slowly smooths the join over time.
For accuracy, play a reference note on the Online Tuner, then sing it and compare. Aim to land within the narrow centre band without sliding into it. Then try intervals: sing a note, then a fifth above it, and check that the second note is exactly seven semitones higher.
Vibrato shows up as a regular wave around the note. A wave that is centred on the note is fine; one that sits mostly above or below it means the average pitch is sharp or flat.
Finding your vocal range
Press Start listening and begin in the middle of your voice. Glide down slowly until the note gets weak, then glide up to the highest note you can sing without strain. The range tracker only counts notes you hold for about a fifth of a second, so brief cracks and squeaks do not stretch the result. Press Reset range to start over.
Once you have covered ten semitones or more, the detector suggests the closest voice type by comparing the middle of your range with these approximate classical ranges.
Soprano: C4 to C6, about 262 Hz to 1,047 Hz.
Mezzo-soprano: A3 to A5, about 220 Hz to 880 Hz.
Alto (contralto): F3 to F5, about 175 Hz to 698 Hz.
Tenor: C3 to C5, about 131 Hz to 523 Hz.
Baritone: A2 to A4, about 110 Hz to 440 Hz.
Bass: E2 to E4, about 82 Hz to 330 Hz.
Treat the hint as a starting point, not a label. Voice teachers classify voices by tone colour, where the voice sits most comfortably and where the registers change, not by range alone. Choir parts also use narrower ranges than these. An untrained voice often gains several notes at both ends with practice.
Limits: polyphony, noise and very low notes
The detector follows one pitch at a time. When two or more notes sound together, as in a chord, a duet or a full song, the repeating pattern belongs to no single note, and the reading jumps or settles on the wrong one. For songs with several instruments, use MP3 to MIDI, which is built for mixes, and the Chord Finder for chords.
Background sound competes with your voice. Music, television and other people talking can push the reading around or produce stray notes. Reverb in a large room keeps old notes ringing after you change pitch. Getting closer to the microphone is the simplest fix.
Very low notes are slow and hard to read. A note around 50 Hz repeats only a couple of times in each slice of audio, and small microphones pick up little of it. Expect the needle to settle more slowly on deep bass voices and bass instruments.
Breathy, very short or very quiet notes may not be shown at all, because their clarity score stays below the threshold. That is deliberate: it is better to show nothing than a wrong note.
What note is this? Finding notes in a recording
Switch to Audio file and choose a recording. The detector reads it from start to finish, draws the pitch over time, and cuts it into notes wherever the pitch moves to a new semitone and stays there. Each note in the list shows when it starts, how long it lasts, its frequency and how far it sits from the equal-tempered note.
Click a point on the chart to play from there, or press the speaker next to any note to hear its reference pitch. When you are happy, export the notes as a .mid file for your DAW.
For a full conversion with quantizing, key snapping and sheet music, open the same file in Voice to MIDI. To find the key of the melody, try the Key Finder.