How this test works
Press Start listening and allow microphone access. The live audio feeds one Web Audio analyser, and on each animation frame the page reads the latest 2,048-sample window. From those same samples it takes the RMS level and a dominant fundamental estimate: frequency, nearest note, cents offset and clarity. It asks the browser to turn echo cancellation, noise suppression and auto gain control off, because noise suppression is built to remove steady tones and a held note is exactly that. Browsers do not always comply, so the lines below the controls show what the browser reports it applied, along with the audio graph's sample rate.
The estimator is a cumulative-mean normalized difference function from the YIN family. It compares the first 1,024 samples of the window with shifted copies of itself and normalizes the squared differences by their running mean. It takes the shortest lag whose score falls below 0.20 and follows it down to its local minimum; taking the deepest dip instead is the classic octave error, because the octave below a note repeats too. Before interpolating between samples, it checks whether three times or twice that lag matches substantially better, which catches a strong harmonic posing as the note. Clarity is one minus the chosen score, so 0.80 is the acceptance floor. The frequency maps onto A4 = 440 Hz and 12-tone equal temperament, named with sharps. Nothing reaches the readout until 5 consecutive windows agree within 50 cents of the first, so a scrape, a door or the start of a word does not flash a note; the 50 cents is an agreement rule between windows, and the reading itself is much finer.
The main limit is octave error: the pattern that repeats most strongly is not always the note you meant. Above the 1,600 Hz ceiling the page does not refuse, so a 1,900 Hz whistle reads about an octave down, and a 4 kHz tone can read near a third of its frequency with full clarity. Two sounds at similar strength are measured as their mixture, which can repeat at a lower note than either one. The page gives no tuning verdict; it shows where the note falls on the grid and leaves the call to you.
Frequently asked questions
How accurate are the note and the cents figure?
Accurate enough to tune by when clarity is high. Clarity is the figure to watch: a reading just above the 0.80 floor can be a semitone out or worse, so play the note again if clarity is near the floor or the note name looks wrong. Cents are shown as whole numbers. At graph rates of 8 or 16 kHz, which some hands-free Bluetooth headsets use, readings near the top of the band are much less precise than at ordinary sound-card rates; the graph rate is printed below the controls. With a real instrument in a real room, the microphone, browser processing, reflections and the note's decay matter more than the arithmetic, so read the cents with a couple of cents of slack and watch which way the figure moves as you turn a peg or move a slide.
Why does it say “no steady pitch”, or show no note?
Each refusal names the check that failed. At or below the −60 dBFS silence floor there is nothing to estimate. Sound that is not periodic enough to reach the 0.80 clarity floor gets “no steady pitch”, which is what breath, key clicks, traffic and a note that has died away all produce. An estimate outside 60–1600 Hz is refused outright, and so is a note whose period is too long for the analysis window, such as a bass guitar's low E; on a device with a raised floor, the message names the raised one. Very low rumble can get “no steady pitch” instead. Even an accepted signal needs 5 consecutive windows agreeing within 50 cents of the first before a note appears, and a refusal or a sudden change clears the note. If Input says blocked or unavailable, check microphone access first.
Which notes fit the 60–1600 Hz band?
It covers every string of a standard guitar (low E at 82.4 Hz), a five-string bass or baritone guitar's low B at 61.7 Hz, and cello C at 65.4 Hz, up to G6 at 1,568 Hz. A bass guitar's or double bass's low E at 41.2 Hz sits below the 60 Hz floor and is refused. Above a graph rate of about 61 kHz, as on a USB interface, an external DAC or a device set to studio quality, the 1,024-sample comparison window raises the floor, and the page prints that session's floor below the controls. A tone just under the 1,600 Hz ceiling can interpolate slightly past it and be refused, while notes well above the ceiling are not refused and can read as a lower note. Chords and other mixtures can read as a much lower note than anything being played, and a second sound at a fifth of the main note's amplitude still pulls the reading by around ten cents, so play one note at a time.
Can it play a reference tone to tune against?
No. This page listens and plays no sound. For a reference, open the
Tone Generator in a second tab and play A4 at 440 Hz; it states its output level and has a hearing-safety warning beside the controls. Stop the reference before you measure your instrument here, or the detector will measure the mixture of both sounds.
What reference pitch and temperament does it use?
A4 = 440 Hz and 12-tone equal temperament, the twelve equally spaced semitones a piano is tuned to. Cents show the distance from the nearest note on that grid: positive is sharp and negative is flat. The reference is fixed, so an orchestra tuning to A = 442 Hz reads about +8 cents at A, and every other note with it. An ensemble at A = 415 Hz is about a semitone below this page's A4, so the readout names the nearest note on its own grid. Just intonation, historical temperaments and the stretched octaves of a real piano can differ from this grid too. The page reports where a note sits against one stated reference, and the judgment stays with you.
