Device Bench

Device test · Audio

Vocal Range Test

Sing or hum down as far as you comfortably can, then up, and this page records the lowest and highest note it was willing to stand behind: the two note names, the span between them in semitones and octaves, and the clarity and the number of separate times each end was held. It refuses rather than guesses — silence, sound with no steady period, and anything landing outside 60–1600 Hz each get a refusal that says which measurement failed, and a refused window never touches the range. Audio stays on this device and nothing is uploaded.

This page listens and plays no sound. If you want a reference to step down and up against, that is the Tone Generator's job — it states its level and duration and carries the hearing-safety copy that belongs with them. Nothing here needs you to match a given pitch, so working without one is fine.

Inputnot requested Lowest— Highest— Span— Singing now— Level—

Idle. Press start, allow the microphone, then sing or hum one steady note at a time — an end of the range fills in once 5 consecutive measurement windows agree with each other.

EndFrequencyCentsClarityTimes held
Lowest — — — —
Highest — — — —

Both ends hold since you started or last cleared — they are the extremes of the run, not of the last second — and “times held” counts the separate occasions a note cleared the gate, never animation frames, because a frame count would be a figure about your display rather than about your singing.

Processing state is reported once the microphone is open.

Sample rate is reported once the microphone is open.

Level is RMS in dBFS — relative to digital full scale, not loudness. A browser has no calibrated reference, so nothing here is a sound pressure measurement, and clarity is a periodicity statistic rather than a mark for how a note sounds.

How this test works

When you press start, the browser asks for microphone access and hands this page a live audio stream. The stream feeds one Web Audio analyser, and on every animation frame the page reads the most recent 2,048-sample window and derives two things from that single window: the RMS level in dBFS, which is what the silence gate tests, and a dominant fundamental estimate — a frequency in hertz plus the clarity figure that backs it. Both readings come from the same samples, so the level you can see and the note derived from it are two views of one measurement rather than two measurements free to disagree. The estimator is the one Pitch Detector uses, in the same shared file rather than in a second copy of it, because that code carries several fixes for the ways this family of detector goes wrong and a copy would quietly un-fix them. Unprocessed audio is requested for every capture, because noise suppression is built to remove the steady tones it reads as background and a held sung note is exactly that; the line under the controls reports what the browser says it actually applied, which is not always what was asked for.

A frequency becomes a note by the standard mapping — A4 = 440 Hz, 12-tone equal temperament, sharps only, since D♯ and E♭ are one key and one frequency. Nothing is recorded until 5 consecutive windows have agreed within 50 cents of each other at a clarity of at least 0.80, so a scrape, a door, a consonant or the moment a slide passes through a note never becomes an end of the range. When a run does clear that gate, the page compares its frequency against the two ends it is holding and moves whichever one it beats; the span is then plain arithmetic on the two note names above — the semitone distance between them on that same grid, and that figure divided by twelve for the octaves — so the number you read and the names you read cannot disagree. Each end also carries the clarity it was accepted at and the number of separate runs that landed on that note, counted per acceptance rather than per frame for the same reason the clipping counter on the microphone test counts inspected windows: consecutive frames overlap or skip depending on the display, so a frame total would be a figure the instrument cannot stand behind. Refusals clear the live field and leave both ends exactly as they were.

The limits are worth stating plainly, and the first one is the biggest. This is a measurement of one run, not a verdict about your voice — it reports the notes one microphone accepted in one session, and it does not name a voice type, does not score anything, and cannot know whether you sang the lowest or highest note available to you. The failure class of this whole detector family is the octave error, and both ends of a sung range sit near it: a voice is harmonically rich, so its second partial can arrive stronger than its fundamental, which is the shape the detector's depth check exists to catch, and above the ceiling there is no refusal left and a very high note is reported as some whole-number division of itself. The floor is set by the analysis window rather than by the band, so at studio sample rates it rises and the page prints the figure it is actually using. Everything before the arithmetic matters more than the arithmetic: a microphone's response is not flat and small capsules give up on low notes first, browsers apply processing this page can request off but cannot switch off, and a room adds reflections and noise of its own. Read the two ends together with the evidence beside them, and treat an end reached once at a clarity near the floor as a reason to sing it again.

Frequently asked questions

What is this actually measuring?
The lowest and highest note this run recorded — not the range of your voice. Those are two different claims and only the first one is a measurement. The page listens for one note at a time, and a note is recorded only after 5 consecutive analysis windows agree with each other within 50 cents at a clarity of at least 0.80; the lowest and the highest note that cleared that gate become the two ends, and the span is the distance between the two note names printed above, in semitones and in octaves. Everything else is deliberately left out. Whether you actually reached the bottom or the top of what you can sing, whether a small microphone reproduced the fundamental of a very low note, and whether something in the room contributed a note you did not sing are all things this page cannot see. So read the two ends as what one microphone accepted in one session, and read the clarity and the times-held figures beside them as the evidence for each: an end reached once at a clarity just above the floor is weak evidence, and singing it again is the cheapest way to strengthen it or to correct it.
Why did it not record the note I just sang?
There are four honest answers. The first three are refusals and the status line names whichever one applied; the fourth is not a refusal at all, and shows as the range simply not moving. Below the −60 dBFS silence floor there is no signal to estimate at all. When nothing in the band is periodic enough — clarity below 0.80 — it reports no steady pitch, which is what breath, consonants, a slide between two notes and a note that has already died away all produce. An estimate that lands outside 60–1600 Hz is refused rather than reported at the edge, and so is a note whose period is simply too long for the analysis window to hold: that is how a very low bass fundamental is answered, and how everything under the raised floor is answered on a device that lifts it. And a note can simply not have been held long enough — the readout shows what is being sung now, but only a run of 5 agreeing windows reaches the range. A refused window never touches the range, which is the whole point of refusing instead of guessing: one squeak the analyser could not stand behind is not allowed to become the top of a range.
Why does it not tell me my voice type?
Because a range does not determine one. Voice type is a judgment about where a voice sits comfortably, how it changes across its registers and what it sounds like doing so — tessitura, timbre and the passaggio — and none of those is a pair of extreme notes. Two singers with the same lowest and highest note are routinely classified differently, so a page that read a label off two numbers would be inventing the part it could not measure. This page therefore does not name a voice type, does not score anything, and does not rank anybody against anybody; it prints the notes it accepted and the evidence behind them, and the judgment stays with you. This page is a music tool and nothing else: it makes no claim about vocal health, carries no medical meaning, and cannot tell you whether a note is safe for you to sing. If singing hurts, stop — but take that to a qualified professional rather than to a web page, because nothing here can see or assess anything about your body.
Can it play a note for me to sing against?
No, and that is deliberate: this page listens and plays no sound. Playing a tone means owning the hearing-safety obligations that come with it — a stated level, a stated duration, and a warning next to the control — and the Tone Generator already carries all three and does that job properly. Open it in a second tab, step down or up from a note you are sure of, and let this page record what it hears. Working with no reference at all is also fine, because nothing about the measurement needs you to match a given pitch: start somewhere comfortable, slide down until the readout stops accepting anything, then slide up the same way.
How accurate are the two ends?
The two ends fail differently, and the bottom one is harder. At the low end a small microphone is usually the limit rather than the arithmetic: phone and laptop capsules roll off well before 100 Hz, so the fundamental of a low note can arrive weaker than its own second harmonic. That is the shape the detector guards against by checking whether twice or three times the chosen period fits the signal better — without that check a low note prints an octave high with its cents figure near zero, a wrong note wearing a perfect tuning. At the top end the failure class is the octave error in the other direction: above the ceiling the page has no refusal to fall back on and reports some whole-number division of what was sung, so a whistled or falsetto note up there can be recorded an octave or more below where it belongs. In the middle of the band, on a clean sustained note at an ordinary sound-card rate, the frequency lands well under a cent — but that figure belongs to the sample rate rather than to the page, and a hands-free Bluetooth headset drops the whole audio stack to 8 or 16 kHz, which is worth tens of cents near the top of the band. The graph rate the page is actually running at is printed under the controls.

Related tests and guides