How HumMatch measures voices, and where we draw the line
Every percent on HumMatch is a computation you can interrogate. This page explains exactly how we measure your voice, where song ranges come from, what a hum can and cannot tell us, and how to correct us when we are wrong.
How your Vocal ID is measured
Pitch detection runs on your device
When you hum, your browser analyzes the audio locally and detects the pitch of each note you hold. We read fundamentals, not vibes: the detected notes, how long you held them, and how steady the pitch stayed.
Sessions merge into a best-of record
One hum is a snapshot. Your Vocal ID keeps the best of every session: reach a higher note on Tuesday than you did on Monday and that note is stored permanently. Your range only ever grows from evidence, never from a guess.
Confidence measures us, not you
The confidence score is not a grade on your singing. It measures how well we know your voice, and it maxes out around five clean sessions. Below 50 we say Calibrating, because that is what it is.
Weak captures never widen your range
A noisy room, a wavering pitch, or a note cut short produces a low-confidence capture. Those captures never extend your stored range. We would rather show you a narrower range that is real than a wider one built on a detection artifact.
Where song ranges come from
Every song range on HumMatch carries a label telling you how we know it. There are three tiers, and we show you which one you are looking at instead of pretending they are all equal.
Manually researched
A person verified this range against published sheet music, with a citation. This is the trusted tier: it earned that trust song by song, and it is never overridden by automation.
Estimated
Derived from data about the artist rather than the specific melody. Estimates are honest starting points, and they stay labeled as estimates everywhere they appear. An estimated range is never presented as verified fact.
Range pending
No trustworthy range yet. We say so, and the song is excluded from range claims until real data exists. An empty label beats an invented number.
The 19-semitone honesty cap
Many estimated ranges were inherited from whole artists: bands with multiple lead singers and stacked harmonies produce spans no single melody actually demands. A single sung melody wider than about 19 semitones is implausible, so when an estimated span reaches that line, scoring uses a 19-semitone window centered on the estimate. The page still shows the true estimated span. Only the math is capped, and researched ranges are never capped at all.
Why we label instead of hiding
We could hide the weak data and the site would look cleaner. But you are choosing songs to sing in front of people. You deserve to know whether a range came from sheet music or from an inference, because you are the one holding the microphone when the bridge arrives.
What we do NOT claim
A hum measures comfortable range, not your ceiling
Humming finds the range you can produce cleanly and comfortably. It does not measure your belt ceiling, the highest note you can hit once with full support and adrenaline. Trained singers usually have edges beyond what a relaxed hum shows, which is exactly why you can correct your own range below.
Tessitura and max range are different things
Your max range is every note you can touch. Your tessitura is where your voice actually lives: the zone where you can sing all night without strain. A song can sit inside your max range and still exhaust you if it parks above your comfortable center. Our scoring weighs where a song sits relative to your center, not just whether the edges fit.
A hum does not capture timbre
Pitch detection reads frequency, not character. Whether your voice is smoky, bright, airy, or brassy is not something three hummed notes can tell us, so we do not pretend it can.
How to correct us
A measurement platform that cannot be corrected is just a platform with opinions. Two correction paths, both built in:
Flag a song range
If a song page shows a range you know is wrong, use the range flag on that page to tell us. Flags land in a review queue and verified corrections update the catalog with a citation.
Set your own range
You may know your trained range better than early calibration does. In My HQ you can set your low and high notes yourself. Your setting is stored as its own layer, clearly labeled, and it never overwrites what we measured. Scoring uses your numbers; the measured record stays intact underneath, and every clean hum keeps confirming your real edges.
Methodology FAQ
How does HumMatch measure my vocal range?
Pitch detection runs on your device while you hum. Each session detects the notes you held and how steady they were, and your Vocal ID keeps the best-of record across all your sessions. Low-quality captures never widen your stored range.
Can three hummed notes really measure a voice?
Three notes start the measurement, they do not finish it. Confidence grows across sessions and maxes out around five clean ones. Until then your profile says Calibrating, and you can correct your own range manually at any time.
Where do HumMatch song ranges come from?
Three labeled tiers: manually researched ranges verified against published sheet music with citations, estimated ranges derived from artist data, and range pending when no trustworthy data exists yet. The label is always shown.
What is the 19-semitone cap?
Estimated ranges inherited from whole artists can be wider than any single melody. When an estimated span reaches 19 semitones, scoring uses a 19-semitone window centered on the estimate. Displayed ranges are never altered and researched ranges are never capped.
What if HumMatch got my range wrong?
Set your own low and high notes in My HQ. Your correction is stored as a separate labeled layer, scoring uses it immediately, and the measured record underneath is never overwritten. Song ranges can be flagged on each song page.
Measure it yourself
The fastest way to judge our methodology is to run it. Hum three notes and look at what comes back.