Your Range Is Not Your Voice: The Hidden Math of Matching a Singer to a Song

Ask someone about their singing voice and, if they know anything at all, they will give you two notes: "I can get from E3 to C5." Lowest note, highest note. Done.

Here is the uncomfortable truth from the science of voice: those two numbers are close to useless for predicting whether that person can actually sing a given song well. Two singers with identical ranges can have completely different experiences with the same chorus: one floats through it, one white-knuckles every phrase. The interesting mathematics lives in everything the two-number summary throws away.

First, turn notes into numbers

To do math on voices, we need pitch as a number, and the cleanest scale is the MIDI note number: the numbering system music software has used for decades. Middle C is 60. Each semitone (one piano key) adds 1. Each octave adds 12. The A above middle C, the note orchestras tune to at 440 Hz, is MIDI 69.

Converting a frequency in hertz to a MIDI number takes one formula:

m = 69 + 12 · log₂(f / 440)

In plain English: take your frequency, ask "how many doublings away from 440 Hz is this?" (that is what the log does), multiply by 12 because an octave (one doubling) contains 12 semitones, and add it to 69, the number for 440 Hz itself. Sing an A3 at 220 Hz (exactly one halving of 440) and the formula gives 69 + 12 · (−1) = 57. Every octave is a clean step of 12.

This matters because human pitch perception is logarithmic. The distance from 220 Hz to 440 Hz sounds the same as the distance from 440 Hz to 880 Hz, even though one gap is 220 Hz wide and the other 440 Hz. MIDI numbers bake that into the scale, so "distance between notes" finally means what your ear thinks it means. From here on, a voice is a set of numbers on a line, and we can do real geometry with it.

Range vs. tessitura: where a voice lives

A famous vocalist gets credited with a four-octave range and it sounds superhuman: 48 semitones of territory. But watch what that voice does across an entire concert and a different picture appears: the overwhelming majority of sung notes fall inside a band barely one octave wide. The four octaves are real estate; the one octave is home.

Voice scientists call that home base the tessitura: the region where a voice sits comfortably, hour after hour, without strain. And the honest way to find it is statistical. Record everything a person sings, convert every pitch to a MIDI number, and pile the numbers into a histogram. You will not see a flat plateau from lowest note to highest. You will see a mountain: a dense peak where the voice lives, and long thin tails of notes that were touched once, at full commitment, on a good day.

The range is the distance between the two most extreme points of that mountain, the two least representative data points the voice ever produced. The tessitura is the fat middle of the distribution. If you want to know whether a song will feel comfortable, the middle is what matters.

Percentiles beat extremes

This is the same reason statisticians distrust minimums and maximums everywhere: extremes are noisy and unrepresentative. One shrieked note at a birthday party technically extends your "range" by three semitones. It moves your median (the note you are most often on) by nothing.

So instead of asking "what is the highest note you have ever hit?", the better question is: "what note are you above half the time?" (the median), and "where do the middle 50% of your notes fall?" (the band from the 25th to the 75th percentile). A singer whose pitch distribution has its median at E4, with the middle half of their singing between C4 and G4, is a fundamentally different instrument from a singer with the same E3–C5 range whose median sits at A3 (even though a range-only comparison says they are identical).

Songs have distributions too. A song is not "F3 to B4"; it is a melody that might spend 80% of its time between A3 and A4 with two brief excursions. A chorus that parks at A4 for four bars is a completely different demand than a chorus that touches A4 once on a passing note. When you match the singer's distribution against the song's distribution (mountain against mountain), you are finally comparing the things that actually collide on stage.

Timbre: why identical ranges sound nothing alike

Now for the dimension range cannot see at all. Two singers hit the exact same A3. One sounds warm and woody; the other sounds bright and cutting. Same pitch, same loudness. Unmistakably different voices. That difference is timbre, and it is measurable.

When you sing a note, you are not producing one frequency. You produce the fundamental (the note itself) plus a whole ladder of overtones above it, and the way energy is distributed across that ladder is the fingerprint of your voice. The classic single-number summary is the spectral centroid: the center of mass of all that energy. Imagine the frequency spectrum as a physical object and ask where it balances. A voice with lots of shimmer in the high overtones balances high: we hear that as bright. A voice whose energy is concentrated near the fundamental balances low: we hear warmth, darkness, velvet.

This is why "you have the same range as this famous singer" so often disappoints. Matching pitch territory while ignoring where the spectral energy sits is like matching two paintings by their frame sizes. A song built to be delivered with a bright, cutting tone will fight a warm, dark voice even when every note is reachable. And vice versa.

Vibrato: the wobble is data too

Even the steadiest held note is not a flat line. Sustained singing oscillates gently around the target pitch (vibrato) and it turns out to be beautifully quantifiable with two numbers.

Rate, measured in hertz: how many times per second the pitch wobbles. Trained singers typically sit between 5 and 7 Hz; a 6.5 Hz vibrato means the pitch rises and falls six and a half times each second.

Extent, measured in cents: how far the wobble strays from center. A cent is 1/100 of a semitone: a unit fine enough to measure deviations no listener could name but every listener can feel. A typical vibrato might swing plus or minus 40 to 60 cents, meaning the pitch orbits within about half a semitone of its target. Under roughly 30 cents reads as a shimmer; past 100 cents it starts to read as a wobble.

Two voices with identical ranges and similar brightness can still differ here: one with a fast, narrow flutter, another with a slow, wide wave. Material that flatters one can expose the other. It is one more measurable axis on which "same range" voices turn out to be different instruments.

The frontier: voices as points in space

Everything so far reduces a voice to a handful of interpretable numbers: a median here, a centroid there. The modern approach goes further: represent an entire voice as a single point in a space with hundreds of dimensions, a voice embedding, where every dimension captures some learned aspect of vocal character.

You cannot picture a 256-dimensional space, but the intuition survives in two: think of every voice as a pin on a map, placed so that similar voices land near each other. Smoky low altos pin themselves into one neighborhood; bright agile tenors gather in another. Distance on the map means dissimilarity of voice.

To compare two voices, you measure the angle between them: a quantity called cosine similarity. Two voices pointing in nearly the same direction from the origin score close to 1: same neighborhood, same character. Unrelated voices score near 0. And this is where matching stops being arithmetic and becomes geometry: "who sounds like you" becomes "who is near you in space," a question with a precise, computable answer. If your voice sits close to a particular artist's, the songs built for that voice are statistically likely to fit yours. Not because your ranges match, but because everything else does.

The two-number summary was never the story

So a voice, properly measured, is not a pair of extremes. It is a pitch distribution with a median and a comfortable middle; a spectral balance point that sets its color; a vibrato rate and extent that shape its texture; and a position in a high-dimensional space that captures the character no single number can. Whether you can sing a song is a question about how all of those align with the song's own distribution and character. Which is why the karaoke classic that "fits your range" can still be a disaster, and the song you never considered can feel tailor-made.

Range is the least interesting fact about your voice. The mathematics of everything else is where the answer lives.

(This is the kind of analysis we obsess over at HumMatch, where humming three notes builds a measured profile of your voice and matches it against thousands of songs.)