Understanding your vocal frequency in Hz

A number in hertz is the most concrete thing you can know about your voice. It is also the easiest thing to misread. This guide covers what the number is, how a detector produces it, what published population data actually says, and how to take two measurements that can honestly be compared.

What a hertz measures

One hertz is one cycle per second. During a voiced sound, your vocal folds open and close in a repeating pattern. The rate of that repetition is the fundamental frequency, usually written F0. At 120 Hz the folds complete 120 open-close cycles every second.

Pitch is the perception that goes with F0, and the two are related but not identical. Loudness, timbre, the listener's hearing and even the surrounding sentence all shift perceived pitch a little. F0 is what a machine measures. Pitch is what a person hears.

How a detector turns sound into a number

A pitch detector does not measure your whole voice at once. It takes a short slice of audio, tests how periodic that slice is, and estimates the repetition rate. In Basaltone that slice is 2048 samples, which lands near 43 milliseconds on a typical 48 kHz microphone, and the analysis runs 30 times per second.

Two things matter more than the algorithm.

The first is rejection. Not every slice deserves a number. Basaltone scores how periodic each slice is and discards anything below 0.75 on that scale, which is what breathy onsets, consonants and room noise look like. It also drops anything outside 60 to 500 Hz, and treats a jump larger than 5 semitones as suspect until several consecutive readings agree, because a lone octave-sized jump is almost always a detection error rather than a real vocal move.

The second is smoothing. A raw frame-by-frame reading jitters far more than your voice does. Accepted readings pass through a short median filter, then an exponential smoother that works in semitones rather than hertz, because equal semitone steps sound equally large to the ear while equal hertz steps do not.

One deliberate choice runs against normal audio practice: Basaltone asks the browser to disable echo cancellation, noise suppression and automatic gain control. Those features make calls sound better by reshaping the signal, which is exactly what a measurement must not do.

What the published numbers actually say

This is the part most pages get wrong, usually by quoting a range with no source and presenting it as a standard to hit.

The largest recent population study on the question, by Berg and colleagues in the Journal of Voice, measured 2,472 adults aged 40 to 79 in Germany. At conversational level it found a mean F0 of about 112 Hz for men and about 169 Hz for women.

The interesting result is not the averages. It is that the female mean sat several semitones below the values older textbooks report. Population norms move with the decade, the country, the recording method and the task the speaker was given.

So treat published ranges as a description of groups, not as a target for a person. A voice outside them has nothing wrong with it, and a voice inside them is not "correct". Your own first measurement is a far more useful reference than any population average, because it is the only number your later measurements can legitimately be compared against.

Taking a measurement you can compare

A single reading tells you very little. Two comparable readings tell you a lot. Most of the work is holding everything except your voice constant.

Two habits matter more than any equipment.

Speak long enough. Basaltone only counts a session toward your progression at 60 seconds or more, because shorter samples are dominated by whichever sentence you happened to say.

Compare medians, not extremes. The lowest note you can briefly produce is not your speaking pitch, and it is not stable from day to day. Basaltone reports a session's low and high as the 10th and 90th percentile of your readings rather than the raw minimum and maximum, so one creaky frame cannot define your range.

What counts as a real change

Hertz are not perceptually even. The same 5 Hz means something different at the top and the bottom of a voice, which is why the app reasons in semitones internally and splits your journey into steps of 0.75 semitone.

Read that table in both directions. A 6 Hz change at 165 Hz is smaller than it looks. A 6 Hz change at 100 Hz is more than a full step. It also means a two or three hertz difference between two sessions is not information, it is measurement scatter.

Because single sessions scatter, Basaltone never promotes you on one result. It takes the median across your last five qualifying sessions and moves at most one step per session, in either direction. How long voice training takes covers how to read that curve over weeks.

The limit of a single number

F0 is one dimension of a voice, and not the one listeners respond to most strongly. It does not measure resonance, formants, vocal weight, articulation, or the rhythm and intonation patterns of your speech. Two people with nearly identical hertz can sound completely different, which is the subject of pitch, timbre, resonance and formants.

It also cannot determine identity, health, or how any particular listener will perceive you. A pitch detector is a tape measure, not a verdict. If a number ever pushes you to force your voice lower, it has stopped being useful: see vocal fatigue, tension and safe practice for the signals that matter more than any measurement, and deeper voice training for how a measurement fits into actual practice.

Basaltone analyzes audio locally in your browser during the public test and stores no recording from it.