In short: Our tools give useful estimates, not laboratory measurements. Music-theory tools follow fixed rules and are exact unless there is a bug. Microphone tools depend on your voice, microphone, device, browser and room, so treat single readings with care and compare several.
This page explains what affects the results, what the tools cannot tell you and how to get repeatable readings. How the tools are built and tested is in our testing methodology.
Accuracy, precision and repeatability
Accuracy is how close a result is to the true value. Precision is how finely it is displayed. Repeatability is whether similar conditions give similar results. A frequency shown to one decimal place is not proof that a phone microphone captured it that exactly, and a microphone can be very repeatable while its own processing still shifts the result.
What testing shows, and what it doesn’t
With clean generated tones, the Pitch Detector read 220 Hz as 220.0 Hz, a tone 20 cents above A4 as 445.1 Hz (+20 cents), a low E1 (41.2 Hz) as 41.2 Hz and a C7 within 1 cent. The Note Identifier found all 16 notes of a test melody with vibrato, in order. These tests prove the maths is right; they do not cover every real voice, microphone or browser. Safari, Firefox, physical Android phones and real singers have not yet been formally tested, which is why real-world results are called estimates.
Microphone and device differences
Built-in, headset and USB microphones differ in sensitivity, frequency response, gain and noise. Phones and laptops may also apply automatic gain, noise suppression and echo cancellation before the page receives the sound. Our main pitch tools ask the browser to switch echo cancellation off, but some devices still process the signal. Browser microphone access works through the getUserMedia() API, which always asks for your permission first. A more expensive microphone is not automatically better: a consistent, clear signal matters most. If a result looks wrong, try the same test on another device.
Background noise, distance and level
Fans, traffic, other voices, music and hum can mask your voice and make a detected note jump or vanish. Too far from the microphone and your voice is weak; too close or too loud and the input clips. Test in a quiet room at a steady, comfortable distance.
How you sing changes the reading
Fast slides, wobbly or very breathy notes, whispering, growls and forced notes at the edge of your range are harder to read. The Voice Range Test only accepts a note after about 0.8 seconds of steady singing for this reason. Tiredness, illness and heavy recent voice use also change your voice from day to day. Do not push through pain to get a better number.
Octave errors and harmonics
A sung note contains a fundamental frequency plus harmonics above it. A pitch detector can sometimes lock onto a harmonic, giving a reading about an octave off. Our tools use the YIN method with a clarity check and a median of three readings to reduce this, but noise, vocal fry or distortion can still cause it. Repeat an unexpected note before you trust it.
A range result is a snapshot
Your lowest and highest detected notes show one session. They are not your permanent limits, your comfortable singing range, your tessitura or your singing ability. Reaching a note once is different from singing it comfortably and repeatedly.
Voice-type results are estimates
The Voice Type Test compares your range with the typical ranges on our vocal range chart. Ranges overlap, so two singers with the same range can be different voice types. Teachers also consider timbre, vocal weight, register changes, training and repertoire. When two types match equally, the tool says you sit between them. More in types of vocal ranges.
What the voice tools cannot diagnose
Pitch, range and stability data are not a clinical examination. The tools cannot detect vocal fold injury, nodules, laryngitis, reflux, neurological or breathing conditions, or hearing loss. For general guidance see the NIDCD page on taking care of your voice, and see a qualified health professional about pain, lasting hoarseness or sudden changes. The disclaimer has more.
Tools that play sound
The Tone Generator and Audio Frequency Test avoid microphone problems, but what you hear depends on your speakers or headphones, volume, equalisation and hearing. Small speakers reproduce very low tones poorly. A tone you can’t hear may be missing from your equipment rather than your ears, so these tools are not hearing tests.
The Online Metronome is fine for practice, but browser audio passes through software buffers, and wireless headphones can add a noticeable delay. Scores in Interval Ear Training describe your answers in that exercise only and improve as you get used to the format.
Music-theory and suggestion tools
Calculators such as the Vocal Range Calculator follow fixed rules: 12 semitones make an octave, frequency doubles each octave, A4 = 440 Hz, and 120 BPM is 0.5 seconds per beat. They can only be wrong through a bug, so please report any output that breaks these rules. The Song Key Finder gives its best match, but borrowed chords, modulation and modal music can make more than one key reasonable. Routines from the Vocal Warm-Up Generator are general and cannot replace a teacher who hears you.
How to get repeatable results
- Use a quiet room.
- Keep the same device, microphone and distance when you compare results over time.
- Hold each note steadily for a second or two instead of sliding.
- Warm up gently; don’t force the extremes.
- Repeat any reading that looks odd, and trust the pattern over a single result.
- If results stay implausible, try another device or browser.
When an online tool is the wrong choice
Use professional equipment or a qualified person when you need a medical or clinical voice assessment, a hearing test, calibrated acoustic measurements, a formal voice classification or legal documentation.
Report a strange result
Tell us the tool, your device and browser, the microphone you used, and what you expected versus what happened: contact@voicerangetest.com or the contact page. We handle reports as described in the editorial guidelines. Privacy questions are answered in the privacy policy and the FAQ.
Last reviewed: 27 September 2026