Voice Test - Free Online

How to read the voice test result

Start with measurement quality rather than the presentation cue. A result based on many clear voiced frames is more useful than one calculated from a whisper, clipped audio, or a noisy room. Confidence describes whether the browser captured enough consistent signal for the summary; it is not confidence about identity, personality, health, or how every listener will react.

Next, read median pitch together with the typical P10–P90 span. The median is the central fundamental frequency in the usable frames, while the span keeps ordinary sentence movement visible without letting one brief extreme dominate the report. Pitch movement and the limited spectral-balance proxy add context, but neither replaces careful listening or direct measurements of articulation, formants, resonance, or vocal quality.

The broad presentation reference combines only the acoustic cues listed on the result card. It is best treated as a way to compare recordings made under similar conditions. Voices and listeners do not fit into clean acoustic boxes, and substantial overlap is normal, so the result must never be used to assign a person a gender, sex, identity, or social category.

Make repeated voice tests comparable

A useful baseline depends on controlling the task. Read the same passage at a natural pace, use the same phone or computer, stay in the same room, and keep the microphone about a hand’s length away. Record at a similar time of day if you are tracking practice. These details reduce variation caused by vowels, emphasis, automatic microphone processing, room reflections, and fatigue.

Do not repeat the test until a preferred label appears. Instead, make two or three ordinary recordings and look for the range in which the measurements usually settle. A gradual trend across comparable sessions is more informative than the highest or lowest result from one day. Write down the passage and setup if you want to keep a private log, because this site deliberately does not create an account or upload result history.

• Use a fixed passage long enough to contain several voiced sentences.
• Speak naturally rather than forcing pitch upward or downward.
• Keep device, room, distance, and input settings as consistent as practical.
• Compare median, range, movement, and confidence together.

Why a voice test changes from one recording to another

Fundamental frequency is dynamic. Questions often rise, emphasis creates local peaks, relaxed endings may fall, and different vowels give the detector different amounts of stable periodic sound. Language, dialect, emotional state, intended audience, speaking volume, posture, and whether you are reading or improvising can all move the summary without indicating a permanent change in your voice.

Recording conditions matter too. A laptop may apply different echo cancellation or gain than a phone, and moving close to the microphone can alter both the signal level and spectral proxy. Fans, music, another speaker, reverberation, and Bluetooth processing may reduce stable pitch coverage. When a result is surprising, improve the recording conditions and repeat the same task before interpreting the number.

What this online voice test can and cannot do

The tool can provide an immediate, private snapshot for personal voice practice, recording setup checks, and educational exploration of pitch. It can help you notice that two readings differ and show which measured cues contributed. It cannot explain the cause of a difference, diagnose a voice condition, authenticate a speaker, identify an individual, or predict how a particular audience will perceive someone.

Never use a recording-level acoustic reference for employment, education, housing, credit, insurance, healthcare, access, moderation, or relationship decisions. Record only yourself or someone who has clearly agreed. If speaking causes pain, a sudden change persists, or hoarseness continues, stop testing and seek advice from an appropriately qualified clinician rather than relying on a browser measurement.

Local processing and technical limits

After microphone permission is granted, JavaScript collects a short mono signal in browser memory. A YIN-based detector estimates fundamental frequency in frames, rejects unreliable or unvoiced portions, and calculates robust summary statistics. The page does not send the recording, waveform, pitch frames, extracted acoustic features, or displayed result to a server. Closing or reloading the page clears the current in-memory session.

Browser pitch estimation is useful but not laboratory instrumentation. Very low fry, breathy or creaky sounds, overlapping speakers, music, and weak microphone levels can produce missing or octave-shifted estimates. The spectral-balance value is a signal-energy proxy affected by the device and room; it is not direct formant, anatomical, placement, breathiness, or clinical resonance analysis.