Detection, explained
What is a voice gender detector?
A voice gender detector is a quick check on how a voice recording may come across acoustically. You record a short sample; the detector measures fundamental frequency (pitch), how that pitch moves through the sample, and a spectral-balance proxy, then places the recording on a broad masculine-leaning to feminine-leaning presentation reference with a visible confidence label.
Everything runs in your browser. Press record, speak for 10–15 seconds, and a YIN pitch-detection algorithm inspects stable voiced frames locally. The reading is a snapshot of this recording — one performance, in one room, on one microphone — not a determination of who you are.
Detector vs analyzer — and why this is never identity detection
People search for a voice gender detector when they want a fast answer, and a voice gender analyzer when they want the full explainable breakdown. On this site both run the same local engine: the detector on this page gives the quick presentation reading, while the homepage analyzer and the voice analyzer tool unpack the same measured cues in more depth.
The wording matters because of what these tools cannot do. Gender identity and biological sex are not acoustic measurements, so no microphone test can detect them. What a detector can honestly do is detect cues in a signal: the median pitch of this recording, its P10–P90 range, its intonation movement, and a brightness proxy. Detection here means detecting what is in the audio — never detecting who a person is.
What this detector reads in a recording
- Pitch — the median fundamental frequency of your recording in Hertz, and the largest single cue in the presentation reference.
- Pitch range — the P10–P90 span, showing how far the pitch moved through the sample.
- Intonation — pitch movement in semitones, describing how expressive the delivery is.
- Spectral balance — a browser-calculated brightness proxy from the recorded signal, not a formant or clinical resonance measurement.
- Recording quality — voiced seconds, estimated background noise, and stability, which combine into the confidence label.
How to run a voice check
- Find a quiet room and keep your microphone about a hand's length away.
- Press the record button and read the fixed passage, or speak naturally for 10–15 seconds.
- Stop, and the detector shows the presentation reading, score, confidence, and the cues that shaped it.
- Repeat with the same passage and setup when you want readings you can compare.
Read the result without over-reading it
Check recording confidence first. It summarizes whether the browser found enough clear voiced material for a stable calculation; it is not confidence about identity. Then read the score together with the contributing cues rather than as a standalone number.
Expect overlap. Voices sit across a broad spectrum, and most recordings land in the wide middle band. A position on the reference describes acoustic cues in one sample; listener perception still depends on resonance, articulation, language, accent, and cultural context — none of which this detector claims to measure.
Limits and responsible use
The detector estimates fundamental frequency and a limited spectral-energy proxy. It does not directly measure formants, vocal-tract anatomy, placement, breathiness, articulation, accent, emotion, or clinical resonance. Noise, clipping, creaky or breathy sound, overlapping speakers, and strong room reflections can reduce accuracy or cause occasional octave errors.
Record only yourself or someone who has clearly agreed. Do not use the reading for employment, education, housing, credit, insurance, healthcare, access, authentication, surveillance, or judgments about another person. This is an educational browser measurement, not a diagnostic or eligibility tool.