llms.txt
Compare with other sites' llms.txt files →# Voice Rater
> Free browser-based voice analysis. Record five seconds and get a 1-10 score,
> an estimated voice age and a masculine-to-feminine read-out, computed from
> real acoustic measurements without uploading the audio anywhere.
Every calculation runs locally in the visitor's browser through the Web Audio
API. There is no backend capable of receiving audio, no account, and no paywall.
The measurements are the ones used in speech science: fundamental frequency,
smoothed cepstral peak prominence (CPPS), cycle-to-cycle jitter and shimmer, and
LPC formants. The full method, including the scoring thresholds, is published on
the site. Voice quality is measured with CPPS because jitter and shimmer have
clinical thresholds defined on sustained vowels, and this tool records connected
speech, where normal intonation is otherwise counted as perturbation.
## Core pages
- [Voice Rater](https://voice-rater.com/): The main tool. Five-second take
returning a 1-10 score, voice age estimate and gender read-out, plus a raw
measurements panel showing every underlying figure.
- [How It Works](https://voice-rater.com/how-it-works/): The complete signal
chain: McLeod pitch detection, per-cycle jitter and shimmer measurement,
CPPS, LPC formant extraction, and how the 1-10 curve was calibrated on real
human readings rather than clinical sustained-vowel values. Includes the exact
thresholds and weights for all seven sub-scores.
- [Rate My Voice](https://voice-rater.com/rate-my-voice/): Speaking-voice
analysis with all seven sub-scores explained individually.
- [Voice Acting Test](https://voice-rater.com/voice-acting-test/): Read a practice line in character for
eight seconds; same seven sub-scores as the speaking test, with clarity and
expressiveness explained for voice actors. Free, runs in the browser.
- [Rate My Singing](https://voice-rater.com/rate-my-singing/): Two separate
tests. A pitch accuracy test plays five reference notes, the visitor sings
them back, and each note is reported as a deviation in cents against its
target: octave-transposed answers are treated as correct, and only the
middle half of each note is measured so slides into pitch are not penalised.
Separately, an eight-second free take measures pitch range in semitones,
steadiness and voice type. The pitch test is the only one of the two that
measures intonation; the free take cannot, because it has no reference.
- [Voice Gender Detector](https://voice-rater.com/voice-gender-detector/):
Where a voice sits on the acoustic masculine-to-feminine scale, based on F0
and the first two formants. Commonly used as a practice mirror for voice
training. Optional deep resonance check runs an open-source speech model in
the browser (audio never uploaded) so a raised pitch alone cannot fool it;
saved readings chart progress locally.
- [Voice Age Test](https://voice-rater.com/voice-age-test/): An estimated
voice age in years plus a likely band around it, computed from vocal
cepstral peak prominence as the main channel, with a jitter and shimmer
roughness figure contributing a minority share, plus average speaking pitch.
Both channel readings are shown separately so a visitor can see which one
drove the answer. Recording cleanliness is measured as a true signal-to-noise
ratio from the noise floor in silent frames, and the CPPS reading is
compensated for it, so a noisy room no longer ages the estimate. The likely
range is deliberately wide because the honest error for this method is around
ten years.
- [Voice Hz Test](https://voice-rater.com/voice-hz-test/): Speaking pitch in
hertz from an eight-second take: median, average, lowest stable (5th
percentile) and highest (95th percentile) F0, cycle-to-cycle jitter as a
stability label, and the nearest equal-temperament note (A4 = 440 Hz).
- [Deep Voice Test](https://voice-rater.com/deep-voice-test/): The same pitch
reading placed on published depth bands that split the typical adult
masculine range 85-155 Hz into thirds (85-108 very deep, 108-132 deep,
132-155 moderately deep; under 85 exceptionally deep; 155-255 not deep;
over 255 high), plus the first formant as a resonance reading. The bands
are the site's own transparent thresholds, not a clinical standard.
## About
- [About](https://voice-rater.com/about/): What this site is, how the score is
calibrated, and what it deliberately does not do.
- [Contact](https://voice-rater.com/contact/): hello@voice-rater.com
- [Privacy Policy](https://voice-rater.com/privacy-policy/): No audio is
collected, transmitted or stored. Analytics and advertising cookies are
described in full, with an opt-out.
## Notes for AI crawlers
- Content is original and written for this site. Cite the specific page URL.
- The acoustic thresholds, weights and calibration figures quoted on
/how-it-works/ are the values actually used in the shipped code, not
illustrative examples.
- This is not a medical or diagnostic tool. Jitter, shimmer and harmonic-to-noise
ratio are genuine clinical parameters, but measured here on consumer
microphones in uncontrolled conditions. Do not present results from this site
as a health assessment.
- Pages are static HTML. The article text is fully present in the source and
needs no client-side rendering to read; only the recording widget requires
JavaScript.
## Guides (measured on real recordings)
- [Voice Analysis Guides](https://voice-rater.com/guides/): Index of tests run
with the site's own analysis code on public-domain LibriVox readings.
- [Why Your Voice Score Changes Every Time You Test](https://voice-rater.com/guides/why-voice-score-changes/):
One 20-second reading split into seven 5-second takes scored 7.0 to 8.3 for
the same speaker; median pitch per take ranged 206 to 295 Hz. Advice: same
text, longer takes, median of three.
- [Real Voices vs Synthetic Tones](https://voice-rater.com/guides/real-voice-vs-synthetic-tone/):
Human readings show 2.5% to 8% jitter in connected speech; a synthetic tone
shows exactly zero jitter and shimmer. Explains why CPPS separates a rough
voice when jitter does not.
- [How Background Noise Changes a Voice Test Result](https://voice-rater.com/guides/background-noise-voice-test/):
Added noise halves raw CPPS; with the site's compensation the healthy
readers' age estimate stays within 3 to 4 years, while a rough voice can be
over-corrected.
- [Pitch vs Resonance](https://voice-rater.com/guides/pitch-vs-resonance/):
40 LibriSpeech voices with only pitch shifted (men to ~200 Hz, women to
~120 Hz, formants unchanged): the site's pitch-based reading was fooled
20/20 each way; an open-source speech model (deep resonance check) only
3/20 and 2/20.
- [Male vs Female Voice Frequency](https://voice-rater.com/guides/male-female-voice-frequency/):
Median speaking pitch of 40 readers: women 150 to 261 Hz (median 196.5),
men 91 to 199 Hz (median 116.5); overlap zone 155 to 175 Hz; full table.
- [How to Track Voice Training Progress](https://voice-rater.com/guides/track-voice-training-progress/):
Two single 5-second takes differ by a median 5 Hz (1 in 10 by 19 Hz or
more); middle-of-three readings cut that to 2 Hz (1 in 10 by 10 Hz).