Vocal Formant Frequency Analyzer

A vowel’s character comes from formants — resonant peaks created by the shape of your throat, mouth, and lips. The tool reads a short slice of your microphone, runs linear predictive coding (LPC) to model the spectral envelope as a set of resonances, and reads the lowest peaks as F1, F2, and F3. F1 tracks how open your mouth is (higher = more open, like “ah”); F2 tracks front-vs-back tongue position (higher = more front, like “ee”). Plotting F1 against F2 places your vowel in the classic vowel space — the same acoustic phonetics framework used in linguistics and speech pathology — and the nearest reference vowel is your closest match. Everything runs live on your device; nothing is recorded.

LPC is an estimate, not a measurement. The number of resonances it finds depends on the model order and your sample rate, and it can miss or merge formants on breathy, nasal, or quiet sounds. The reference vowels are adult-average values from the Peterson & Barney (1952) and Hillenbrand datasets, which are a widely cited benchmark in speech science — your absolute formant frequencies scale with vocal-tract length, so a child’s or higher voice will read higher than the chart. Delivering real-time formants in the browser makes this kind of speech analysis accessible without any software installation. For precise spectral analysis of any sound, the Audio Spectrum Analyzer can complement formant work by showing the full frequency envelope. Use sustained, clear vowels for the steadiest readings, and treat the vowel guess as a rough indicator rather than a clinical measurement.