Bioacoustic Resonance Lab

Real-Time Formant & Harmonic Overtone Synthesizer

Real-Time FFT Spectrum FFT 1024 Hz

F1 / F2 Vowel Space Plane Acoustic Vocal Tract Map

Harmonic Overtone Decay H1 - H5 Amplitudes

Acoustic Control Racks

Fundamental Pitch (F0) 340 Hz

F1 Formant Frequency 920 Hz

F2 Formant Frequency 1450 Hz

Vibrato / Resonance Depth 0.15

Timbre Warmth Index

94.20

Formant Ratio (F2/F1)

1.58

Harmonic Clarity

88.5%

Active Acoustic Mode

Human

Same note, different vowel: how formants shape timbre

A one-minute painted explainer of the lab's source-and-filter model, followed by the real vowel measurements behind it and where to read more.

Read the narration

Sing 'ah', then 'ee', on the same note. The pitch stays put, yet the sound changes. This lab rebuilds that trick from a buzz and two filters. First, the source: a sawtooth buzz at three hundred forty hertz. Like your vocal folds, it holds a ladder of harmonics, at six-eighty, one thousand twenty, thirteen-sixty and up, each weaker than the last. Then, the filter. Your throat and mouth resonate at frequencies called formants. The tool models the first two as band-pass filters: F one at nine hundred twenty hertz, and F two at fourteen-fifty. Only harmonics near the formants pass strongly. The fundamental is mostly filtered out, while the rungs from about seven hundred to fourteen hundred hertz ring loudest. That reshaped ladder is the timbre. Vowels are mostly formant positions. Dropping your jaw raises F one; moving your tongue forward raises F two. Here F two over F one is one point five eight, like an open 'ah'. A bright 'ee' pushes it past eight. Change the pitch and the rungs slide, but the formants hold, so 'ah' still sounds like 'ah'. Drag F one and F two and hear the vowel morph.

What the lab models, and what speech research measures

Source plus filter

Voiced speech is modelled as a buzzy source (the vocal folds, rich in harmonics at whole-number multiples of the pitch) shaped by the resonances of the throat and mouth. The lab copies this: a 340 Hz sawtooth fed through two parallel band-pass filters, F1 at 920 Hz (Q 4) and F2 at 1450 Hz (Q 5).

What actually comes out

Run the sawtooth's harmonics through those two filters and the loudest are the 2nd and 3rd (680 and 1,020 Hz), then the 4th (1,360 Hz); the 340 Hz fundamental ends up weakest of the first five. The lab's 'Harmonic Overtone Decay' bars use a separate simplified rule and still draw H1 tallest, so trust your ears over that chart.

Vowels live on the F1–F2 map

In Peterson and Barney's classic 1952 measurements, men's 'ee' (as in heed) averaged F1 ≈ 270 Hz and F2 ≈ 2,290 Hz, a ratio near 8.5, while 'ah' (as in hod) averaged about 730 and 1,090 Hz. Women's and children's formants run higher because their vocal tracts are shorter.

The tool's indices are its own

Formant Ratio is simply F2 ÷ F1 (1.58 at the defaults). 'Timbre Warmth Index' (94.40) and 'Harmonic Clarity' (96.3%) are the tool's own scoring formulas, based on distance from 280 Hz pitch and 880 Hz F1, and on vibrato depth; they are not standard acoustic measures.

About the canine preset

The 'Canine Recognition Preset' values (220 Hz, F1 850, F2 1,220) match Peterson and Barney's average for women saying 'ah' rather than any measured dog vocal tract, so treat it as just another setting.

Try it in the tool

Press ▶ Start Synthesizer, then drag F1 Formant Frequency down toward 300 Hz and F2 Formant Frequency up toward 2300 Hz: the 'ah' turns into an 'ee'.
Leave the formants alone and slide Fundamental Pitch (F0) from 340 down to 150 Hz: the note drops but the vowel colour stays.
Watch Formant Ratio (F2/F1) as you move the formant sliders: front vowels like 'ee' read high, open vowels like 'ah' read low.

Sources and further reading

Control Methods Used in a Study of the Vowels (Peterson & Barney) · Journal of the Acoustical Society of America, 1952
The classic F1/F2 measurements of American English vowels from men, women and children.
Acoustic characteristics of American English vowels (Hillenbrand et al.) · Journal of the Acoustical Society of America, 1995
A modern replication and extension of Peterson and Barney's formant data.
Web Audio API — BiquadFilterNode · W3C
Defines the band-pass filter the lab uses to model each formant.
Formant · Wikipedia
Overview of formants, typical vowel values and the source–filter idea.

The lab is a two-formant toy model; real voices have three to five formants, a source whose harmonics fall off differently from a sawtooth, and formant values that vary by speaker, accent and language.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.