CORIS.THIRI.AI / EXPERIMENTS / WIDGET 09

Moveable-Do Chromatic Solfège & Formant Alignment

Standard TTS Phonetic Engine
Raw Grapheme Failure ("thee-ree" / "thigh-ree")

Naive text-to-speech engines mispronounce brand phonetics and fail to shape acoustic vocal tract formants F1/F2.

THIRI Acoustic Formant Pipeline
Phonetic Token Replacement ("Theory" /ˈθɪəri/) + Dual Biquad Formants

Passed phonetically as "Theory" to speech models, locked with exact oral cavity resonances for sung Solfège degrees.

FORMANT SPECTRUM ANALYZER: Oral Cavity Resonances F1 (Jaw) & F2 (Tongue) VOWEL FORMANT: RAW RECTIFIED (NO TRACT FILTER)
Interactive Moveable-Do Chromatic Keybed (Root: C3 · 130.81 Hz)
Tonic Root:
Do 1
Ra ♭2
Re 2
Me ♭3
Mi 3
Fa 4
Fi ♯4
Sol 5
Le ♭6
La 6
Te ♭7
Ti 7
Do' 8
Vowel Formant Profile:
/a/ (Ah)
/e/ (Eh)
/i/ (Ee)
/o/ (Oh)
/u/ (Oo)
Select vocal tract acoustic aperture
Formant Resonances (F1 / F2):
F1: 800 Hz • F2: 1200 Hz
Dual biquad bandpass oral cavity filters
Phonetic Speech Engine:
TOKEN: "Theory" → /ˈθɪəri/
Displays "THIRI", voices "Theory"
Pitch Fundamental ($f_0$) 130.8 Hz (C3)
Formant Bandwidth (Q) 5.5 Q
Phonetic Pronunciation "theory"
Vocal Tract Realism 98.4% NATURAL
👤 FOR CREATORS & LISTENERS · VOCAL MOUTH-SHAPE & BRAND EAR TEST Instant A/B Audition
The "Robot Kazoo vs. Human Choir" & Brand Pronunciation Test

Why AI voices fail on brand names & singing: First, if you type "THIRI" into a standard AI voice, it guesses "thee-ree" instead of "Theory" (/ˈθɪəri/). Second, when singing musical notes, raw sound waves sound like a flat plastic kazoo unless shaped by a human jaw (F1) and tongue (F2).

1. COMPARE BRAND PRONUNCIATION • 2. AUTO-PLAY SUNG SOLFÈGE MOTIF:
👄 Active Mouth Shape (/a/ "Ah"): Jaw dropped wide open (F1: 800 Hz), tongue relaxed center (F2: 1200 Hz).
💻 FOR DEVELOPERS · PHONETIC MIDDLEWARE & BIQUAD FORMANT BANK
Phonetic Alias Transform & Dual-Bandpass Formant Synthesis

      

    


    
    

      

        Acoustic Anatomy: Human Formant Frequencies
      

      

        The human vocal tract functions as a series of acoustic resonators. The fundamental frequency \(f_0\) produced by the vocal folds is shaped by the pharynx and oral cavity, creating distinct spectral peaks called formants:
        F1 (first formant) responds to jaw opening (low for /i/ and /u/, high for /a/), while F2 (second formant) responds to tongue position (high for front vowels like /i/, low for back vowels like /u/).
        Standard speech models that synthesize naive text often produce flat, unnatural spectral distributions. THIRI filters harmonic saw waves through authentic dual biquad formants to produce organic, human vocal timbre on every chromatic scale degree.