​Hi everyone!
​I’m excited to share FORMANT-1, a lightweight speech synthesis engine built for the Bytebeat environment. This isn’t just a simple wave generator — it’s a mathematical model that simulates the biological mechanics of the human vocal tract.
​In this demo, the engine says "HELLO".
​Key Features:
​LF-Source: Uses an asymmetric glottal impulse model for a natural "chesty" male timbre.
​Nasal Cavity: Features an independent resonance channel (~280Hz) to add warmth and realism.
​Dynamic Coarticulation: Real-time smoothing of formants for fluid speech transitions.
​🚀 How to run it:
​Open any online Bytebeat player (e.g., Dollchan Bytebeat or GreasyFork Bytebeat).
​Set the Mode to JavaScript (or float).
​Set the Sample Rate to 44100 (this is crucial for the filters to sound correct).
​Paste the code below into the editor and hit Play.
​💻 The Code:
t === 0 ? (window.m = {h:new Float32Array(16).fill(0),f1:500,f2:1500,p:100,a:0,n:0.1}) : 0,
sc = [[500,1500,0,0,0.8,0],[500,1800,100,1,0.1,0.1],[400,1000,102,0.8,0,0.3],[450,850,95,1,0,0.1],[0,0,0,0,0,0],[0,0,0,0,0,0]][Math.floor((t/44100*1.4)%6)]||[0,0,0,0,0,0],
window.m.f1 += (sc[0]-window.m.f1)*0.25,
window.m.f2 += (sc[1]-window.m.f2)*0.25,
window.m.p += (sc[2]-window.m.p)*0.12,
window.m.n += (sc[5]-window.m.n)*0.15,
ph = (t/44100*window.m.p)%1,
gl = window.m.p>0 ? (ph<0.6 ? Math.sin(3.14ph/0.6) : -Math.sin(3.14(ph-0.6)/0.4)*0.25) : 0,
src = (gl+(Math.random()-0.5)0.06)sc[3] + (Math.random()-0.5)*sc[4],
window.m.a += ((sc[2]>0||sc[4]>0.5?1:0)-window.m.a)*0.3,
r1 = 0.996, c1 = 2r1Math.cos(6.28*window.m.f1/44100),
o1 = src + c1window.m.h[0] - r1r1*window.m.h[1],
window.m.h[1] = window.m.h[0], window.m.h[0] = o1,
v1 = o1(1-r1r1),
r2 = 0.994, c2 = 2r2Math.cos(6.28*window.m.f2/44100),
o2 = v1 + c2window.m.h[2] - r2r2*window.m.h[3],
window.m.h[3] = window.m.h[2], window.m.h[2] = o2,
v2 = o2(1-r2r2),
r3 = 0.994, c3 = 2r3Math.cos(6.28*280/44100),
o3 = src + c3window.m.h[4] - r3r3*window.m.h[5],
window.m.h[5] = window.m.h[4], window.m.h[4] = o3,
v3 = o3(1-r3r3),
Math.tanh(((v20.8) + (v3window.m.n2.0)) 12) * window.m.a

📚 Vowel Reference Map (Create your own words!)
​Modify the sc array in the code using these F1/F2 frequency values:

Vowel F1 (Hz) F2 (Hz) Mouth Position
[A] (Father) 750 1200 Open
[E] (Bed) 500 1800 Mid-Front
[I] (See) 300 2300 High-Front
[O] (Hot) 500 900 Mid-Back
[U] (Boot) 350 800 High-Back
Tips:
​To change Pitch, modify the 3rd value in the sc sub-arrays.
​To make it Whisper, set the 4th value to 0 and increase the 5th value (Noise).
​Enjoy the synthesis! Let me know if you find any cool new sounds.