a·ARTICULATIONATLAS / THE FORENSIC LINGUIST
+ A STUDY OF SPEECH, FROM THE INSIDE OUT

The architecture of sound.

Explore the invisible choreography behind every voice.

INTERACTIVE RESEARCH STUDIOVOL. 01 / ARTICULATIONFrom movement to meaning
01 / THE ARTICULATORY MODEL Real-time 3D
FIG. 01ORAL APERTURE
Hard palateUpper incisorsTongue bodyVelum
YX
Drag to rotate · scroll to zoom
ILLUSTRATIVE ANATOMYNOT TO SCALE
Target [ɑ]0.50 / 1.20 s
02 / THE ACOUSTIC PICTURE

See what you hear.

Illustrative synthesis · not measured data

Waveform & spectrogram

01—02

One synthesized gesture, two perspectives.

A01.2 s
6k0
Time (s) →Frequency (Hz) ↑

Vowel space

03

F1 × F2 · select a point to explore.

2004507009002500200015001000500iyɪeøɛæaɑɔoʊuə← F2 (Hz) · frontnessF1
[ɑ] selectedF1 750 · F2 1100

Spectral slice

04

At target · FFT of the synthesized signal.

F1F2F301.5k3k4.5k6kFrequency (Hz) →Relative dB
2,048-point Hann window24 kHz sample rate

Articulatory gesture

05

Movement from rest to target and back.

100.00.30.60.91.2
[ɑ] target activationTime (s) →

Inside the articulation

06

Live model parameters · relative scale, 0–1.

Tongue height0.08
low → high
Tongue position0.22
back → front
Jaw opening0.90
closed → open
Lip rounding0.00
spread → rounded
Nasal opening0.00
oral → nasal
Filled = current poseTick = target
✳

A model is a beginning. Every voice is different.

This atlas makes articulatory relationships visible. Anatomy, gesture timing, and acoustic values are illustrative; real speech varies with speaker, accent, and context.

RESEARCH NOTES / 01

Built for exploration.
Grounded in phonetics.

This proof of concept connects an original, stylized 3D oral model to hand-authored articulatory targets for 36 selected IPA sounds. It is an exploratory instrument, not a calibrated biomechanical simulation.

Anatomy & movement

Independent tongue body and tip, jaw, lips, and velum move through schematic rest–target–release gestures. Cutaway and airflow paths are illustrative. The model is procedural Three.js geometry; no third-party anatomical mesh is bundled.

Acoustics & charts

The listening reference uses harmonic synthesis at 125 Hz for voiced sounds, resonant noise for fricatives, and bursts for plosives. The waveform, spectrogram, and FFT use those actual synthesized samples. Vowel formants are authored synthesis settings, not population means; consonants have no vowel-space point. Model parameters use relative units.

Scope & next research step

The inventory is a curated subset, not the complete IPA. There is no coarticulation, dialect model, measured recording analysis, or speaker identification. Expert review, MRI-derived geometry, articulatory calibration, and licensed human reference recordings are needed before research or teaching claims.

Primary references

International Phonetic AssociationSound classification and chart conventionsBirkholz, Jackèl & Kröger · 2006Construction and control of a three-dimensional vocal tract modelUSC SPAN · real-time MRI IPA atlasObserved articulations and speaker variationPraat · acoustic analysis documentationFormant frequency, bandwidth, and time