Our Technology

The technology that makes intelligibility measurable.

Transparent speech engineering for objective assessment, training, and certification.

Built for institutional trust: transparent scoring logic, reproducible calibration, and a runtime architecture designed for auditability, not black-box claims.

Defensibility

2 pending utility patents + calibration IP

Transparency

Outputs tied to acoustic evidence

Scalability

Browser-native real-time extractor

Setup → Tension → Resolution.

Vowel production determines what an air traffic controller hears. Yet vowels can't be taught conventionally — they have no articulatory "touch points." And legacy norms, pooled from decades-old datasets, fail diverse speakers in both directions.

Our resolution is a closed loop: METHOD trains the production, ENGINE measures it against a word-specific corridor, and CALIBRATION ensures the corridor itself reflects how the system actually measures.

SAE VOWEL CORRIDORLEARNERF1F2 (Hz)F1 (Hz)/ɑ//ɛ//i//u/
Technology architecture visual placeholder
METHOD

ORONYM™

Forced Prosodic Alignment via Native-Language Motor Patterns

An oronym is a sequence of words or syllables that, when spoken using phonological patterns already familiar to the learner, produces acoustic output that approximates the target pronunciation. Learners achieve correct vowel production through articulation they already know — without conscious manipulation of tongue or lip position.
ENGINE

VCAM™

Vowel-Centered Accent Modification — Real-Time Acoustic Scoring

VCAM defines word-specific acoustic targets for each phrase in the training sequence. Real-time measurement compares the learner's production against perceptually weighted intelligibility boundaries, generating a transparent 0–100% score with immediate in/out-of-target feedback.
CALIBRATION

Pierson (2026)

Speaker-Calibrated Corridor Norms — Replacing Legacy Pooled Data

The Pierson (2026) corridor set replaces pooled historical datasets with norms derived from a calibrated native SAE speaker, using the same validated extraction pipeline used at runtime. This eliminates systematic drift between what the system considers acceptable and what it actually measures.

Formant Extractor

If measurement is unstable, scoring is meaningless. Our formant extractor — running in WebAssembly in the browser — is the prerequisite for "objective."

6–14×
Frame-success improvement over prior extractor
51–63%
Male frame success (up from 4–9%)
<0.5ms
Per-frame latency, real-time scoring

Patent Portfolio

US
63/755,979
Vowel-Centered Speech Modification
Filed February 7, 2025 • Provisional
US
63/903,553
Vowel-Centered Accent Modification in Pilots
Filed October 22, 2025 • Provisional

Proprietary methods protected under pending US patent applications.

Request Technical Diligence BriefReview Research Validation