SpeakKit

Voice Calibration

Speak naturally for a few seconds so we can detect your voice type and optimise the analysis parameters.

Or select manually:

Consonant Analysis

Target sound:

—

Energy level — adjust threshold in Settings

Low (50–500 Hz) Mid (500–2k Hz) Hi-Mid (2–5k Hz) High (4–8k Hz) V.High (8–16k Hz)

        Waiting for mic start...
      

Word Practice

Loading speech model...

How it works:

Press Start Listening
Say the target word clearly
Green = matched, Red = heard something else
On correct, the next word appears automatically

Target word:

Say: GO

I heard:

—

0 Attempts

0 Correct

— Accuracy

Settings

Visible toolkit panels (keep the tab bar focused on current work)

Level calibration (speak at your loudest for 3 seconds) Not calibrated

Onset trigger threshold (drag the white line to adjust)

Green bar = current mic level. Drag the white threshold line left (more sensitive) or right (less sensitive). When the bar passes the line, a capture fires. Mic gain is set at the OS level (System Preferences → Sound → Input).

Noise floor calibration (re-run if background noise changes or you move room) Not calibrated

Microphone processing (uncheck all for accurate speech analysis)

Auto Gain Control Noise Suppression Echo Cancellation

These are browser features enabled by default. They distort the audio signal — uncheck all for accurate spectral analysis. Re-calibrate noise floor after changing.

Microphone

Audio output (for playback and audio feedback)

Camera (for face tracking)

Real-time Spectrogram

Scrolling FFT spectrogram with formant tracks overlaid. F1 · F2 · F3

Max freq:

▸ Tuning Parameters

Voice Preset

LPC Order: 12

Downsample: 3

Min Formant: 90 Hz

Max Bandwidth: 400 Hz

Voicing Threshold: 0.30

Tracking Smoothing: 0.30

Voicing: —

SpeakKit

Welcome

Voice Calibration

Consonant Analysis

Word Practice

Face & Lip Tracking

Settings

Patients

Admin: Login Names

Real-time Spectrogram