Voice Activity Detection

A reference consumer using only voice-core's public API. No game code. Connect the mic, then run each calibration step independently: DSP level calibration, phoneme/sound training, and wake-word training are all separate, explicit actions. The current training state and live wake-word diagnostics are shown below.

Training state

DSP tuned— DSP fingerprint— Classifier— Sound labels— Wake-word model— Words— Word sequences— Detect threshold—
0.90

Live wake-word scores

(train a wake word to see live scores)
○ idle