Machine LearningShipped
wake-word-ml
TC-ResNet8 wake word detector for Alfred
About
A 65K-parameter TC-ResNet8 trained on synthetic macOS TTS voices, hard negatives (Hey Albert, Hey Alfredo), and personal recordings. Processes 40-dimensional MFCCs from 1-second 16 kHz audio and exports to a roughly 250 KB CoreML model. Sub-millisecond per frame on Apple Silicon, well under 2% CPU at 12 inferences per second. The training pipeline supports retraining with more personal samples, additional ambient noise, and tunable sensitivity.
Status
Trained, exported, and deployed inside Alfred. Hits sub-millisecond per frame on Apple Silicon.
Stack
PythonPyTorchtorchaudiocoremltoolsMFCCTC-ResNet8