coyotte508/momo-wake-word

Momo Wake Word

Wake word detection model for "Momo", trained via openwakeword.com.

Designed for the home-agent project — a local AI home assistant running on dedicated GPU hardware.

Files

File	Format	Use case
momo.onnx	ONNX	Python (openwakeword), browser (onnxruntime-web)
momo.tflite	TFLite	Mobile, microcontrollers, edge devices

Usage (Python)

pip install openwakeword sounddevice numpy

from openwakeword.model import Model
from openwakeword.utils import download_models
from huggingface_hub import hf_hub_download

download_models()  # one-time: downloads preprocessing models
model_path = hf_hub_download("coyotte508/momo-wake-word", "momo.onnx")
oww = Model(wakeword_models=[model_path], inference_framework="onnx")

Training

Trained at openwakeword.com with the wake phrase "Momo". The model uses openwakeword's mel-spectrogram + Google speech embedding pipeline for lightweight, CPU-friendly inference.

Language

This model is optimized for French speakers. For other languages, train a new model at openwakeword.com.