Multiple Wake Word Support

Capability

openWakeWord can run 15-20 models simultaneously on a single Raspberry Pi 3 core.

Use Cases

Implementation

from openwakeword.model import Model

# Load multiple wake word models
models = {
    "hey_lars": Model("models/hey_lars.tflite"),
    "hey_case": Model("models/hey_case.tflite"),
    "hey_lena": Model("models/hey_lena.tflite")
}

def on_wake_word(wake_word: str):
    if wake_word == "hey_lars":
        route_to_lars()
    elif wake_word == "hey_case":
        route_to_mobile_ai()
    elif wake_word == "hey_lena":
        route_to_nexus()

Training Each Wake Word

1. Use Google Colab notebook for each phrase
2. ~1 hour per wake word
3. Export separate .tflite files
4. Load all models at startup

Resource Usage

• Each model adds minimal CPU overhead
• All run in parallel, listening simultaneously
• Detection is instant when triggered