ViolaWake documentation

Start here if you want to use the SDK rather than the marketing site.

Quickstart

pip install "violawake[oww]"

from violawake_sdk import WakeDetector

with WakeDetector(model="temporal_cnn", threshold=0.80) as detector:
    for chunk in detector.stream_mic():
        if detector.detect(chunk):
            print("Wake word detected")
            break

Important SDK facts

The canonical package import is "violawake_sdk".
Wake detection requires the OpenWakeWord runtime backbone. Install the "oww" extra unless your environment already provides it.
The default wake head is a TemporalCNN ONNX model.
Detection runs locally on device.
The SDK includes wake detection, async detection, VAD, optional STT/TTS paths, confidence helpers, power management, and model discovery.

Primary references

Training custom wake words

Use the Console if you want a browser workflow. Use the CLI if you want local control:

violawake-train --word "jarvis" --positives samples/jarvis --output models/jarvis.onnx

For production, collect more than the minimum. Include multiple speakers, microphones, rooms, and hard negatives.

Evaluation tools

ViolaWake documents EER, FAR, FRR, ROC AUC, d-prime, and streaming false alarms per hour. Run evaluation on your own target audio before product deployment.