wakewords detection for your apps & webpages

Add your wakewords like "alexa", "hey siri" in your apps.

• Works in the browser with Javascript, Node.js, Swift and Python.
• Train your own custom wakewords detection model in 30 min (Yes, you can do "Jarvis", "Computer", etc)

Quick Start: Detect wakewords in your web pages

npm install wakewords

import { Wakewords } from "wakewords/browser";

const wakewords = await Wakewords.load();
const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
const listener = wakewords.createListener({ stream });

listener.addEventListener("prediction", (event) => {
  console.log(event.detail.label, event.detail.probability);
});

await listener.start();

Default wake words The default model is trained for wakewords from Google Speech Commands dataset v2. List of words below.

backward	bed	bird	cat	dog	down
eight	five	follow	forward	four	go
happy	house	learn	left	marvin	nine
no	off	on	one	right	seven
sheila	six	stop	three	tree	two
up	visual	wow	yes	zero

It does include 12 other words mentioned in the labels.json in this repo. That shouldnt matter if you only want the words from the Speech Commands dataset.

Quick Start: Train for your own wakewords

pip install wakewords

Use uv run for below commands if applicable to you.

wakewords init

# Edit your custom wake words
$EDITOR config.json

export CARTESIA_API_KEY=your-api-key

# Generates synthetic dataset
wakewords generate --lang en --all-voices
wakewords augment

# Train your model
wakewords download
wakewords manifest
wakewords train

# Export to onnx and use whereever
wakewords export

Docs

• JavaScript API
• Python training
• Swift package

License

Copyright © 2026 Akash Manohar John, under MIT License (See LICENSE file).