Introduction

We provide the tools to easily train and deploy highly accurate, custom wake words for your voice-enabled applications.

What is a wake word?

A wake word is a specific phrase (like “Hey Siri” or “OK Google”) that a device continuously listens for to trigger an action. This allows your system to stay in a low-power, privacy-preserving state until the user is ready to interact.

Training Your Model

You can build and train your custom wake words instantly using the Model Dashboard. We offer a generous free tier for hobbyists and small companies. Please see our Pricing Page for full details.

Running Your Model

Once your model is trained, you can easily integrate it into your application using our official libraries. Choose your preferred environment below for installation guides and API references:

• Python — sonnetics-py for desktop or server.
• JavaScript — sonnetics-js for web browsers and Node.js.
• Rust — sonnetics-core for Rust applications.

If your preferred language is not yet supported, please get in touch.

How do I know my audio is private?

All audio processing happens entirely on your device — nothing is sent to the cloud. Our models are fully downloadable, and our client libraries are completely open source, giving you full visibility to inspect the code before you run it.