Talk to Natro on Windows

Download the voice models

Run the setup script again, this time without -SkipModels:

powershell -ExecutionPolicy Bypass -File scripts\setup-pc.ps1

Set up speech recognition

Natro hears you with Google Cloud Speech-to-Text (Chirp 3).

1. Create a project in the Google Cloud Console.
2. Enable the Cloud Speech-to-Text API.
3. Create a service account with the Cloud Speech Client role.
4. Save its JSON key as google-key.json at the top of the repository, and set GOOGLE_CLOUD_PROJECT in .env.

API keys don’t work with Chirp 3. It needs the service account’s key.

Start Natro

In two terminals:

.venv\Scripts\python.exe -m natro_agent.server

.venv\Scripts\python.exe -m natro_pc.app --wake

Say “OK Natro” once the listener is ready. Without --wake, press Enter to record; --type lets you type instead.

Or use the tray app

.venv\Scripts\pythonw.exe -m natro_pc.tray

The talk key is Ctrl+Alt+Space. Change it with NATRO_HOTKEY in .env.