-
Notifications
You must be signed in to change notification settings - Fork 2
08 Voice
AgentBridge can listen and speak. On Windows you can dictate your request from the microphone instead of typing it, and the assistant answers by voice as well as in writing. Every conversation can become a natural spoken exchange with your own AI assistant.
The assistant speaks with the built-in Kokoro voice engine: its voices and model are included with AgentBridge, so speech works the moment you run the program, on any machine, with nothing to install. The engine catalog is designed to grow — future engines will appear in the same /ttsengine list.
Type /voice in the chat to dictate from the microphone, and /tts to hear the assistant speak the last answer aloud. Type /ttsengine to see which voice engine is active and which engines are available. The same voice engine also speaks on the phone and in the podcast tool, so your assistant sounds the same everywhere.
On Windows, the speech-recognition component travels with the release — it is part of the package you install, in the voiceagent folder next to the program. If it is missing on an outdated installation, the program does not show technical details: it simply says that the speech-recognition component is missing from this installation and that updating the app to the latest version brings it, since voice comes bundled with it.
Everything runs on your machine, locally, with no downloads.
The next guide explains how to reach your assistant by telephone, so it can help you even when you are away from the computer.