name	squawkbox
description	Voice capability for agents — speak text aloud, manage voices, save audio to files. Wraps pocket-tts and miniaudio. Triggers on: 'speak', 'say', 'read aloud', 'voice', 'audio', 'narrate', 'announce', 'text to speech', 'tts'. Gives you a voice. How you use it is up to you and the user.

squawkbox — Voice for Agents

You can speak. Here's what you have:

Tool	What it does
`speak(text, voice?, output_path?)`	Say text aloud. Optionally save to .wav.
`list_voices()`	See available voices (built-in + custom).
`add_voice(name, audio_file_path)`	Register a custom voice from a .wav reference.
`remove_voice(name)`	Remove a custom voice.

alba (default), marius, javert, jean, fantine, cosette, eponine, azelma

Speak with default voice:

speak("Hello, I'm ready to help.")

Speak with a specific voice:

speak("Important update.", voice="marius")

Save audio to file without playing:

speak("Saved for later.", output_path="./message.wav")

List what's available:

list_voices()

Audio streams — playback starts before full text is synthesized
First call loads the model (~1-2s), subsequent calls are fast
Voice cloning from custom .wav requires the gated HuggingFace model (hf auth login)
Built-in voices work without any setup

name	squawkbox
description	Voice capability for agents — speak text aloud, manage voices, save audio to files. Wraps pocket-tts and miniaudio. Triggers on: 'speak', 'say', 'read aloud', 'voice', 'audio', 'narrate', 'announce', 'text to speech', 'tts'. Gives you a voice. How you use it is up to you and the user.