Skip to content

Getting Started

This page covers everything you need to get Soundboard Plugin running with Claude Code.

RequirementMinimumNotes
voice-soundboard>= 2.5.0The TTS synthesis engine. Install with Kokoro backend for local inference.
Python>= 3.10Required for both the engine and the plugin.
PlatformWindowsPrimary target. Uses winsound for playback.

The voice-soundboard engine is a separate package that handles the actual speech synthesis. Soundboard Plugin is the Claude Code integration layer that sits on top of it.

Terminal window
cd voice-soundboard
pip install -e ".[kokoro]"

The [kokoro] extra pulls in the Kokoro TTS backend for fully local inference. No API keys or cloud services required.

Terminal window
cd soundboard-plugin
pip install -e .
Terminal window
claude plugin add /path/to/soundboard-plugin

This tells Claude Code where to find the plugin. The plugin starts as a stdio MCP server when Claude Code launches.

Once installed, open Claude Code and run these commands:

/soundboard:speak Hello! I'm your coding assistant.

You should hear spoken audio through your default speakers. Try a few more:

/soundboard:narrate src/server.py

This walks through a source file with adaptive pacing, splitting at sentence boundaries.

/soundboard:notify Build succeeded with 0 warnings

Spoken workflow notification using the announcer preset.

/soundboard:voices

Lists all 12 available voices and their presets.

/soundboard:voice-status

Shows engine health, backend info, and enforced limits. Useful for troubleshooting.

If the voice engine is not running or not installed, the plugin stays loaded but all voice tools return descriptive error messages instead of crashing. Check status with /soundboard:voice-status to diagnose issues.

  • Commands — full reference for all slash commands and MCP tools.
  • Voices — browse the voice library and configure emotion routing.
  • Architecture — understand the speech pipeline and security model.