MCP TTS VOICEVOX

What is MCP TTS VOICEVOX?

MCP TTS VOICEVOX is a text-to-speech MCP server that uses the VOICEVOX engine. It enables AI assistants in MCP clients like Claude Desktop to speak, offers an interactive UI audio player with client-side playback, and supports multi-speaker conversations, segment editing, and WAV export.

How to use MCP TTS VOICEVOX?

Requires Node.js 20+ (or Bun) and a running VOICEVOX Engine (or use Docker Compose). Install via npx -y @kajidog/mcp-tts-voicevox (or bunx) in your Claude Desktop configuration, then restart Claude Desktop. Optional: install ffplay for streaming playback. Environment variables configure engine URL, speakers, playback options, and UI player settings.

Key features of MCP TTS VOICEVOX

• Text-to-speech for AI assistants (Claude Desktop, ChatGPT, etc.)
• Interactive UI audio player with play/pause and track controls
• Multi-speaker dialogue with segment editing and reordering
• WAV export, user dictionary manager, and cross-session state restore
• Streaming playback via ffplay (optional) or fallback temp-file playback
• Cross-platform: Windows, macOS, Linux (including WSL and Docker)

Use cases of MCP TTS VOICEVOX

• Make an AI assistant speak aloud in conversation
• Create multi-character voice dialogues with different speakers per segment
• Generate and export audio files for archival or further editing
• Manage custom pronunciation via the built-in user dictionary

FAQ from MCP TTS VOICEVOX

What is the difference between voicevox_speak and voicevox_speak_player?

voicevox_speak plays audio on the server side (requires a local audio device). voicevox_speak_player renders an interactive UI audio player in the chat, playing audio on the client side (browser/app) and works even on headless servers.

Do I need ffplay to use MCP TTS VOICEVOX?

No. ffplay is optional and enables low-latency streaming playback. Without it, playback falls back to temporary file‑based playback using system players (PowerShell, afplay, aplay, etc.).

Can I run MCP TTS VOICEVOX with Docker?

Yes. Use Docker Compose to start both the VOICEVOX Engine and the MCP server (HTTP mode). Server‑side playback (voicevox_speak) is disabled by default because the container has no audio device; use voicevox_speak_player instead.

Which MCP clients are supported?

ChatGPT works via HTTP (requires VOICEVOX_PLAYER_DOMAIN). Claude Desktop works out of the box via stdio, or via HTTP using mcp-remote. The UI player requires a host that supports MCP Apps.

How do I configure environment variables?

Set variables like VOICEVOX_URL, VOICEVOX_DEFAULT_SPEAKER, VOICEVOX_USE_STREAMING, and UI player settings directly in the shell or environment. See the README's configuration section for a full list.