Edge TTS for OpenClaw

OpenClaw plugin: voice replies using Microsoft Edge TTS via edge-tts CLI.

Natural-sounding Chinese/English speech synthesis for your AI agent.

Why?

Text is great, but sometimes you want to hear your agent's replies. This plugin synthesizes OpenClaw's responses using Microsoft Edge's neural TTS engine via edge-tts — no API key required.

⚠️ Privacy: reply text is sent to Microsoft's Edge TTS cloud service for synthesis. Do not use this plugin if your agent handles sensitive or private information.

Features

Install

Prerequisites

pip install edge-tts
# Audio player (install at least one)
sudo pacman -S mpv       # recommended
# or: ffplay / pipewire / vlc / pulseaudio

Local development

cd /path/to/edge-tts-for-openclaw
npm install
openclaw plugins install --link .

From ClawHub (once published)

openclaw plugins install clawhub:edge-tts-for-oc

Configuration

After installing, enable the plugin:

// ~/.openclaw/openclaw.json
{
  plugins: {
    allow: ["edge-tts-for-oc"],
    entries: {
      "edge-tts-for-oc": {
        enabled: true,
        config: {
          voice: "zh-CN-YunxiaNeural",  // default voice
          autoSpeak: "auto",             // "always" | "auto"
        },
      },
    },
  },
}

Then restart:

openclaw gateway restart

⚠️ Expose the speak tool to the model

OpenClaw gates plugin tools behind its tool policy. If your config sets tools.profile to an allowlist profile (for example the default coding), the plugin's speak tool is not visible to the model until you explicitly allow it:

{
  tools: {
    profile: "coding",
    alsoAllow: ["speak"], // expose the plugin's speak tool
  },
}

Alternatively, add it via CLI:

openclaw config set tools.alsoAllow '["speak"]' --strict-json --merge

If tools.profile is unset or full, the speak tool is exposed automatically and no alsoAllow entry is needed.

Config fields

Usage

Auto-speak mode (always)

Every reply is automatically spoken aloud.

Manual mode (auto)

The model can call the speak tool to read text aloud:

speak(text: "构建成功！所有测试通过。")

Listing available voices

edge-tts --list-voices | grep -i zh-CN

Dependencies

• TTS engine: edge-tts (Python package)
• Audio playback: mpv (recommended), or ffplay / pw-play / cvlc / paplay
• Desktop: Linux

Privacy

This plugin sends reply text to Microsoft's Edge TTS cloud service (edge-tts CLI) for speech synthesis. No text is logged or stored locally. If your agent handles sensitive or private content, disable auto-speak or avoid this plugin.

License

MIT