openclaw-telegram-audio-chat

OpenClaw plugin that auto-sends Telegram voice bubble replies from assistant text using local TTS.

What it does

• Adds /audio_chat on|off|status|max <n>
• Stores per-chat enable/disable state in the OpenClaw state directory
• Cleans markdown before TTS
• Skips long messages with a tip instead of generating voice
• Debounces sends to reduce duplicate voice bubbles

Install

After publication, install with:

openclaw plugins install openclaw-telegram-audio-chat

OpenClaw checks ClawHub first and falls back to npm automatically.

Config

Use extensions.audioChat (or audioChat) in gateway config:

{
  "extensions": {
    "audioChat": {
      "enabledByDefault": false,
      "channels": ["telegram"],
      "defaultMaxChars": 150,
      "tooLongTip": "Message too long, skipped voice reply",
      "voice": "zh-CN-YunxiaNeural",
      "access": {
        "directOnly": true,
        "allowUserIds": []
      }
    }
  }
}

Dependencies

This plugin requires:

• ffmpeg
• edge-tts available via Python (python3 -m edge_tts) or another dedicated Python environment

Privacy and safety

• No hardcoded personal user IDs or chat IDs
• Voice synthesis uses Microsoft Edge TTS via the internet (edge-tts connects to speech.platform.bing.com). Text sent for synthesis is processed by Microsoft's servers.
• The plugin only sends voice notes to chats where it is enabled
• Recommended default: keep enabledByDefault: false
• Recommended default: keep access.directOnly: true

Compatibility

Notes

Currently focused on Telegram voice-note delivery.