CyberHorn (赛博小喇叭)

Let OpenClaw speak in Feishu: turn any text into a native voice message (not a file link) in a Feishu chat.

When to use

• The user asks to "say something in Feishu" or "send a voice message to the group".
• The agent should announce or alert via voice in a Feishu room.
• You want replies as playable voice notes instead of text bubbles.

How it works

1. TTS — Text is synthesized with Edge TTS (default, no API key) or ElevenLabs (custom voices). Set TTS_PROVIDER=EDGE or ELEVEN in .env.
2. Encode — Audio is converted to Opus (mono 16 kHz) via FFmpeg for Feishu’s voice message format.
3. Send — The file is uploaded to Feishu and sent as a voice message to the given chat.

Setup

• Required (all modes): FEISHU_APP_ID, FEISHU_APP_SECRET in .env or OpenClaw config. FFmpeg must be on PATH or set FFMPEG_PATH.
• Optional default target chat: FEISHU_DEFAULT_CHAT_ID can be set in .env so you don't have to pass a chat ID every time.
• Edge TTS (default): No extra keys. Optional EDGE_VOICE (e.g. zh-CN-XiaoxiaoNeural).
• ElevenLabs: Set TTS_PROVIDER=ELEVEN, and add ELEVEN_API_KEY, VOICE_ID.

Invocation

From the skill directory (or with PYTHONPATH set):

python main.py "<text to speak>" "[feishu_chat_id]" [receive_id_type]

• Arg 1: Text to speak.
• Arg 2 (optional): Feishu chat ID (or other receive_id). If omitted, the skill will use FEISHU_DEFAULT_CHAT_ID from .env when available.
• Arg 3 (optional): receive_id_type, default chat_id (can be open_id etc. per Feishu API).

OpenClaw can call this with env vars injected and the same two (or three) arguments.