SiliconFlow TTS Generation Skill

Text-to-Speech using SiliconFlow API with CosyVoice2 model. Supports 8 preset voices, multiple languages, and Chinese dialects.

Features

• 🎙️ 8 Preset Voices: 4 male + 4 female voices
• 🌍 Multilingual: Chinese, English, Japanese, Korean
• 🗣️ Chinese Dialects: Cantonese, Sichuan, Shanghai, Tianjin, Wuhan
• ⚡ Ultra Low Latency: 150ms first packet delay
• 🎵 Voice Cloning: 3-second rapid voice cloning
• 💾 Auto Download: Saves audio files locally

Requirements

• Environment Variable: SILICONFLOW_API_KEY
• Optional Config File: ~/.openclaw/openclaw.json (for auto-detect)

Installation

npx clawhub install siliconflow-tts-gen

Configuration

Set your SiliconFlow API key:

export SILICONFLOW_API_KEY="your-api-key"

Usage

List Available Voices

python3 scripts/generate.py --list-voices

Generate Speech

# Basic usage (default voice: alex)
python3 scripts/generate.py "你好，世界"

# Specify voice
python3 scripts/generate.py "Hello World" --voice bella

# Adjust speed
python3 scripts/generate.py "你好" --voice claire --speed 0.9

# Save to file
python3 scripts/generate.py "欢迎收听" --output welcome.mp3

# Change format
python3 scripts/generate.py "Hello" --format wav

Available Voices

Male Voices

Female Voices

Parameters

Security Notes

• This skill requires an API key to call SiliconFlow services
• The script reads ~/.openclaw/openclaw.json only to auto-detect API keys
• No sensitive data is transmitted except to api.siliconflow.cn
• Review the code at scripts/generate.py before providing credentials

Author

MaxStorm Team

License

MIT