Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Fixed initialization errors in setup.sh script.
What This Skill Does
Generates text-to-speech audio using Alibaba Cloud's DashScope (Bailian) Qwen TTS API via WebSocket streaming. Supports over 30 mood presets, 35 female voices, and 10 languages including Japanese and Korean. The agent picks moods based on its own emotional reaction to the conversation, not the user's state.
The 30+ named mood presets (nervous, jealous, yandere, asmr, etc.) produce emotionally nuanced speech that generic TTS APIs can't replicate without custom prosody tuning.
When to use it
- Voicing AI companion replies with contextually matched emotions
- Sending opus-format voice messages to Telegram users
- Generating multilingual audio in Japanese, Korean, or Spanish
- Recording WAV or MP3 narration for video or podcast projects
- Producing roleplay character voices with presets like tsundere or maid
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: 'Say good morning to me in Japanese with a gentle voice'
- 1Translate 'good morning' to Japanese internally: 'おはようございます'
- 2Select mood 'gentle' to match a warm companion response
- 3Call tts.py --voice Maia --mood gentle --language Japanese --format mp3
- 4Receive streaming audio from DashScope WebSocket
- 5Send tts_output.mp3 and original translated text to user
MP3 audio of 'おはようございます' in Maia's gentle voice
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.