qwen3-tts-instruct

What This Skill Does

The qwen3-tts-instruct skill provides high-fidelity, emotional text-to-speech capabilities via the Alibaba Cloud Bailian (DashScope) engine. It is designed for OpenClaw agents to communicate with natural, human-like cadence and complex emotional variance. The skill supports over 30 mood presets, allowing the agent to adapt its delivery based on the context of the conversation. Unlike static TTS services, this skill requires the AI to actively decide the tone, ensuring that the voice aligns with the persona's internal emotional state rather than just the literal meaning of the text.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/yanmoon321/qwen3-tts-instruct

Ensure your DashScope API credentials are properly configured in your environment variables as per the OpenClaw documentation.

Use Cases

• Roleplay Agents: Utilize mood presets to express vulnerability, excitement, or hesitation in character-driven narratives.
• Multilingual Support: Providing native-sounding audio responses for global users, provided the text is pre-translated.
• Interactive Assistants: Creating a warmer, more engaging user experience by matching the voice's emotional inflection to the user's input.
• Content Creation: Generating spoken audio for tutorials, reminders, or notifications that require specific emotional beats.

Example Prompts

1. "I am feeling really sad today, can you talk to me in a soft, comforting voice?"
2. "Please translate your response into Japanese and read it out loud using an excited, energetic tone."
3. "Summarize the news for me, but keep the tone professional and neutral."

Tips & Limitations

• Pre-Translation is Mandatory: This skill reads the text exactly as provided. If you intend to speak in a foreign language, you must instruct the LLM to perform the translation task before triggering the TTS skill.
• Contextual Awareness: Always map the user's emotion to an appropriate, non-reactive emotional response. If a user is hostile, don't trigger an 'angry' mood; choose 'submissive' or 'nervous' to maintain conversational safety.
• Format Flexibility: Use the --format flag to switch between pcm, wav, mp3, and opus depending on your application's bandwidth requirements.
• Voice Consistency: Select a voice persona that matches the brand or character identity consistently throughout the session.