Xeontts

What This Skill Does

XeonTTS is a localized speech synthesis skill designed for the OpenClaw agent ecosystem, specifically optimized for QQBOT workflows. It leverages OpenVINO-accelerated Qwen3-TTS models (Base and Custom variants) to provide high-performance voice synthesis directly on your machine. By running two dedicated local services (Flask on port 5002 for inference and Node.js on port 9002 for workflow management), it ensures efficient, low-latency audio generation without relying on external cloud APIs.

Installation

To install XeonTTS, execute the following commands in your OpenClaw environment:

1. Run clawhub install openclaw/skills/skills/aurora2035/xeontts.
2. Navigate to the installation directory: cd "$HOME/.openclaw/workspace/skills/xeontts".
3. Execute the setup script: bash install.sh. This script will create a dedicated Python 3.10 virtual environment, download the required INT8-quantized Qwen3 models, register systemd services for automated startup, and perform a health check to ensure both the Flask and Node.js services are operational.

Use Cases

XeonTTS handles two primary audio synthesis tasks:

• Voice Cloning: The skill guides users through a two-step process requiring a 3-5 second reference audio sample to generate custom audio output.
• Tone-Specific TTS: Users can request text to be read using specific emotions or styles (e.g., "happy," "serious") using the Custom model.

Note: This skill strictly ignores speech-to-text or transcription requests; those are reserved for the xeonasr skill.

Example Prompts

1. "我要克隆音色。" (Triggers the cloning workflow and prompts for a reference audio file).
2. "用开心的语气朗读这首诗：[文本内容]" (Directly generates audio using the Custom model with a specific tone).
3. "把这段文字生成语音文件，我要保存到本地。" (Initiates TTS synthesis and provides a local path).

Tips & Limitations

• Safety & Compliance: All reference and generated files in references/ and outputs/ are automatically purged after 7 days.
• Constraint Management: Reference audio must be 3-5 seconds long; otherwise, the request will be rejected. Output audio length is capped at 20 seconds for Base/Clone models and 30 seconds for Custom models. If your text is too long, please split it into smaller segments.
• Integration: XeonTTS is designed to coexist with existing ASR tools. It will not override your current media or transcription settings, as it is configured solely through the channels.qqbot.xeonTts path.