Feature Request: Local TTS Model Support (Qwen3-TTS)

Summary

Add support for local TTS models, starting with Qwen3-TTS (千问3 TTS), as an alternative to cloud-based providers like ElevenLabs, OpenAI TTS, and Edge TTS.

Motivation

1. Cost reduction: ElevenLabs quota consumption is very fast for active users
2. Privacy: Keep voice synthesis completely local
3. Offline capability: TTS works without internet connection
4. Quality: Qwen3-TTS provides excellent quality, especially for Chinese language
5. No rate limits: Local models have no API quota concerns

Proposed Solution

Add a new TTS provider option: qwen3-tts (or generic local-tts with model selection)

Configuration example:

{
  "messages": {
    "tts": {
      "provider": "qwen3-tts",
      "qwen3": {
        "modelPath": "/path/to/qwen3-tts/model",
        "voiceId": "zh-CN-female-1",
        "device": "cpu"
      }
    }
  }
}

Technical Details

• Qwen3-TTS (千问3) is an open-source TTS model from Alibaba
• Supports Chinese and English with natural-sounding voices
• Can run locally on consumer hardware (CPU or GPU)
• Model repository: https://github.com/QwenLM/Qwen-TTS

Benefits for OpenClaw Users

• Eliminates dependency on cloud TTS services
• Reduces operational costs for heavy TTS users
• Better privacy and data sovereignty
• Consistent performance without API rate limits

Alternative Consideration

If implementing specific models is out of scope, consider adding a generic local TTS plugin interface that allows users to integrate any local TTS backend (CosyVoice, Bark, Coqui TTS, etc.) through a standardized API.

Current TTS Limitations

As of OpenClaw 2026.3.13, only three cloud providers are supported:

• ElevenLabs (expensive)
• OpenAI TTS (requires API key)
• Edge TTS (free but cloud-based)

Adding local model support would make OpenClaw more flexible and cost-effective for users with high TTS usage.

Thanks for considering this feature!