openclaw-speaches

Local Speaches speech providers for OpenClaw.

This plugin registers one speaches provider for both:

• realtime speech-to-text for @openclaw/voice-call streaming
• text-to-speech / telephony TTS using Speaches' OpenAI-compatible /audio/speech endpoint, tested with Kokoro

Install

openclaw plugins install clawhub:openclaw-speaches

Example voice-call config

{
  "streaming": {
    "enabled": true,
    "provider": "speaches",
    "providers": {
      "speaches": {
        "baseUrl": "http://127.0.0.1:8000/v1",
        "model": "Systran/faster-distil-whisper-small.en",
        "apiKey": "cant-be-empty",
        "silenceDurationMs": 500,
        "vadThreshold": 0.5,
        "prefixPaddingMs": 0,
        "convertTwilioMulaw": true
      }
    }
  },
  "tts": {
    "provider": "speaches",
    "providers": {
      "speaches": {
        "baseUrl": "http://127.0.0.1:8000/v1",
        "apiKey": "cant-be-empty",
        "model": "speaches-ai/Kokoro-82M-v1.0-ONNX",
        "voice": "af_sky"
      }
    }
  }
}

Global TTS config

{
  "messages": {
    "tts": {
      "provider": "speaches",
      "providers": {
        "speaches": {
          "baseUrl": "http://127.0.0.1:8000/v1",
          "apiKey": "cant-be-empty",
          "model": "speaches-ai/Kokoro-82M-v1.0-ONNX",
          "voice": "af_sky"
        }
      }
    }
  }
}

Environment variables

• SPEACHES_BASE_URL — default http://127.0.0.1:8000/v1
• SPEACHES_API_KEY — default cant-be-empty
• SPEACHES_STT_MODEL — default Systran/faster-distil-whisper-small.en
• SPEACHES_TTS_MODEL — default speaches-ai/Kokoro-82M-v1.0-ONNX
• SPEACHES_TTS_VOICE — default af_sky

Notes

For latency-sensitive phone calls on CPU, smaller faster-whisper models usually perform better than large models. This plugin was tested with OpenClaw voice-call, Speaches, faster-distil-whisper-small.en, and Kokoro ONNX.