openclaw - 💡(How to fix) Fix TTS voice parameter not respected in ElevenLabs provider configuration

Description

The talk.providers.elevenlabs.voice configuration parameter is accepted by the gateway but not actually used when making TTS API calls to ElevenLabs.

Steps to Reproduce

1. Set talk.provider to elevenlabs
2. Configure talk.providers.elevenlabs.apiKey with valid ElevenLabs API key
3. Set talk.providers.elevenlabs.voice to a valid ElevenLabs voice ID (e.g., N2lVS1w4EtoT3dr4eOWO for Callum)
4. Restart the gateway
5. Make a TTS call via the gateway

Expected Behavior

The ElevenLabs API call should include the specified voice ID in the request, resulting in the configured voice being used for text-to-speech generation.

Actual Behavior

The specified voice ID is not being passed to the ElevenLabs API. Responses always use the default/previous voice instead of the configured voice.

Environment

• OpenClaw Version: 2026.5.22
• Provider: ElevenLabs
• Verified ElevenLabs API Key: Working (tested direct API calls)
• Verified Voice ID: N2lVS1w4EtoT3dr4eOWO (Callum - valid voice in ElevenLabs account)

Configuration Attempted

{
  "talk": {
    "provider": "elevenlabs",
    "providers": {
      "elevenlabs": {
        "apiKey": "sk_...",
        "voice": "N2lVS1w4EtoT3dr4eOWO"
      }
    }
  }
}

Workaround

Currently unable to change TTS voice at runtime. The gateway configuration parameter exists but is not being utilized.

Additional Notes

• The talk provider configuration is properly accepted by the gateway
• No errors are thrown when setting the voice parameter
• Gateway logs show configuration changes but no indication of voice parameter being used
• Direct ElevenLabs API calls with the voice parameter work correctly

Impact

Users cannot customize the voice used for text-to-speech responses through the gateway configuration, limiting personalization options.