openclaw - 💡(How to fix) Fix [Feature]: ElevenLabs Realtime Voice Provider for Talk

Feature Request: ElevenLabs Realtime Voice Provider Is your feature request related to a problem? Please describe. Currently, talk.realtime.provider only supports openai and google. Users who prefer ElevenLabs (better voice quality, lower latency with Flash v2.5, specific voice identities) cannot use native realtime voice conversations without routing through OpenAI or Google backends. Describe the solution you'd like Add elevenlabs as a supported talk.realtime.provider, leveraging ElevenLabs' existing realtime capabilities: Streaming TTS: Eleven Flash v2.5 (sub-300ms latency) via WebSocket streaming Realtime STT: Scribe v2 Realtime (already bundled for Voice Call plugin) Conversational AI: ElevenLabs Conversational AI API (WebSocket-based, supports turn-taking, interruption detection) Describe alternatives you've considered stt-tts mode: Works with ElevenLabs TTS but lacks the sub-300ms latency and smooth turn-taking of true realtime Voice Call plugin: Only for telephony use cases, not browser Talk Additional context ElevenLabs APIs that could be integrated: Text-to-Speech Streaming API (/v1/text-to-speech/{voice_id}/stream with optimize_streaming_latency=4) Conversational AI API — full-duplex WebSocket conversation with built-in VAD, interruption handling, and agent configuration Scribe v2 Realtime — already partially integrated for Voice Call streaming STT Configuration example (desired):

{
  talk: {
    realtime: {
      provider: "elevenlabs",
      providers: {
        elevenlabs: {
          apiKey: "${ELEVENLABS_API_KEY}",
          model: "eleven_flash_v2_5",  // or "eleven_v3"
          voice: "N2lVS1w4EtoT3dr4eOWO", // Callum - Husky Trickster
          latencyTier: 4,  // 0-4, maps to optimize_streaming_latency
          // Conversational AI specific:
          agentId: "<optional-convai-agent-id>", // for custom agents
        },
      },
      mode: "realtime",
      transport: "webrtc", // or "provider-websocket"
      brain: "agent-consult",
    },
  },
}

Use case: User has ElevenLabs API key configured for TTS and wants the same voice identity + low latency in browser Talk realtime mode, without needing OpenAI API access. Target Repository: https://github.com/openclaw/openclaw/issues Labels: enhancement, voice, talk, elevenlabs Related: ElevenLabs Conversational AI API, Streaming TTS API, Scribe v2 Realtime

Submitted by: Nas