Feature Request

Current behavior: Talk Mode only supports ElevenLabs for text-to-speech synthesis, requiring audio to be streamed from the cloud on every response.

Proposed behavior: Add support for on-device TTS as a Talk Mode provider option, starting with iOS AVSpeechSynthesizer. This would allow the iPhone to speak responses locally using Apple's built-in voices instead of streaming audio from ElevenLabs.

Why this matters

• Zero latency on the TTS step — no network round-trip for audio streaming
• Works offline — voice responses don't depend on an external API
• No API costs — ElevenLabs usage adds up, especially for frequent Talk Mode users
• Apple's premium voices are good — the newer downloaded Siri voices sound natural and are completely on-device
• Privacy — response text stays on-device rather than being sent to a third-party TTS API

Suggested implementation

• New Talk provider option (e.g., "provider": "system" or "provider": "native") in talk config
• On iOS, use AVSpeechSynthesizer with configurable voice identifier
• Could extend to macOS (NSSpeechSynthesizer) and Android (TextToSpeech) as well
• Keep ElevenLabs as the default for users who prefer cloud voices

Config example

{
  talk: {
    provider: "native",  // or "elevenlabs" (default)
    // native-specific settings
    nativeVoiceId: "com.apple.voice.premium.en-US.Zoe",
    nativeRate: 0.5,
  }
}

Platform: iOS (OpenClaw iOS node app)
Talk Mode docs: docs/nodes/talk.md