Summary

OpenClaw currently supports ElevenLabs, OpenAI, and Edge TTS as built-in providers. It would be great to add Google Gemini TTS (Cloud Text-to-Speech API using Gemini models like gemini-2.5-flash-preview-tts) as a native provider.

Why Gemini TTS?

• Best-in-class Japanese support — ElevenLabs' Japanese quality is inconsistent, Edge TTS is decent but limited in expressiveness
• Style prompts — Gemini TTS supports expressive style instructions (e.g. speak softly, speak cheerfully), which is a unique differentiator
• Multi-voice in one model — supports multiple voices (Leda, Kore, Puck, Aoede, etc.) without separate voice cloning
• Cost — free tier available via Google AI API key; also works via Vertex AI for enterprise
• Google Cloud alignment — many OpenClaw users already have GCP/Vertex AI credentials configured

Proposed Config

{
  messages: {
    tts: {
      provider: "gemini",
      gemini: {
        apiKey: "...",          // or use existing google auth profile
        model: "gemini-2.5-flash-preview-tts",
        voice: "Leda",
        languageCode: "ja-JP",
        stylePrompt: "Speak naturally and warmly",
      },
    },
  },
}

Workaround Attempted

Tried creating a local OpenAI-compatible proxy server that wraps Gemini TTS, but messages.tts.openai does not support baseUrl for custom endpoints — the gateway rejects it as an unrecognized key and crashes.

Environment

• OpenClaw v2026.2.21
• Using Gemini TTS via standalone script successfully (gemini-2.5-flash-preview-tts)
• Would love native integration for Discord voice channels + auto-TTS