Text-to-Speech Configuration

Route: /ttsSidebar Group: Capabilities Access: Owner (Cross-tenant)

---

Overview

TTS (Text-to-Speech) allows GoClaw to read agent responses aloud using synthesized voice. The system supports multiple providers and can be configured to automatically play audio based on conditions.

---

Configuration Guide

1. Select the Primary Provider

Go to /tts and select a provider from the list:

• None (Disabled) — disable TTS
• OpenAI — use OpenAI TTS API
• ElevenLabs — use ElevenLabs API (high-quality voices)
• Edge (Free) — use Microsoft Edge TTS (free, no API key required)
• MiniMax — use MiniMax TTS API

2. Configure Auto-Apply Mode

Value	Description
off	Agent can use TTS tool manually
always	All replies get audio
inbound	Only when user sends voice/audio
tagged	Only when reply contains special tag

3. Reply Mode

• Final only — only play audio for the final complete response
• All — play audio for all turns in the conversation

4. Limits and Timeout

• Max Text Length — maximum text length before truncation (default: 1500 characters)
• Timeout (ms) — maximum wait time for API call (default: 30000 ms)

5. Per-Provider Settings

Each provider has its own parameters:

OpenAI TTS

• API Key (from env OPENAI_API_KEY)
• Model: default gpt-4o-mini-tts
• Voice: alloy, echo, fable, onyx, nova, shimmer

ElevenLabs

• API Key (from env ELEVENLABS_API_KEY)
• Voice ID: default pMsXgVXv3BLzUgSXRplE
• Model ID: default eleven_multilingual_v2

Edge TTS

• Enabled toggle
• Voice: default en-US-MichelleNeural
• Speech Rate: reading speed (e.g., +10%, -20%)

MiniMax

• API Key + Group ID
• Model: default speech-02-hd
• Voice ID: select a voice

---

User Interface (UI)

The TTS configuration page includes:

• Status card — displays the active provider and connection status
• General Settings — primary provider, auto-apply mode, reply mode, max text length, timeout
• Per-provider settings cards — each provider has a dedicated card with corresponding fields
• Save button — save all settings
• Refresh button — reload from server

---

Example

Configure TTS using Edge TTS (free), auto-play for inbound messages:

{
  tts: {
    provider: "edge",
    auto: "inbound",
    mode: "final",
    max_length: 1500,
    timeout_ms: 30000,
    edge: {
      enabled: true,
      voice: "vi-VN-HoaiMyNeural",  // Vietnamese voice
      rate: "+0%"
    }
  }
}

---

Notes

• TTS provider API keys are not stored in config.json — use environment variables or .env.local
• Edge TTS does not require an API key but depends on internet connectivity
• auto: "tagged" requires the agent to return a special tag in its response — check the corresponding agent documentation
• When mode: "all", the number of API calls increases significantly — monitor costs with OpenAI/ElevenLabs

---

See Also

• Configuration Reference — TTS section
• Provider Configuration