OpenClaw Voice & TTS Setup: Add Speech to Your AI Agent

February 08, 2026

OpenClaw Voice & TTS Setup: Add Speech to Your AI Agent

By David Park | February 7, 2026

OpenClaw Voice & TTS Setup: Add Speech to Your AI Agent

Voice interaction makes OpenClaw feel more natural. Whether you want voice responses, voice commands, or full conversational audio, OpenClaw supports multiple TTS (text-to-speech) providers.

---

Why Add Voice?

Use Case	Benefit
Hands-free interaction	Use OpenClaw while driving or cooking
Accessibility	Audio output for visually impaired users
Natural conversation	Voice feels more personal than text
Mobile convenience	Send voice notes instead of typing

---

TTS Provider Options

1. OpenAI TTS (Recommended for Starters)

Setup:

# Configure OpenAI API key openclaw config set providers.openai.apiKey "sk-..."
Enable OpenAI TTS
 openclaw config set tts.openai.enabled true openclaw config set tts.openai.voice "alloy" openclaw config set tts.openai.model "tts-1"

Available voices:

alloy — Neutral, balanced

echo — Clear, slightly deeper

fable — Warm, storytelling

onyx — Deep, authoritative

nova — Bright, friendly

shimmer — Light, cheerful

HD quality (slower):

openclaw config set tts.openai.model "tts-1-hd"

2. ElevenLabs (Best Quality)

Setup:

# Get API key from elevenlabs.ai openclaw config set providers.elevenlabs.apiKey "YOUR_API_KEY"
Enable ElevenLabs
 openclaw config set tts.elevenlabs.enabled true openclaw config set tts.elevenlabs.voiceId "21m00Tcm4TlvDq8ikWAM"  # Rachel

Popular voices:

Voice ID	Name	Style
21m00Tcm4TlvDq8ikWAM	Rachel	Warm, female
AZnzlk1XvdvUeBnXmlng	Domi	Neutral, female
nPczCjz82KWscJtnM8EX	Charlie	Casual, male

3. macOS Say Command (Free, Local)

# Enable built-in TTS openclaw config set tts.macossay.enabled true
List available voices
 say -v "?" | head -20
Set a specific voice
 openclaw config set tts.macossay.voice "Samantha"

Available on macOS only. No API key required.

4. Edge TTS (Free, Cloud)

# Enable Edge TTS openclaw config set tts.edge.enabled true openclaw config set tts.edge.voice "en-US-AriaNeural"

---

Voice Wake (Hands-Free Activation)

What is Voice Wake?

Voice wake allows you to activate OpenClaw by saying a trigger phrase, like "Hey OpenClaw" or "Computer."

Setup Voice Wake

# Enable voice wake openclaw config set voiceWake.enabled true
Set trigger phrase
 openclaw config set voiceWake.trigger "Hey OpenClaw"
Sensitivity (0.1 to 1.0)
 openclaw config set voiceWake.sensitivity 0.7

Using with Microphone

Voice wake requires a microphone and speech recognition:

# Test microphone openclaw voice test-microphone
Enable speech recognition
 openclaw config set stt.enabled true openclaw config set stt.provider "whisper"

---

Voice Mode (Conversational Audio)

Enable Full Voice Mode

# Enable voice mode openclaw config set voiceMode.enabled true
Set input mode
 openclaw config set voiceMode.input "microphone"  # or "voice_note"
Set output mode
 openclaw config set voiceMode.output "tts"  # or "voice_message"

Conversation Flow

1. You speak or send voice note 2. OpenClaw transcribes speech to text 3. Agent processes the request 4. Response converted to speech 5. Audio sent back to you

---

Telegram Voice Integration

Telegram supports voice notes natively:

# Enable voice responses openclaw config set telegram.voiceEnabled true openclaw config set telegram.voiceProvider "elevenlabs"

OpenClaw will respond with voice notes instead of text messages.

---

Configuration Examples

Basic TTS Setup

{   "tts": {     "provider": "openai",     "openai": {       "enabled": true,       "voice": "nova",       "model": "tts-1"     }   } }

Multi-Provider Fallback

{   "tts": {     "provider": "elevenlabs",     "elevenlabs": {       "enabled": true,       "voiceId": "21m00Tcm4TlvDq8ikWAM"     },     "openai": {       "enabled": true,       "voice": "nova"     },     "macossay": {       "enabled": true,       "voice": "Samantha"     }   } }

OpenClaw will try ElevenLabs first, fallback to OpenAI, then macOS say.

---

Testing Your Setup

Test TTS

# Say something openclaw tts speak "Hello, I can speak!"
Test all providers
 openclaw tts test --all

Test Voice Wake

# Listen for trigger openclaw voice wake --test
Say "Hey OpenClaw" and watch for activation

Check Audio Quality

# Generate sample audio openclaw tts sample --provider openai --voice nova
Play the file
 afplay sample.mp3  # macOS or
 play sample.mp3    # Linux with sox

---

Troubleshooting Voice Issues

Issue	Solution
No audio output	Check `tts.enabled` and provider API keys
Robotic voice	Switch to ElevenLabs for natural audio
Voice wake not detecting	Increase sensitivity, check microphone
Telegram voice not sending	Enable `telegram.voiceEnabled`
High latency	Use `tts-1` (not HD) for faster response

Verify Configuration

# Check TTS settings openclaw config get tts
Test provider connection
 openclaw tts test --provider openai
View audio output path
 openclaw tts paths

---

Raspberry Pi Voice Setup

On Raspberry Pi, use local TTS for speed:

# Install eSpeak (free, local) sudo apt-get install espeak
Configure OpenClaw
 openclaw config set tts.espeak.enabled true openclaw config set tts.espeak.voice "en"

For better quality on Pi:

# Install Piper (Neural TTS, runs locally) pip install piper-tts
Configure
 openclaw config set tts.piper.enabled true openclaw config set tts.piper.model "en_US-lessac-medium.onnx"

---