Voice Setup Tutorial
Complete guide to setting up Voice Wake, Talk Mode, and hands-free voice interaction with OpenClaw
Complete guide to setting up Voice Wake, Talk Mode, and hands-free voice interaction with OpenClaw
I set up voice so I could talk to OpenClaw without touching the keyboard—wake word plus Talk Mode plus ElevenLabs for responses. It takes about 20–30 minutes. Voice Wake is macOS/iOS/Android; Talk Mode works everywhere but needs an ElevenLabs key for TTS. Here's what actually worked for me.
Wake word not triggering: Check microphone permissions and that Voice Wake is enabled in settings. On macOS, the menu bar app must be running. If you're on a different platform, confirm Voice Wake is supported (macOS, iOS, Android).
No speech output or ElevenLabs errors: Verify your ElevenLabs API key and quota. If you're using a different TTS provider, make sure it's configured in the voice section of your config. Test with a short message first.
Latency or lag: Voice recognition and TTS add round-trips. For faster feedback, use a model with low latency and a close-by region for ElevenLabs if available.
By the end you'll have:
Before starting:
Voice Wake enables always-on speech recognition. Let's configure it:
On macOS, Voice Wake integrates with the system speech recognition:
Alternatively, configure Voice Wake in your OpenClaw configuration file:
{
"voice": {
"wake": {
"enabled": true,
"wakeWord": "hey openclaw"
}
}
}
Save the configuration and restart the Gateway:
openclaw gateway restart
On iOS and Android, Voice Wake is available through the companion apps:
For detailed platform setup, see macOS, iOS, or Android guides.
Talk Mode enables two-way voice conversation with text-to-speech responses. We'll use ElevenLabs for high-quality voice synthesis.
Add Talk Mode configuration to your OpenClaw config:
{
"voice": {
"talk": {
"enabled": true,
"provider": "elevenlabs",
"apiKey": "your-elevenlabs-api-key",
"voice": "default"
}
}
}
Security Note: Store your API key securely. Consider using environment variables or a secrets manager instead of hardcoding it.
ElevenLabs offers multiple voice options. You can:
Update the "voice" field in your configuration with your chosen voice ID.
Now let's test both Voice Wake and Talk Mode:
You can also test via the CLI:
openclaw agent --message "Say hello in a voice message"
Customize your wake word to something personal:
{
"voice": {
"wake": {
"enabled": true,
"wakeWord": "hey assistant"
}
}
}
On mobile devices, you can use push-to-talk instead of always-on listening:
Configure audio quality for Talk Mode responses:
{
"voice": {
"talk": {
"enabled": true,
"provider": "elevenlabs",
"voice": "default",
"stability": 0.5,
"similarity_boost": 0.75
}
}
}
Now that everything is set up, here's how to use your voice assistant:
openclaw gateway restartopenclaw gateway logsNow that you have voice features set up, explore these related topics:
Voice features work best when combined with other OpenClaw capabilities. Try asking OpenClaw to: