Grok Voice is a voice architecture on Agent settings. One model hears the caller and speaks the reply. You pick a Grok voice. You do not choose a separate listener, thinker, and speaker.
It uses Grok Voice 3.0. The choice applies to both inbound and outbound calls.
Why it matters for your business
- Save money: AI usage is $0.09 per started minute, billed from your Voicetta wallet. You do not add an xAI account.
- Work faster: select the card, choose a voice, and save.
- Standardize service: the greeting and instructions you already wrote still apply.
- Delight every Guest: callers hear one Grok voice. Each option in the menu includes a short description.
How it compares
| Card | What it does |
|---|---|
| Built-in | Older speech-to-speech voice. Still available. You can keep that voice or attach a custom voice provider. |
| Built-in Live | Recommended. The agent can listen and speak at the same time. AI is $0.06 per started minute, plus carrier and LiveKit fees. |
| Grok Voice | Grok Voice 3.0. Grok hears and speaks, using Grok's own voices. |
| Mix & match — Voicetta | You choose who listens, thinks, and speaks. Voicetta bills the AI. |
| Mix & match — Your keys | The same split, using API keys from Settings. Those providers bill you directly. |
Grok Voice takes turns. Built-in Live is the card for listening and speaking at the same time.
What you choose
- The Grok Voice card under Voice architecture.
- The voice callers hear. Eve is the default.
- Your existing greeting and instructions.
You do not pick a separate transcription provider or a custom voice provider on this card.
Grok Voice is for phone calls. SMS and WhatsApp keep the text model configured for those channels.
Step-by-step in the app
- Open Agent → Agent settings.
- Under Voice architecture, select Grok Voice.
- Choose the voice callers will hear.
- Save.
- Place a test call.
Switching to another architecture can change the voice. Test again after you switch.
Billing
Grok Voice AI is $0.09 for each started minute. One second and sixty seconds both cost $0.09. Sixty-one seconds cost $0.18. The charge comes from your Voicetta wallet.
Phone calls also add the usual carrier fee and the LiveKit voice line. Those charges are separate and follow their existing rules.
You do not add an API key for Grok Voice.
Best practices
- Run a test call after you select Grok Voice.
- Read the voice descriptions and pick the one that matches your brand.
- Keep business rules in the prompt. The voice card does not replace them.