Skip to content
Nappai Documentation

Grok Voice

Grok Voice is one of the models you can connect to a Voice Agent Realtime. A single xAI model hears the caller, decides when they have finished, thinks and answers in its own voice.

It plugs into the Realtime model connector of the Voice Agent Realtime node. It holds the model, voice and credential the agent uses on every call.

When to use it

  • You already use Grok and want the same credential for your voice agent.
  • You want the greeting and fixed farewells said exactly as you write them.
  • You want a per-minute call price that is easy to work out.

Configuration

You need a Grok API credential with your xAI account key. See Credentials.

FieldWhat it does
Credential (required)The Grok API credential used to connect to the service.
Modelgrok-voice-think-fast-2.0 (default) answers soonest. grok-voice-think-fast-1.0 costs less per minute. grok-voice-fast-1.0 is also available. grok-voice-latest always moves to xAI’s newest model, at its price too.
VoiceThe voice it speaks with: ara (default), eve, leo, rex, sal and more. All of them speak any language.

Advanced options

FieldWhat it does
Answer EagernessHow much silence it waits before answering: auto and high answer after 0.2 s (the quickest), medium waits 0.4 s and low 0.7 s, for callers who pause or noisy places.

Output

The output holds the model settings. It connects to the Realtime model connector of the Voice Agent Realtime node.

Example

A language school wants an agent that gives course information and always reads the same legal notice at the start.

  1. Add Grok Voice and connect it to Realtime model on the Voice Agent Realtime.
  2. In Credential, choose the Grok API credential.
  3. Leave Model on grok-voice-think-fast-2.0 and pick a voice in Voice.
  4. Write the notice in the agent’s Greeting: Grok will say it word for word.
  5. Test the agent in the playground and listen to how it sounds.

Good practices

  • Try several voices in the playground before deciding: all of them speak your language, but each has its own tone.
  • Use grok-voice-latest only if you are happy for the model, and its price, to change without notice.
  • Response time varies. On a test call on our development server on 14 October 2026, grok-voice-think-fast-2.0 took between 1.6 and 1.8 seconds from the moment the caller stopped speaking to the first sound of its answer. Check it in the playground with your own calls.
  • xAI bills the usage to your own account, per minute of call. At its public prices on 14 October 2026, about $0.08/min with Think Fast 2.0 and $0.05/min with Think Fast 1.0. Check the current price there.