Skip to content
Nappai Documentation

Gemini Live

Gemini Live is one of the models you can connect to a Voice Agent Realtime. A single Google 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, language and credential the agent uses on every call.

When to use it

  • You already use Gemini and want the same Google key for your voice agent.
  • You take calls in one particular language and want the agent never to switch.
  • You want to choose among the Gemini voices, which speak any language.

Configuration

You need a Google Gemini credential with your Google AI key, the same one the Gemini model uses. See Credentials.

FieldWhat it does
Credential (required)The Google Gemini credential used to connect to the service.
Modelgemini-3.8-live (default) is the stable one. gemini-3.1-flash-live-preview is a preview that uses tools one at a time. gemini-2.5-flash-native-audio-preview-12-2025 is also available.
VoiceOne of the Gemini voices, such as Kore (default), Puck, Zephyr, Charon or Leda. All of them speak any language.
LanguageThe language it listens and answers in. Default: es-ES. With a fixed language, a noise is never taken for another language. Empty, Gemini detects it by itself, which can make it switch language and voice mid-call.

Advanced options

FieldWhat it does
Answer EagernessHow soon it decides the caller’s sentence is over: auto (default, Google’s own setting), low, medium or high. high answers sooner but may cut people off; low waits longer.

Output

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

Example

A clinic wants an agent that always answers in Spanish.

  1. Add Gemini Live and connect it to Realtime model on the Voice Agent Realtime.
  2. In Credential, choose the Google Gemini credential.
  3. Leave Model on gemini-3.8-live, pick a voice in Voice and leave Language on es-ES.
  4. Test the agent in the playground and listen to how it sounds.

Good practices

  • Set Language if all your calls are in the same language: that way the agent never switches language because of a noise.
  • Try several voices in the playground before deciding: all of them speak your language, but each has its own tone.
  • The Greeting and fixed farewells may come out with a word or two changed: the model says them its own way.
  • Response time varies. On a test call on our development server on 14 October 2026, gemini-3.8-live in es-ES took between 1.0 and 1.9 seconds from the moment the caller stopped speaking to the first sound of its answer. Check it in the playground with your own calls.
  • Google bills the usage to your own Google account. Check the current price there.