Skip to main content
Gemini Live brings real-time, natural voice conversations to your Digital Twin using Google’s native audio API. Unlike text-based approaches that convert speech to text and back, Gemini Live processes audio natively — resulting in lower latency and more natural-sounding responses.
Gemini Live is ideal for interactive learning, tutoring, and conversational scenarios where low-latency, natural dialogue is important.

​
How It Works

User Speaks

Audio is captured from the user’s microphone and streamed over a WebSocket connection directly to Google’s Gemini API.

Gemini Processes

Gemini processes the audio natively — understanding speech, generating a response, and synthesizing audio output in a single model pass.

User Hears

The AI’s audio response streams back in real time. Both sides of the conversation are transcribed for the chat history.

​
Prerequisites

Before enabling Gemini Live, ensure:
  • A Gemini API key is available — either saved on this Digital Twin, or the platform-wide default provided by Praxis AI
  • Convo is switched on for the Digital Twin, with its Speech-to-Speech Model set to one of the Gemini Live models below — choosing one of those models is what switches the twin to Gemini for voice
Supported Gemini Live models (as the Speech-to-Speech Model picker shows them):

​
Admin Setup

1

Make sure there's a Gemini API key

Pria looks for the key in two places, in order:
  1. This Digital Twin — Gemini API Key, under API Keys on Configuration and Integrations (used if present)
  2. Platform-wide — a default key provided by Praxis AI (used when the twin has none)
You only need to add your own key if you want Gemini usage billed to your Google account.
This differs from standard Gemini chat, which can also use a key saved on an individual AI model. Gemini Live never does — only the two places above.
2

Turn Convo on

On Configuration and Integrations, under Realtime / Convo, tick Enable Convo (Speech-to-Speech). The same section holds Convo Enable Text Input and Convo Show Transcript, described under Features.
3

Choose Gemini as the voice

On Personalization and AI Models, in Convo (Live Voice & Avatar Calls), set Speech-to-Speech Model to one of the Gemini Live models above. Two Gemini-only settings then appear beneath it.
4

Decide who picks the voice

Gemini Voice starts on User’s Choice — each person picks a voice in Convo Mode, and their pick is remembered.Choose a specific voice instead and everyone hears that voice: the picker disappears from Convo Mode, so nobody can change it. Use that when the voice is part of your Digital Twin’s identity.Main voices:
The list has more than twenty further voices — Callirrhoe, Autonoe, Enceladus, Iapetus, Algieba, Despina, Gacrux and others — each labelled with its style. Preview Gemini voices on Google AI Studio.
5

Set how easy it is to interrupt

Interruption Sensitivity decides how readily the twin stops talking when someone speaks over it:
  • Low (default) — ignores background noise, but people have to speak up to cut in
  • High — easy to talk over, but may react to room noise
Keep Low for classrooms and open offices; try High for one-to-one use in a quiet room.

​
Features

During a voice conversation, the AI can invoke tools just like in text mode — searching the web, looking up files in IP Vault, querying Canvas, and more. Tool results are incorporated into the audio response naturally.When a connector tool needs approval, the microphone pauses and an approval card appears in the transcript (or over the call when the transcript is hidden). The twin cannot ask out loud on Gemini Live, so watch for the card: you have 45 seconds to approve or decline.
Both the user’s speech (input) and the AI’s spoken responses (output) are transcribed in real time and — when Convo Show Transcript is on — displayed in the Convo Mode panel. Transcripts are saved to conversation history for later review.
The microphone stream is captured with browser noise suppression enabled, and Gemini’s voice activity detection is tuned to filter brief background noises — improving recognition accuracy in noisy environments like classrooms or open offices.
When Convo Enable Text Input is on, users can type messages during an active voice conversation. The typed text is sent alongside the audio stream, and the AI responds in the conversation.
Each voice session uses a single-use, short-lived ephemeral token minted by the server and locked to the session’s model and configuration. No long-lived API keys are exposed to the browser.

​
User Experience

When a user opens Convo Mode with Gemini Live configured, they see:
  1. Voice selector — Dropdown with available Gemini voices (the choice is remembered per user). It isn’t shown when the administrator has fixed the voice
  2. Start/Stop button — Begins and ends the voice session
  3. Text input — Type during a voice conversation (when enabled by the administrator)
  4. Live transcript — Real-time display of both user speech and AI responses (when enabled)

​
Troubleshooting

  • Check Enable Convo (Speech-to-Speech) is on and the Speech-to-Speech Model is a Gemini Live model
  • If you saved your own Gemini API Key, check it’s valid — a wrong key on the twin is used in preference to the platform default
  • Ensure the browser has microphone permissions granted
  • Use headphones to prevent echo/feedback
  • Move to a quieter environment or closer to the microphone — brief background noises are filtered, but sustained noise degrades recognition
  • Check your internet connection — Gemini Live requires a stable WebSocket connection
  • Confirm the tools are switched on for the Digital Twin
  • Some tools may not be available during voice mode — check the admin Tools panel