Configure live avatars

This document describes how to configure live avatars in Gemini Live API. Live avatars lets you generate real-time video of a talking avatar synchronized with synthesized speech responses from the gemini-3.8-live model. You can choose from a variety of built-in avatars and pair them with supported prebuilt voices in both Cloud de Confiance console and Gemini Live API.

Configure live avatars in Cloud de Confiance console

Do the following:

  1. In the Cloud de Confiance console, go to the Agent Platform > Studio > Stream realtime page.

    Go to Stream realtime

  2. Click Switch model.

  3. From the models list, select gemini-3.8-live.

  4. In the main panel, select Live Avatar.

  5. From the Avatar list, select an avatar.

  6. From the Voice list, select a voice.

  7. Optional: Add system instructions to guide the behavior and output.

  8. Optional: Enable the camera input to share video with Gemini Live API.

  9. In the main panel, click Start Session.

Configure live avatars using the API

Do the following:

  1. In generation_config set "response_modalities": ["VIDEO"].

  2. In the avatar_config object, specify the prebuilt avatar name.

  3. In the speech_config object, specify the prebuilt voice.

WebSocket configuration

WebSocket connections are established with a standard WebSocket handshake. The endpoint is regional and uses OAuth 2.0 bearer tokens for authentication. The authentication token is typically passed in WebSocket headers, such as Authorization: Bearer [TOKEN].

The following is a sample WebSocket configuration:

import asyncio
import json
import websockets

# Replace PROJECT_ID and LOCATION with your Project ID and location.
PROJECT_ID = "PROJECT_ID"
LOCATION = "LOCATION"

# Authentication
token_list = !gcloud auth application-default print-access-token
ACCESS_TOKEN = token_list[0]

# Configuration
MODEL_ID = "gemini-3.8-live"
MODEL = f"projects/{PROJECT_ID}/locations/{LOCATION}/publishers/google/models/{MODEL_ID}"

config = {
    "response_modalities": ["VIDEO"],
    "speech_config": {
        "voice_config": {
            "prebuilt_voice_config": {
                "voice_name": "Puck",
            }
        }
    },
}

avatar_config = {
    "avatar_name": "Ben",
}

# Construct the WSS URL
HOST = f"{LOCATION}-aiplatform.googleapis.com"
URI = f"wss://{HOST}/ws/google.cloud.aiplatform.v1.LlmBidiService/BidiGenerateContent"

async def main():
    headers = {"Authorization": f"Bearer {ACCESS_TOKEN}"}

    async with websockets.connect(URI, additional_headers=headers) as ws:
        print("Session established.")

        # Send Setup (Handshake)
        await ws.send(json.dumps({
            "setup": {
                "model": MODEL,
                "generation_config": config,
                "avatar_config": avatar_config,
            }
        }))
        # Send and receive audio/video streams...

if __name__ == "__main__":
    asyncio.run(main())

Python SDK configuration

The following sample shows how to configure live avatars using the Google Gen AI SDK:

from google import genai
from google.genai import types

client = genai.Client()

config = types.LiveConnectConfig(
    response_modalities=["VIDEO"],
    speech_config=types.SpeechConfig(
        voice_config=types.VoiceConfig(
            prebuilt_voice_config=types.PrebuiltVoiceConfig(
                voice_name="Puck",
            )
        )
    ),
    avatar_config=types.AvatarConfig(
        avatar_name="Ben",
    ),
    system_instruction=types.Content(
        parts=[types.Part.from_text(text="You are a helpful customer service assistant.")]
    ),
)

async with client.aio.live.connect(model="gemini-3.8-live", config=config) as session:
    # Interact with the live session...
    pass

Use a custom avatar

Instead of choosing a prebuilt avatar, you can provide a reference image of a face, and Gemini Live API generates the avatar video from that likeness. Custom avatars are configured per session in the customized_avatar field of avatar_config.

Reference image requirements

The reference image must meet the following specifications:

  • Image format: PNG recommended. PNG uses lossless compression, and alpha transparency prevents a visible box around the avatar in light and dark modes.
  • Color mode: RGB.
  • Image size: 704 x 1280 pixels minimum (portrait), 1280 x 704 pixels minimum (landscape).
  • Resolution: 720p or higher.
  • Aspect ratio: 9:16 (portrait) and 16:9 (landscape) is standard.
  • File size: Less than 5 MB
  • Quality: No blur or compression artifacts.

For the best results, also follow these composition guidelines:

  • Framing: Use a bust shot, where the head and shoulders occupy more than 60% of the frame. Crop out the lower body, arms, hands, and any held objects such as microphones or phones.

  • Background: Use a simple, uncluttered backdrop with no other people or distinct objects.

  • Perspective: Face the camera directly, with the head level and the eyes looking into the lens.

  • Expression: Keep the expression neutral and at rest, with no smiling, squinting, or visible teeth.

  • Safety: Don't use images of minors, celebrities, or offensive content.

Configure a custom avatar in Cloud de Confiance console

Do the following:

  1. In the Cloud de Confiance console, go to the Agent Platform > Studio > Stream realtime page.

    Stream realtime

  2. From the models list, select gemini-3.8-live.

  3. In the main panel, click Live Avatar.

  4. Select Upload under the Avatar's options and upload an avatar image.

  5. Click Start Session, and then click Start conversation.

Configure a custom avatar using the API

Configure the custom avatar in the initial setup payload that you send over the WebSocket connection:

import base64
import json
import websockets

# Load custom avatar image (PNG format, 704x1280 min)
with open("/path/to/your/sample_avatar.png", "rb") as f:
    avatar_b64 = base64.b64encode(f.read()).decode("utf-8")

GENERATION_CONFIG = {
    "response_modalities": ["VIDEO"],
}

# Setup avatar config with custom face image
AVATAR_CONFIG = {
    "customized_avatar": {
        "image_data": avatar_b64,
        "image_mime_type": "png",
    }
}

# Session setup message
SETUP_MESSAGE = {
    "setup": {
        "model": "gemini-3.8-live",
        "generation_config": GENERATION_CONFIG,
        "avatar_config": AVATAR_CONFIG,
        "system_instruction": {
            "parts": [{"text": "Your system instruction here"}]
        },
        "input_audio_transcription": {},
        "output_audio_transcription": {},
    }
}

The preceding sample uses the following fields:

Field path Type Description
generation_config.response_modalities list[string] Must be set to ["VIDEO"] to enable avatar video output.
avatar_config.customized_avatar.image_data string Base64-encoded image string.
avatar_config.customized_avatar.image_mime_type string Image MIME format ("png").

You can pair a custom avatar with a prebuilt voice or with a custom voice.

Available prebuilt voices

You can pair prebuilt avatars with any of the prebuilt HD voices in Gemini Live API. For more information on prebuilt voices, see Prebuilt voices.

What's next