Build real-time voice & vision agents with our lowest latency yet

Email not displaying correctly? View it online

Build real-time conversational voice & vision agents

Hi Jacek,

We just launched the first real-time audio model in the Gemini 3 family giving you the tools to build more conversational, multimodal agents. The upgrades include:

Higher task completion rates in noisy environments: The model now performs significantly better in function calling and tool use in real-world scenarios.
Improved instruction following: The model now adheres to complex system instructions a lot better, even when the conversation takes unexpected turns.
More natural and low-latency dialogue: The model is much more effective at recognizing acoustic nuances like pitch and pace, and is now a lot faster in response time.

Speak to vibe code and experience real-time instruction following and tool-calling capabilities

Try the demo app

Get started in minutes

You can use our GenAI SDKs on your server, connect directly to the WebSocket from your client, or use one of our partner integrations for WebRTC support.

import asyncio
from google import genai
 
 client = genai.Client(api_key="YOUR_API_KEY") 
 
model = "gemini-3.1-flash-live-preview"
config = {"response_modalities": ["AUDIO"]}
 
 async def main(): 
    async with client.aio.live.connect(model=model, config=config) as session:
        print("Session started")
        # Send content...
 
 if __name__ == "__main__": 
    asyncio.run(main())

Ready to start building?

Documentation

Explore code samples, technical details, and best practices.

Read the docs

Examples on GitHub

Check out our example apps to get up and running in no time.

Clone example apps

Happy building,

The Google AI Studio team

Prompt models

Manage API keys

Vibe code

Get the latest product updates

Upcoming global events

This email was sent to jaczad@gmail.com because you signed up to receive emails about Google AI. If you do not wish to receive these emails, please unsubscribe.