Google expands its Search Live multimodal AI to over 200 countries and territories, enabling real-time, voice- and camera-enabled conversations alongside traditional search results.

The rollout introduces Gemini 3.1 Flash Live, a new audio model designed for natural, multilingual interactions and rapid responses within the Google app on Android and iOS.

A standout feature is camera-powered real-time assistance using the phone camera and Google Lens to answer questions about real-world objects, including setup instructions for gadgets or appliances.

Performance benchmarks show improved handling of multi-step instructions and real-world conversational tasks, with better understanding, instruction-following, and consistency.

Gemini 3.1 Flash Live offers enhanced tonal understanding, finer acoustic nuance recognition (pitch and pace), and dynamic response adaptation to user expressions such as frustration or confusion for voice-first use in noisy environments.

The model emphasizes natural-feeling conversations with quicker responses and sustained context over longer exchanges.

It aims to reduce response times, improve handling of long and complex queries, and maintain conversational continuity for more human-like interactions.

Users can build voice-ready agents to handle complex tasks at scale, with validation from companies like Verizon, LiveKit, and The Home Depot praising its improved natural conversations.

The feature prioritizes language inclusivity by supporting multiple languages to broaden digital participation in diverse markets.

All audio generated by Gemini 3.1 Flash Live includes an inaudible SynthID watermark to identify AI-generated audio and combat misinformation.

A SynthID watermark is embedded in AI-generated audio to enhance transparency and curb misinformation.

Every audio output from 3.1 Flash Live carries the SynthID watermark to aid detection of AI-generated content and support safety efforts.