Google has updated its Search Live experience in the United States with the Gemini 2.5 Flash Native Audio model. The same audio technology is also extending to other Gemini-powered products, including Gemini Live, Google AI Studio, and Vertex AI.
Google updates Search Live with Gemini model upgrade
Google announced that Search Live now uses the Gemini 2.5 Flash Native Audio model for real-time spoken interactions inside Google Search. Users can have back-and-forth voice conversations in AI Mode and hear more natural-sounding responses.
According to Google, the updated voice experience is rolling out this week in the United States. Users can also slow down voice responses, which Google positions as helpful for instructional content and step-by-step explanations.
Key details:
- Product updated: Search Live within Google Search.
- Model: Gemini 2.5 Flash Native Audio.
- Capabilities: Real-time voice conversations in AI Mode with more fluid, expressive spoken replies.
- Control: Option to slow spoken responses for clearer guidance.
- Geography: Rollout starting in the United States this week.
"When you go Live with Search, you can have a back-and-forth voice conversation in AI Mode to get real-time help and quickly find relevant sites across the web. And now, thanks to our latest Gemini model for native audio, the responses on Search Live will be more fluid and expressive than ever before."
Google outlined the changes in its Search blog post, You can now have more fluid and expressive conversations when you go Live with Search.
Broader Gemini native audio rollout
Google positions the Search Live upgrade as part of a wider Gemini 2.5 Flash Native Audio deployment. The same native audio model is becoming available in Gemini Live within the Gemini app, in Google AI Studio, and through Vertex AI.
In its Gemini blog post, Improved Gemini audio models for powerful voice interactions, Google describes Gemini 2.5 Flash Native Audio as processing spoken audio in real time and producing fluent spoken output. The company says the updated model more consistently triggers external functions during conversations, follows complex instructions, and maintains context across multiple dialogue turns.
Google is also adding live speech-to-speech translation built on the Gemini audio model. The system can continuously translate ambient speech into a target language or support two-way conversations between speakers of different languages.
According to Google, the translation feature preserves vocal characteristics such as rhythm and emphasis so translated speech stays conversational. The company highlights broad language coverage, automatic language detection, multilingual input support, and noise handling as core capabilities.
Background context and sources
Google presents Search Live as a way to use voice, camera, and Gemini to access web results and get real-time help with what is happening around a user. The experience combines camera and microphone input so people can talk with Gemini about what they are seeing while still reaching relevant sites across the web. For practical examples, Google offers 5 ways to get real-time help by going Live with Search on its Search blog.
In an earlier announcement on voice search, Google also described a research model called Speech-to-Retrieval (S2R), a neural network trained on large datasets of paired audio queries. That work, summarized in Google announces a new era for voice search, provides additional context for how audio-native models are influencing Search.
Official sources:
.svg)
.svg)
.svg)