Google launched Gemini 3.1 Flash Live, a new audio-first AI model designed for real-time voice conversations. The model supports over 90 languages, filters background noise more effectively, and maintains conversation context for twice as long as previous versions.

What Happened

Gemini 3.1 Flash Live is Google's highest-quality model for audio and voice interactions. It improves on the 2.5 Flash Native Audio model with better acoustic recognition, including pitch and pace detection, and stronger background noise filtering. The model is available now across Google products and in preview via the Gemini Live API in Google AI Studio.

Google is also using 3.1 Flash Live to roll out Search Live globally to over 200 countries, enabling voice-driven search with real-time AI responses.

Why It Matters

Voice-first AI has struggled with two problems: conversations that feel robotic, and models that lose context mid-discussion. Flash Live tackles both. The improved tonal understanding produces more natural dialogue rhythm, while doubled conversation length means longer brainstorming sessions stay coherent.

For creators building voice-powered applications, the model also delivers significantly improved tool calling, which means AI agents can trigger external actions (fetching data, running workflows) during live voice sessions without breaking conversation flow.

Key Details

  • Languages: 90+ for real-time multimodal conversations
  • Conversation length: 2x longer context than previous versions
  • Noise handling: Improved background noise filtering
  • Tool calling: Enhanced ability to trigger external tools during live conversations
  • Safety: All generated audio watermarked with SynthID
  • Access: Gemini Live API in Google AI Studio (preview)

The launch coincides with Google making Gemini Live available as a standalone experience across more devices and regions.

What to Do Next

Developers building voice agents or real-time audio applications should evaluate Flash Live against their current stack. The API preview in Google AI Studio provides direct access for testing. For context on Google's broader AI rollout, see our coverage of Gemini 3.1 Pro and Google Lyria 3 Pro.