Google has announced the launch of Gemini 3.1 Flash Live, hailed as the current "highest-quality audio and voice model," designed to power major upgrades for Gemini Live and Search Live. The model is available in preview through the Gemini Live API on Google AI Studio, featuring lower latency compared to the 2.5 Flash Native Audio version and enhanced ability to recognize acoustic details like pitch and speech rate.
In real-world conversation, Gemini 3.1 Flash Live excels at distinguishing and extracting valid human voices from ambient sounds like traffic or TV, while more effectively filtering background noise. Google notes that Gemini Live on Android and iOS will deliver faster responses with fewer awkward pauses, and it can now maintain conversation context for twice as long to support extended brainstorming and follow-up questions. The system also dynamically adjusts answer length and tone to fit the interaction.
