Google is making its Gemini voice model more fluid, natural and precise while lowering latency and improving precision with the launch of Gemini 3.1 Flash Live.
Gemini 3.1 Flash Live powers the recently announced Search Live, as well as Gemini Live. Gemini Live now offers faster responses compared to the previous model, and can follow your conversation for twice as long, meaning you can have even longer brainstorms.
Gemini 3.1 Flash Live is watermarked with SynthID, which is interwoven into the audio output, and allows people to detect that the audio is AI-generated.
The new voice model is also available via the Gemini live API in Google AI Studio, allowing users to develop voice-first agents that can complete complex tasks. And for Gemini Enterprise for Customer Experience, the voice model is better at recognizing acoustic nuances like pitch and pace, and can change dynamically to adjust responses to users’ expressions of frustration or confusion.
Source: Google Blog
