ICYMI: Google launches Gemini 3.1 Flash Live on AI Studio and Gemini

What's new? Gemini 3.1 Flash Live is available via the Gemini API and Google AI Studio for real time voice and vision agents; it cuts latency and isolates speech in 90+ languages;

· 1 min read
AI Studio

Google has introduced Gemini 3.1 Flash Live, now available to developers through the Gemini API and Google AI Studio. This release is designed for those building voice and vision agents that require fast, reliable, and natural conversations in real time. Targeting developers and businesses creating voice-first AI products, Gemini 3.1 Flash Live aims to improve latency, task completion in noisy environments, and instruction adherence. The model supports more than 90 languages and can process both audio and visual inputs, making it suitable for a global developer audience.

Compared to earlier models, Gemini 3.1 Flash Live offers lower latency and better acoustic understanding, distinguishing speech from background noise more effectively. It also surpasses the prior 2.5 Flash Native Audio model in recognizing pitch and pace, resulting in smoother conversations. With improved ability to follow complex system instructions, it maintains operational guardrails even during unpredictable conversations. Technical capabilities include robust session management for ongoing dialogues and ephemeral tokens for secure, short-lived sessions.

Google, the company behind this launch, is focusing on enabling production-ready, real-time AI systems. Developers are already integrating Gemini 3.1 Flash Live into applications such as:

  1. Design critique tools
  2. AI companion devices for older adults
  3. AI-powered RPG game masters

The API is built for scalability, supporting diverse use cases from live video streams to phone calls, and integrates with partner platforms for global deployment.

Source