Google Gemini 3.8 Live enables smooth voice control and simultaneous task execution

Google is introducing new voice interaction models aimed at developers, businesses, and consumers. The basic version, Gemini 3.8 Live, aims for efficiency and flexibility. The more powerful version, Gemini 3.8 Live Extended Thinking, is tailored to solve extremely demanding multi-step tasks.

Gemini 3.8 Live enables near real-time processing of visual input, seamlessly adapting and switching between 97 supported languages during a conversation. A unique feature of the solution is the ability to run tools and call application programming interfaces (APIs) in the background without interrupting the voice conversation. This allows the user to continue communicating while the system completes complex tasks in the background.

For more complex workflows, Extended Thinking offers deep thinking, using natural speech responses to validate requests and explain progress in real time. This model ranks first on the Speech to Speech Quality Index with a score of 82.6 and scores 97.7 % on the Big Bench Audio test. The integration is already available in Google Workspace services, including Docs Live, Gmail Live, and Keep Live, as well as Search Live.

To ensure security and prevent misinformation, a SynthID watermark is seamlessly embedded into all generated audio. Although the new feature promises smooth voice control and high accuracy, it is worth being a little cautious when using it in production environments and testing its operation thoroughly.