Launch of Gemini 3.1 Flash Live: Google's New AI Audio Model

Google has introduced its latest audio model, Gemini 3.1 Flash Live, which aims to make voice interactions more natural and reliable. This model offers improved accuracy and reduced latency, resulting in smoother and more precise voice interactions.

Gemini 3.1 Flash Live is designed for real-time dialogue and is available to developers through the Gemini Live API in Google AI Studio, while enterprises can leverage it to enhance customer experience. Users worldwide can experience it via Search Live and Gemini Live, now supporting over 200 countries.

The model showcases significant improvements in tonal understanding, enabling more natural conversations. It allows developers to build voice agents capable of handling complex tasks more reliably. On the ComplexFuncBench Audio benchmark, it outperformed its predecessor.

Gemini 3.1 Flash Live has also enhanced its recognition of acoustic nuances such as pitch and pace, making interactions more intuitive. Companies like Verizon and The Home Depot have already noted the positive impact of the model on their workflows.

With the launch of Gemini 3.1 Flash Live, users can now engage in multilingual real-time conversations, opening new avenues for communication. All audio generated by this model is watermarked with SynthID, helping to prevent the spread of misinformation.