Google DeepMind updated its voice model to reduce latency and increase precision. The changes aim to make real-time interactions feel more fluid and natural. This is an incremental performance boost for Gemini 3.1 Flash Live. Developers can now expect tighter response times and higher reliability during live audio streams.