Google’s Gemini AI is enhancing its conversational capabilities with the latest update, Gemini 2.5 Flash Native Audio, specifically aimed at live voice agents. Key improvements focus on three areas:
-
Sharper Function Calling: Enhanced reliability allows Gemini to trigger external functions while gathering real-time information, ensuring a seamless conversation flow.
-
Improved Instruction Following: The adherence rate to developer instructions has risen from 84% to 90%, enabling Gemini to manage complex instructions more effectively.
-
Smoother Conversations: The AI now retrieves context more efficiently, leading to more cohesive interactions.
Additionally, Gemini Live won’t interrupt users mid-sentence for pauses, and users can mute their microphone while the AI speaks. These updates enhance Gemini’s capability to handle intricate workflows and engage in natural-sounding dialogues. The rollout affects Gemini Live, Search Live, Google AI Studio, and Vertex AI, alongside announced changes for the Translate app.
