- Published on
Microsoft Foundry Introduces Native Real-Time Voice Agents, Directly Challenging Google Gemini Live
- Authors
- Name
- DeepHorus
Microsoft is further strengthening the real-time voice capabilities of the Foundry platform, providing developers with native voice agents that can directly accept speech input and produce speech output, making it easier for enterprises to build real-time conversational AI experiences similar to Google Gemini Live.
The key change in this Microsoft update is that Foundry Agent Service now supports native real-time voice capabilities. Unlike traditional AI agents, which first convert speech to text, pass it to a text model for processing, and then convert the text back into speech, the native voice architecture allows real-time voice models to directly process users' speech input and generate spoken responses, thereby reducing intermediate steps and the resulting latency.
For users, this means AI agents can hold continuous conversations more naturally. Users do not need to wait for speech recognition to finish first, nor do they need to wait for the model to generate a complete written answer before speech synthesis begins. The system can operate more like real-time communication between people and can handle key aspects of voice interactions, such as interruptions and turn detection.
This new capability from Microsoft is built on Voice Live API. This interface brings together speech recognition, generative AI, text-to-speech, turn detection, and interruption handling in a unified real-time voice interface, allowing developers to add real-time voice interaction capabilities to their AI agents without having to build multiple voice components separately.
