Post by Lana Noor

With the latest Microsoft Foundry updates, including Voice Live API availability in Microsoft Foundry Next Generation and the release of newest Azure OpenAI voice model gpt-realtime-1.5, voice-to-voice conversational AI is becoming much easier to build for enterprise applications. In this article, I explore three different Microsoft PaaS approaches for building voice-enabled Agentic RAG systems: Azure AI Speech (STT + TTS), GPT-Realtime via WebRTC, and the Voice Live API. The article breaks down how each technology works architecturally and compares them across engineering complexity, latency, and ideal use cases, helping teams understand when to use each approach depending on the application requirements and customer experience goals.