Microsoft · AI-103
Speech for agents
AI-103 · Lesson 13 of 15 · 1:40:21
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-103 exam (Azure AI Apps and Agents Developer Associate). In this lesson: speech to text and text to speech with the Azure Speech SDK (recognize_once vs continuous recognition, fast and batch transcription, LLM speech, SSML), custom speech and custom voices, speech as an agent modality (the Voice Live API connected to a Foundry agent, voice agents, the Speech MCP server), reasoning over audio (input_audio, Realtime API), and speech translation. Includes three original practice questions.
Preview the topics
- An app transcribes a forty-minute recorded call with recognize_once, and gets back only the first sentence. What should it use instead?
- Call center recordings are full of background noise and regional accents, and the word error rate is high. What improves recognition the most?
- A hotel wants a voice line where guests talk naturally to an existing Foundry agent, interrupt it, and call from noisy places. What should you use?