In AI-901, this objective means understanding how a simple app can use Azure Speech in Foundry Tools for speech input or speech output.
Microsoft lists this exact skill under Implement AI solutions for text and speech by using Foundry in the AI-901 study guide.
What Azure Speech in Foundry Tools does
Azure Speech supports common speech workloads such as:
Requirement Capability
Convert spoken audio into text Speech to text
Convert written text into spoken audio Text to speech
Translate spoken language Speech translation
Identify speakers in audio Diarization
Microsoft describes Foundry Tools as letting you try Azure Speech features in a web interface, with authentication, resource management, and deployment handled in the Foundry portal.
Lightweight app pattern
A lightweight app is usually a small console app, script, or simple web app.
User speaks / enters text
↓
Lightweight application
↓
Azure Speech in Foundry Tools
↓
Speech-to-text or text-to-speech result
↓
Application displays or plays result
Common exam scenarios
1. Speech to text
Use when the app must transcribe audio.
Example:
“A company wants to convert call centre recordings into written transcripts.”
Correct capability:
Speech to text
Microsoft’s speech-to-text quickstart covers real-time transcription and options such as language, profanity policy, speaker diarization, and viewing transcript/JSON output.
2. Text to speech
Use when the app must generate spoken audio from text.
Example:
“An app needs to read alerts aloud to users.”
Correct capability:
Text to speech
Microsoft’s text-to-speech quickstart covers converting text into speech, supported audio formats, voice options, and SSML for finer control.
Key AI-901 distinction
If the question says… Choose…
“Transcribe”, “captions”, “spoken words to text” Speech to text
“Read aloud”, “spoken response”, “generate voice” Text to speech
“Different speakers in a meeting” Diarization
“Convert speech from one language to another” Speech translation
Summary
For AI-901, remember:
Azure Speech in Foundry Tools = build simple apps that recognise speech, synthesize speech, or translate speech.
The lightweight app does not train the model. It simply calls the Speech capability and returns the result.