Question

A developer is building a text-to-speech feature using Azure Speech in Foundry Tools for an audiobook app. The publisher wants certain passages read in a cheerful tone and others in a sad tone, using the same neural voice. What should the developer use to control these expressive styles in the synthesized output?

Accepted Answer

Pass Speech Synthesis Markup Language (SSML) with the mstts:express-as style attribute in the synthesis request

Answer

Deploy a separate custom neural voice model for each emotional tone

Answer

Increase the temperature parameter on the speech synthesis endpoint

Answer

Send plain text and rely on automatic sentiment detection to adjust the tone