AI-901 ยท Text and speech solutions
Voices, speech translation and SSML
Choosing neural voices and languages, speech translation, and shaping synthesized speech with SSML such as pauses, rate and pitch.
Key points
- Voice: speech_config.speech_synthesis_voice_name or <voice name=...> in SSML.
- Neural voices speak other-language text with an accent, so match voice to text.
- Translation: SpeechTranslationConfig, speech_recognition_language = source locale, add_target_language(code) per target, TranslationRecognizer, result.translations[code].
- SSML: <speak> root, <voice>, <break> for pauses, <prosody rate|pitch|volume> for delivery.
Source: https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-synthesis-markup-voice
Practice questions
Which SSML element inserts a pause between words?
- <prosody>
- <voice>
- <break>
- <emphasis>
Show answer
Answer: C. <break>
<break> adds silence. <prosody> changes rate, pitch or volume, and <voice> selects the speaker.
Which attributes can the SSML <prosody> element adjust?
- rate
- pitch
- volume
- locale
- targetLanguage
Show answer
Answer: A. rate, B. pitch, C. volume
<prosody> controls rate, pitch and volume (plus contour). Locale and translation settings are not prosody attributes.
If you give a neural voice text in another language, it automatically translates the text into the voice's own language.
Show answer
Answer: False
False. Synthesis does not translate. The voice reads the text as written, typically with an accent. Translation is a separate capability.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
