Amazon Polly
Amazon Polly for AIF-C01: text-to-speech with lifelike voices, lexicons, SSML markup, voice engines and speech marks for lip-sync and word highlighting.
DashboardAmazon Polly
- Turn text into lifelike speech using deep learning
- Provides many voices and languages
- Lexicons: we can customize pronunciations of specific words and phrases
- Define how to read certain specific pieces of text
- AWS => “Amazon Web Services”
- W3C => “World Wide Web Consortium”
- SSML format: Speech Synthesis Markup Language
- Markup for our text to indicate how to pronounce it
- Gives control over emphasis, pronunciations, breathing, whispering, speech rate, pitch, pauses
- Example:
<speak>Hello,<break />how are yoy?</speak>
- Voice engine: generative, long-form, neural, standard
- Speech Marks:
- Can encode when sentence/word starts or ends in the audio stream
- Useful for lip-synching animation or highlighting words as they’re spoken