Amazon 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