Ref: https://learn.cantrill.io/courses/1820301/lectures/42176865

• Summary:
• Contents:

Amazon Polly 101

• 🔧 Text-to-Speech (TTS) service
  • ❗ NO translation done in the conversion, same language!
• Voice engines:
  1. Standard TTS = Concatenative (concatenates phonemes)
     • 💡 Phoneme = smallest unit of sound in a language
  2. Neural TTS = Phonemes → spectograms → vocoder → audio
     • Much more human/natural sounding, but much more complex and computationally heavy
  3. Newer engines: long-form, generative (more powerful)
• Features
  • Integration with other services and apps
    • e.g. WordPress plugin to read WordPress articles out loud
  • Many output formats supported (MP3, PCM, Ogg Vorbis…)
  • Supports Speech Synthesis Markup Language (SSML) → markup tags provide additional control over how speech is generated
    • e.g. emphasis, pronunciation, whispering, over-exaggerated “newscaster” speaking style
  • Lexicons: define how to read certain specific text
    • e.g. “AWS → Amazon Web Services”
  • Speech mark: encode where a sentence/word starts or ends in audio
    • helpful for e.g. lip-syncing or word highlighting
• Screenshot