shiichan

Amazon Polly Generative engine upgrade: new voices, plus Hindi and Finnish!

Hi, it's Shii-chan! Today's news is about Amazon Polly voices getting more natural!

AWS What's New aws.amazon.com

What was announced?

According to AWS What's New, there is a new model behind the Amazon Polly Generative engine. It delivers more natural-sounding, accurate speech with low latency.

This is a drop-in model upgrade, and SSML tags work just as before.

The story so far

Polly is a fully managed service that turns text into natural-sounding speech. But complex content such as numbers, email addresses, and street addresses sometimes needed fine-tuning.

This upgrade is designed to make that kind of content sound more reliable, so you need less tuning of the speech output!

What changes

  • Higher overall voice quality and a richer voice portfolio
  • Hindi and Finnish are generally available on the Generative engine for the first time
  • The low-latency streaming API makes it easier to build real-time apps such as contact center experiences, chatbots, and LLM-based voice agents
  • It also suits news, media, entertainment, e-learning, and accessibility use cases

Dive Deep

The announcement says there are 14 new voices across nine locales. These are the voices it names:

  • US English: Kenneth, Nate
  • Australian English: Russell, Curtis
  • Hindi and Indian English: Kiara, Arjun
  • German: Tina
  • US Spanish: Carlos
  • Brazilian Portuguese: Leo, Carolina
  • Korean: Hyunduk
  • Finnish: Suvi

The new Generative voices are available in these Regions:

  • US East (N. Virginia)
  • US West (Oregon)
  • Europe (Frankfurt)
  • Europe (Zurich)
  • Europe (London)
  • Asia Pacific (Sydney)
  • Asia Pacific (Seoul)
  • Asia Pacific (Singapore)
  • Asia Pacific (Tokyo)
  • Canada (Central)
  • AWS GovCloud (US-West)

The Generative offering now reaches AWS GovCloud (US-West), which helps public sector and regulated industries. The announcement does not include pricing, and points to the Amazon Polly pricing page.

Wrap-up

  • A new Generative engine model brings more natural, accurate, low-latency speech
  • Numbers, email addresses, and addresses are designed to sound more reliable
  • Hindi and Finnish reach GA on the Generative engine for the first time
  • New voices across nine locales, in 11 Regions including AWS GovCloud (US-West)
  • SSML works as before

This is for you if you build voice agents or contact center apps, or need multilingual text-to-speech!