Amazon's text-to-speech voices get more natural, at the same price

Amazon has made the computer-generated voices in its Polly service sound more natural, and it did not raise the price. Polly turns written text into spoken audio. The upgrade is live now for businesses already using it, with no work on their end.

What changed

Amazon upgraded what it calls the Generative engine, the part of Polly that produces its most lifelike voices. It says the speech is now more expressive and more accurate, and still fast enough for live conversation. It also added 14 new voices across 9 locales. Hindi and Finnish get generative voices for the first time.

For anyone already using these voices, the voice names and the technical requests behind them are unchanged. Amazon calls it a drop-in upgrade. Existing apps keep working, and the better voices arrive in place.

Who it affects

Polly is built for companies that need spoken audio at large scale. Amazon lists several uses:

• Customer service phone menus that read back order numbers and addresses
• Live voice assistants that speak answers as they are written
• Audio for video, ads, podcasts and audiobooks
• Training courses that can be updated without re-recording
• Read-aloud tools for people who prefer to listen

Amazon points to accuracy as the key gain for phone systems. A voice that reads a tracking number or confirmation code correctly can help a caller finish without reaching a human agent.

A real example

Sanoma Media Finland uses Polly to turn news articles into audio. It says it publishes over 10,000 articles a month with a Finnish voice. Its product manager said the new Finnish generative voice, called Suvi, is a clear improvement over the older one and sounds noticeably more natural. That is one customer's view from its own testing.

What it costs

Amazon says the upgrade comes at the same price. There is no cost increase and no new pricing tier. Polly is pay-as-you-go, and the company says businesses can save and replay fixed audio prompts, such as a standard greeting, at no extra charge. The material does not give actual prices. Amazon points readers to its Polly pricing page.

What to do

Businesses that already use Polly generative voices do not need to act. To hear the difference first, Amazon offers sample audio. New users can reach the voices through the AWS Management Console, a command line tool, or software development kits. The material does not list the 14 new voices or the 9 locales beyond Hindi and Finnish.