AWS has upgraded the generative speech engine in Amazon Polly, its text-to-speech service. The company says the new engine produces more natural, expressive and accurate speech, with low latency. It is generally available in production, and existing applications using generative voices receive the upgrade without changes to their code or API calls.
The update adds 14 voices across nine locales. Hindi and Finnish are available on Polly’s generative engine for the first time. AWS says the catalog now has 57 generative voices across 11 languages and 25 locales, and the service is available in 11 AWS Regions.
What changes for developers
Applications can continue using the same Amazon Polly API. AWS says the upgrade costs the same as before, with no new product SKU. Developers can also use the service’s streaming API to send text and receive audio at the same time, a setup intended for real-time voice applications.
AWS says the voices can better reflect emotion and speaking style based on the text, without manual setup. The company offers sample audio, including a voice reading an order confirmation with a tracking number, address, date and price. These examples illustrate the intended improvements, but AWS does not provide independent benchmark results in the post.
The 14 new voices are available in the US East (N. Virginia), US West (Oregon), Europe (Frankfurt, Zurich and London), Asia Pacific (Sydney, Seoul, Singapore and Tokyo), Canada (Central) and AWS GovCloud (US-West). Voice availability can vary by Region, so AWS recommends checking its current list when deploying.
Limits and customer feedback
Standard and neural voices are unaffected. Generative voices still do not support speech marks, a feature that aligns generated speech with its text; standard and neural voices continue to support them. Generative voices do support SSML, a way to add some control over speech output using markup.
Sanoma Media Finland says it publishes more than 10,000 real-time news articles a month using Polly’s Finnish neural voice. Its product manager said the new Finnish generative voice sounded noticeably more natural in the company’s testing. That is customer feedback, not a published independent comparison.
Why it matters
Developers already using Polly’s generative voices can get the updated engine without changing their integrations or paying more, according to AWS. Hindi and Finnish voice options may also help teams producing spoken content for those language audiences, though voice availability depends on the AWS Region and the post gives no independent performance measurements.
