AWS upgrades Amazon Polly Generative engine with 14 new voices

AWS has made an upgraded Amazon Polly Generative engine generally available, adding 14 voices across nine locales while keeping existing voice names and API calls unchanged. Applications already using generative voices do not need code changes for the in-place upgrade.

The additions expand Polly’s generative catalog to 57 voices across 11 languages and 25 locales, according to AWS. Hindi and Finnish are available in the Generative engine for the first time, with Arjun and Kiara for Hindi and Suvi for Finnish. The remaining voices cover US, Australian and Indian English, German, US Spanish, Brazilian Portuguese and Korean.

AWS also expanded the engine to AWS GovCloud (US-West), taking announced availability to 11 Regions. All 14 new voices are available across those Regions. Existing applications can continue using SynthesizeSpeech, while the bidirectional StartSpeechSynthesisStream API accepts text and returns synthesized audio at the same time. Standard and neural voices are unaffected. Generative voices still do not support speech marks, and supported SSML tags remain available.

AWS said the upgraded model improves expressiveness and accuracy while retaining low latency. The reviewed evidence, however, included no independent benchmark or methodology for those claims. The company also said the upgrade brings no price increase, new SKU or integration change.

At the reporting cutoff, AWS documentation still listed 43 generative voices and 10 commercial Regions. The October 22 announcement listed 57 voices and 11 Regions, leaving the guide out of sync with the launch announcement.