What changed
Amazon Polly has introduced a major upgrade to its Generative engine, delivering more natural and accurate speech output with low latency. The update includes 14 new voices across 9 locales, with Hindi and Finnish now generally available for the first time. The upgrade is fully backward compatible, meaning existing applications can continue to use the same API calls and voice names without code changes.
What this means in practice
The new voices, including Arjun and Kiara for Hindi, and Suvi for Finnish, enhance the ability to deliver expressive and context-aware speech. These voices can reflect emotion and prosody based on the text input, improving the realism of automated interactions. The upgrade also expands the total number of generative voices to 57 across 11 languages and 25 locales.
The improved accuracy in speech generation is particularly beneficial for applications requiring precise pronunciation of numbers, currency, addresses, and other entities. For example, the new voices can accurately read back order confirmations, including tracking numbers, dates, and times. Additionally, the upgrade supports real-time applications through the Bi-Di Streaming API, enabling simultaneous text-to-speech synthesis.
Limitations and availability
The upgrade is now generally available in production across 11 AWS Regions, including AWS GovCloud (US-West). The new voices are accessible via the SynthesizeSpeech API, and no changes to existing integrations are required. Pricing remains on a pay-as-you-go basis with no cost increase. The supplied announcement does not specify exact performance benchmarks or independent testing results.
