Amazon Polly now clones voices as a gated, consent-based feature in seven languages
AWS announcement, 19 November 2026: Amazon Polly voice cloning is generally available as a gated release. A few minutes of recordings from a consenting speaker produce a cloned voice that runs on Polly's generative engine in English, Spanish, German, French, Italian, Portuguese or Hindi. Building a voice is free; usage is pay-as-you-go.
AWS announced on 19 November 2026 that voice cloning in Amazon Polly is now generally available as a gated release. Customers give Polly a few minutes of audio recordings from a speaker who has agreed to be cloned, and Polly builds a cloned voice that runs on its generative engine. The launch covers seven languages: English, Spanish, German, French, Italian, Portuguese and Hindi, with each cloned voice speaking the language it was recorded in.
Gated access and consent mechanics
Access is enabled for select existing Polly customers; others apply through their AWS account team or an access request. Per the AWS Machine Learning blog, cloned voices are automatically encrypted and access-controlled so only the owning customer can use them, customers confirm the speaker's consent each time a voice is created, and AWS deletes its copy of the source recordings after voice creation. The feature is available in all AWS Regions that support the Polly generative engine, including GovCloud (US). Building a cloned voice is free; synthesizing speech with it is billed pay-as-you-go.
What AWS pitches it for
AWS positions the feature for a consistent brand voice across IVR systems, voice agents, apps and advertising; contact centers that can update prompts without new studio sessions; media production such as audiobooks, courses and game characters; localization with a consistent persona per language; and accessibility uses. The launch post includes before-and-after audio from named Amazon staff and AWS chief evangelist Jeff Barr, each a real recording followed by their cloned voice reading text they never recorded.
Why the gating matters
Voice cloning is one of the AI capabilities most associated with impersonation fraud, so the launch mechanics are part of the story: a named-consent requirement per voice, gated onboarding rather than self-serve availability, and per-customer isolation of the resulting voices. AWS is not the first major cloud to offer custom synthetic voices, but the gated, consent-confirmed design is the pitch: clone your own spokesperson, not anyone else.