Amazon Polly adds gated voice cloning in seven languages

Amazon Web Services has made voice cloning in Amazon Polly generally available, but only as a gated release that approved accounts must be granted access to. The feature builds a custom text-to-speech voice from a few minutes of recordings of one speaker who has agreed to it.
What the feature does
Customers supply a few minutes of audio from a speaker who has consented to having their voice cloned. Amazon Polly uses those recordings to build a custom voice that can read new text in that speaker's voice. The cloned voices run on the Amazon Polly generative engine, which AWS says delivers naturalness and accuracy similar to its other generative voices.
Recordings can be sent directly in the CreateVoice request or uploaded through the console. Each cloned voice speaks the language in which it was recorded.
Languages and regional availability
At launch, voice cloning supports seven languages: English, Spanish, German, French, Italian, Portuguese and Hindi. It is available in all AWS Regions where the Amazon Polly generative engine runs, including the AWS GovCloud (US) Region.
Access, security and pricing
AWS announced the release on November 19, 2026. Access is enabled per AWS account. Customers already selected for access can use the feature now, while others can contact their AWS account team or apply for access.
AWS says the cloned voices are automatically encrypted and access-controlled, so that only the customer can use them. Building a cloned voice is free; usage of cloned voices is charged on a pay-as-you-go basis, according to the Amazon Polly pricing page.
Intended uses
AWS points to a consistent brand voice across interactive voice response systems, contact centres, apps, in-store announcements and connected devices, so prompts can be regenerated when scripts change without a new studio session. It also names audiobooks, courses, podcasts, game characters and videos, as well as localised content maintained in one voice identity per language and reading aids for accessibility.
AWS's own demonstrations include a customer-service recording and a game character. These samples come from AWS and its authors, and the company has not published independent evaluations of them.
Open questions
- What criteria does AWS apply when deciding which accounts receive access?
- How will consent from the recorded speaker be verified and documented?