Amazon Polly launches voice cloning capability

Executive Briefing
“Amazon introduced voice cloning in Amazon Polly, allowing organizations to build custom...”
Key Takeaways
- 1.Amazon Polly builds custom voices from one to ten audio recordings provided by a consenting speaker.
- 2.The feature supports seven languages including English, Spanish, German, French, Italian, Portuguese, and Hindi, with each cloned voice speaking the language it was recorded in.
- 3.Voice creation is asynchronous and typically takes 5 to 10 minutes, with free voice creation and pay-as-you-go pricing for usage.
- 4.Cloned voices are automatically encrypted and access-controlled through AWS IAM, and generated speech carries an automatic watermark.
Amazon introduced voice cloning in Amazon Polly, allowing organizations to build custom text-to-speech voices from a few minutes of clean source recordings. The capability is now generally available as a gated release across AWS regions supporting the generative engine.
Product Features & Key Highlights
Amazon Polly builds custom voices from one to ten audio recordings provided by a consenting speaker.
The feature supports seven languages including English, Spanish, German, French, Italian, Portuguese, and Hindi, with each cloned voice speaking the language it was recorded in.
Voice creation is asynchronous and typically takes 5 to 10 minutes, with free voice creation and pay-as-you-go pricing for usage.
Cloned voices are automatically encrypted and access-controlled through AWS IAM, and generated speech carries an automatic watermark.
Practical Assessment & Evaluation
Integrating voice cloning directly into the Amazon Polly generative engine helps enterprises maintain a consistent brand persona across interaction channels without repeatedly booking studio sessions.
Mandating speaker consent confirmation via API parameters or the console, alongside customer-managed encryption keys, helps mitigate the risks of audio impersonation and unauthorized cloning.
Pricing, Availability & Next Steps
The initial quota of 10 cloned voices per region can be increased by contacting AWS support, but specific approval timelines and evaluation criteria remain unspecified.
The real-world accuracy and handling of complex domain-specific terminology in commercial environments remain unverified outside of the provider's standard benchmarks.







