Amazon Polly is a fully managed Amazon Web Services product that converts text into spoken audio on demand. It uses deep learning to produce lifelike voices in dozens of languages and offers Standard, Neural, Long-Form and Generative voice engines. Developers call it through an API, then store the resulting MP3 or OGG files for replay at no extra cost. SSML markup and custom lexicons let you control emphasis, intonation, phrasing and the pronunciation of acronyms or company names. Billing is pay-as-you-go by character: Standard voices cost $4.00 per million characters, Neural $16.00, Generative $30 and Long-Form $100.00, with free tier allowances for new accounts.
Polly fits software developers and engineering teams that need to add speech to an application without building a voice model of their own. Product teams building mobile and IoT apps use it, as do contact center architects who need spoken prompts for automated phone menus. Media producers and e-learning teams with scripts to narrate can also work with it, as can accessibility specialists who want written content available as audio. Because it runs inside AWS, teams already on that cloud will find billing, security and access control familiar, and AWS says it does not retain the text you submit.
Typical uses include reading articles, RSS feeds and web pages aloud, and generating voiceovers for animations, games and training videos directly from scripts. Companies feed it text for interactive voice response systems that prompt callers, and reuse cached audio to keep costs low on repeated phrases. Multilingual projects can produce the same message in several languages, with speech duration adjusted to help with dubbing. Newsrooms convert written stories into audio editions, and educators turn course material into spoken lessons for learners who prefer listening. Speech Marks output adds timing metadata for synchronizing audio with highlighted text or animation.
