What do you send to make a spoken announcement?

Choose a voice and its engine, the speech model that produces the audio. Provide the text and choose an output format such as MP3. Check that your chosen voice and engine are available in your AWS region. Keep the plain text so people can read the announcement too.

Do repeat plays require another synthesis request?

You can save the generated clip and play it again without asking Polly to generate it each time. Your host may charge separately for storing or delivering audio. Keep the script beside the audio file so you know which announcement the clip contains.

What changes what you pay?

AWS lists different synthesis rates by voice engine, priced per amount of text processed. Estimate how many characters you will send each time the announcement changes. Check credit eligibility and expiration separately from the listed rates.

What should you check before sending private text?

Polly is a hosted service. AWS says it may store and use text inputs for service maintenance and improvement, and documents an AWS Organizations opt-out policy for improvement. Check the policy that applies to your AWS account before sending anything sensitive. A local playback file does not change where the original synthesis happened.

How would you update an announcement without regenerating every time?

Sketch this with fictional material. Leave the real site, accounts and services alone.

Write two versions
Draft fictional opening hours for a shop and then change its closing time.
Label each audio file
Give the old and new script distinct names so your app can select the right clip.
Mark when generation happens
Draw one new synthesis request when the script changes and several later plays of the saved file.