Should your app wait for a clip or play audio as it arrives?

For a fixed announcement, send the text and save the returned audio file. For a reply that needs to begin sooner, Sonic can stream audio in pieces. Your app needs to play those pieces in order and handle interruptions. If the user stops a reply, stop playback and give them the text version.

What does the hosted speech service receive?

The service receives the text to speak and the voice selection. Leave unnecessary personal details out of the text. Before sending sensitive text, confirm how the service handles it under your plan. For a browser feature, Cartesia documents short-lived access tokens and says to keep the account API key on your server.

Choose a voice the service provides and check your usage rights. Cartesia requires verified speaker consent for voice clones.

What counts against your Sonic allowance?

Cartesia lists one credit per character for Sonic, with a break tag counting as a credit. The Free card lists monthly credits. The Pro card adds commercial-use rights. Check your plan and output rights before using speech in a customer product.

How would you handle someone stopping a spoken reply?

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

Write the reply
Use a fictional two-sentence update about a delivery. Mark which text should always stay readable.
Split the audio on paper
Put the first sentence on one card and the second on another. Say when playback begins and when it should stop.
Write the expected result
If the user stops during sentence one, mark that sentence two stays silent and the full text is still readable.