Speech & transcription

Use speech generation for character voice messages and audio content. Use transcription to turn recorded user audio into text for your conversation workflow.

Speech generation

A typical speech job receives text and an assigned voice, then returns an audio result. The voice, language, output format and supported controls depend on your configured service.

The following is an illustrative request body for the integration preview:

{
  "model": "YOUR_ASSIGNED_SPEECH_MODEL",
  "voice": "YOUR_CHARACTER_VOICE",
  "input": "I was hoping you would call."
}

The assigned production contract determines the endpoint and whether the response is audio bytes or a job/result reference.

Recorded previews

Provide a short, representative sample before asking a visitor to start a live call. Playback should begin after a user action, show a play/pause state and stop when a live session begins.

Amora's homepage uses the same recording as the character demonstration, with an audio-format fallback for compatibility.

Transcription

A transcription workflow accepts an audio file or an approved file URL, submits it for processing and retrieves the result. For asynchronous jobs, retain the job identifier and check status rather than submitting the same recording repeatedly.

Language selection, supported file formats, payload limits and transcript retention are confirmed in your production service configuration.

Content delivery

Deliver voice messages through your asset storage and access-control layer. Paid messages should be available only to users with the required entitlement. Decide whether to retain source text and audio as part of the content workflow.