Flow authors design flows in Architect. Architect then transforms the flows into VoiceXML (VXML) documents that run on the Edge. The VXML interpreter, built into the Edge, analyzes the documents, and the VXML rules determine how the system handles audio.

When a flow processes an action that includes an audio file or a text-to-speech (TTS) operation, the system does not play them immediately. Instead, these items enter the queue for later playback, and the flow execution continues normally. However, in VXML, flow execution continues immediately after the document loads, and continues until any of the following behaviors occur:

Behavior Result
The flow issues a transfer or disconnect action.
  • Terminates the flow document execution.
  • Causes queued audio to play before it is processed.
The far-end service disconnects.
  • Terminates the flow document execution.
The flow issues a request for user input, for example, a field in VXML.
  • Pauses the flow document execution until the user enters a response, or the system times out. 
  • Causes queued audio to play before it is processed.

Note: A request for user input is especially important because it is the only non-terminating VXML operation that causes queued audio to flush, or play.

For more information on how audio buffering and flushing work, see Architect audio buffer and flush FAQs.