# CVG 1.24.3 (18-Oct-2022)

This release provides minor but for some customers very important improvements to our Call API.

## Call API Improvements

### Barge-In with DTMF and / or Speech

For announcements played via [`/call/say`](https://cognitivevoice.io/specs/#/call/say), [`/call/play`](https://cognitivevoice.io/specs/#/call/play) or [`/call/prompt`](https://cognitivevoice.io/specs/?urls.primaryName=Call%20API#/call/prompt), a bargeIn flag (`true` or `false`) could previously be used to specify whether these announcements should be interrupted by caller input or not. This is still possible, alternatively it is now possible to specify in a more fine-grained way that only DTMF inputs or only voice inputs of the caller should interrupt the announcements of the bot. 

### Number Prompt starts collecting DTMF Input immediately 

With the number prompt, i.e. a [`/call/prompt`](https://cognitivevoice.io/specs/#/call/prompt) request with type name `number`, the DTMF inputs are now collected immediately with the `/call/prompt` request, regardless of how long the announcement runs in the prompt. 

The prompt timeout starts as before after the announcement has been played completely. 

### Audio Playback Error Handling

Playing an audio file via [`/call/play`](https://cognitivevoice.io/specs/#/call/play) may fail due to an incorrect link to the audio file, network problems, or an incorrect format of the audio file. To avoid silence in this case, the [`/call/play`](https://cognitivevoice.io/specs/#/call/play) endpoint now accepts an optional `fallbackText` which is then announced (no SSML formatting possible here).

In addition, a Health Warning is generated if an audio file is not played successfully.