How can I start and stop speech recognition while streaming

By default speech recognition is active as soon as you send a stream to an endpoint of your actived workflow. So naturally speech recognition stops when the stream stops or, the stream stops and the workflow is deactivated. However there are scenarios, that require the speech recognition to stop while a stream is active. Currently a workflow can perform speech recognition with a single language at a time. The input language is chosen by the user by streaming to the correct ingest endpoint. For details see.

In some scenarios, for example a panel discussion or a conference with a sequence of speeches in different languages, you will want to stop speech recognition for languages that are a mismatch with the ingest endpoint you are currently streaming to. If the ingest endpoint receives speech in another language than expected the speech recognition will produce incorrect results with low confidence values. If it is unavoidable to send mismatching languages to an endpoint while continuing to stream results (e.g. burned into video) it is better to stop the speech recognition.

To stop speech recognition, go to the "Streams" tab of your workflow:

1. Go to the currently active stream (highlighted in green colour).
2. Click on the three-to-symbol,
3. From the menu choose "pause speech recognition" to stop the speech recognition, 
4. Click on "start speech recognition" to start it up again.