I would like to ask if there is any way to make the detections run in realtime instead of recorded files?

And if you would have any idea on how to chunk the audio for the detections ?