Is it possible for this model to be used in streaming use cases? For example, if I have a live audio feed and want to find when a certain sound happens?