There is a fundamental problem with the whole idea:
The code runs on the SERVER.
How will a person sitting in front of a browser running on there computer that is located, let's say, 1000 miles away hear what is coming out of the server, if it even has speakers? Those would have to be some very loud speakers.
Speech in an application is neat-o, but in a web application it makes no sense.
Also, a simple semantic/style point: wrapping code in a try...catch that simply re-throws the error is not considered error handling. It adds absolutely no value, in fact it obfuscates the original source of the problem.