I might have voted for "using the ancient HTML voice", but that would actually be pretty cool. You can use it any time to read text on the page, or you can feed text to it. Sometimes the user has different voices already loaded, but you can get some different sounding voices by varying the speed and pitch.

Did you know that the browser has this already installed? It's independent of the operating system and it always works. It's an old part of the accessibility system, and barely anyone every uses it.

... it might be able to simulate "crowd noise", not sure yet ...