The speech service is a speech synthesis service. It will take strings of characters and attempt to change them into audible speech. There are two parts to the Speech service, a front end and a back end. The front end simply accepts a string, but the back end can be switched from FreeTTS or Google.
FreeTTS is an open source speech synthesis system.
Google is an online service, so initially this service needs to be run with an internet connection. The audio generated is saved locally, so future utterences of the same phrase will not need the internet connection.