This morning I made a road trip back to my town, guided by a GPS system the German-language voice output of which also gave town steet and names (‘turn right on Heidenheimer Strasse’, ‘continue 20 kilometers to Tübingen’).
Of course, unlike set phrases such as ‘turn right after … meters’, a text-to speech system cannot have all possible street, town names etc. in storage as recordings, but needs to synthesize the pronounciation of these proper names based on the text->pronounciation rules of the language. With German-language proper names this works pretty well as pronounciation is regular (you just need to be sure of how to pare a name into syllables, e.g. Neckarsulm is Nek|ar|sulm not Nek|ars|ulm).
Now, the pronounciation of English-language toponyms and personal names (the latter figuring in street names) is much more irregular. Which makes me curious: How well does English-language text-to-speech typically cope with street names etc.?