Looks nice so far, although I would have to think a while to find a case where I needed predominantly name and address dummy data :-)

A few more sets of possible data geared towards specific regions might be nice. For example right now the output is exclusively ASCII which is unlikely to hit any legacy charset mismatches.