Comment by adrianbg

My guess would be that they have separate speech recognition / synthesis models. Sorry, I should try and get this thing out to other countries too.