Speech synthesis output and voice recognition would get my vote.