I think that automatic phonetic transcription is a feat beyond programs for the time being. Phonemic transcription may be a more realistic goal but even then, the program should be able to tell the difference between pairs such as "row" /r@U/ ('row a boat') and /raU/ ('quarrel').