>Why can't crypton do English voices this well when at&t can do it while barely even trying.
On second thought, this is a good question. Clear sounding text-to-speech is easier since there's no manual tuning involved unlike singing. That applies to Japanese voicebanks too (e.g. comparing VOICEROID voicebanks with their VOCALOID counterparts).
However, I don't know why the English Vocaloids themselves don't sound as good the Japanese ones.