High quality text-to-speech synthesis system with efficient duration models developed using coding schemes based on vowel production characteristics.