American English male speaker ("Kurt") for Festival.
This voice provides an American English male voice using a residual excited
LPC diphone synthesis method. It uses the CMU Lexicon for pronunciations.
Prosodic phrasing is provided by a statistically trained model using part of
speech and local distribution of breaks. Intonation is provided by a CART
tree predicting ToBI accents and an F0 contour generated from a model
trained from natural speech. The duration model is also trained from data
using a CART tree. |