Improvements in Speech Synthesis