Acoustic Modeling for Speech Synthesis