from the NetBSD maintainer:

This voice provides an American English female voice using a residual
excited LPC diphone synthesis module, both created at OGI.  It uses
a lexicon compiled from MOBY and CMU lexicons, and other trained
modules used by CSTR voices.
 
This voice can be activated via (voice_tll_diphone) .

WWW:  http://cslu.cse.ogi.edu/tts