For several years, speech-to-text has been touted to have the potential to completely eliminate hand-based inputs. Unfortunately, speech-to-text has had limited success because of lack of reliability and robustness. Our system has taken a novel approach that bridges the gap between hand-based input methods and speech-to-text. We are using a framework called Predictive Speech-to-Text to enhance the accuracy. Using VoicePredict, a user speaks a word/phrase, types a letter or two, and the word/phrase appears on the mobile screen. This approach guarantees a near 100% task completion- accurate system because the system automatically chooses between the hand and voice inputs to generate "best of both worlds". The speech recognition system is based on CMU's Sphynx system. Both of these systems will be integrated into a robust speech-to-Text messaging system.