IVR – Text-to-Speech or the Real Deal?

At some point during your IVR application planning and development, you will be faced with the decision to either implement Text-to-Speech (TTS) or use the recording of a live person – or possibly both – to deliver the desired information to the end user.  Although TTS has come a long way from its electronic-sounding origins, most end users would agree that it still does not meet expected sound quality expectations, in that it doesn’t sound like a live person.  There are more than a dozen TTS vendors in the marketplace, some of which have mastered the synthesis of human voice better than others.  I’m of the opinion that a recording by a live person is the best method to deliver an optimal customer experience when replacing a live interaction with IVR.  My rule of thumb is to use a live person to record the body of the IVR application, while sticking to TTS for database driven content.  TTS can work very well to read back dates, currency amounts, or short bits of information which can change dynamically, or even for reading names from a database.  Ultimately, you want your IVR application to deliver the most natural-sounding experience.  Only through user testing will you find the right balance between the use of files recorded by a live person and TTS.