Prosody is the structure around language that tells us how to take it in.
More than pitch
Prosody is often reduced to a change in pitch. In conversation it is much broader: timing, stress, tempo, and silence all shape how a sentence arrives.
The same words can signal confidence, care, uncertainty, or urgency depending on the pattern around them.
Learning the right amount
Expression works best when it has restraint. An overperformed voice can be as distracting as a flat one, especially in a long conversation.
Our evaluation focuses on appropriateness: whether the voice makes the moment clearer, rather than merely more animated.
A useful future for synthetic speech
The opportunity is not to imitate every human inflection. It is to make spoken interfaces easier to follow, kinder in difficult moments, and more capable of holding attention without demanding it.
That standard is a better guide than realism alone.
Keep a good signal close.
New research and product notes from Sunesis, sent occasionally and thoughtfully.
Subscribe to the journal