This article discusses a research paper on using an LLM to select nonverbal behaviors that differ from a virtual human’s spoken words.