LLM-Based Selection of Incongruent Verbal and Nonverbal Behavior for Virtual Humans
IndustryAI 55
Source: arXiv cs.ROPublish time unverified
arXiv:2608.22731v1 Announce Type: cross Abstract: Nonverbal behavior generation systems for virtual agents often take an utterance as input and generate nonverbal behaviors that emphasize or illustrate the content of the verbal channel. However, human nonverbal behavior is shaped by more than the content of the speech. It is also influenced by speaker roles, interpersonal relationships, social context, and the cognitive and emotional states of the interactants.