본문 바로가기
  • Home

Analysis of the usability of facial expressions generated by generative artificial intelligence

  • Journal of Software Assessment and Valuation
  • Abbr : JSAV
  • 2025, 21(2), pp.103~112
  • Publisher : Korea Software Assessment and Valuation Society
  • Research Area : Engineering > Computer Science
  • Received : June 3, 2025
  • Accepted : June 20, 2025
  • Published : June 30, 2025

Eun-Mi Ji 1 Yoonsik Shim 1

1배재대학교

Accredited

ABSTRACT

This paper aims to analyze whether generative AI based on currently developed large language models (LLMs) can generate avatars that display appropriate facial expressions in context when interacting with humans in avatar form. Furthermore, it evaluates the usability of the generated facial expressions by comparing them with existing facial expression recognition datasets across various categories such as gender, expression type, age, and identity. Lastly, this comparative evaluation seeks to identify and analyze the differences and characteristics among the results of various existing AI models, thereby assessing the potential applicability of current AI systems in generating realistic and contextually appropriate facial expressions.

Citation status

* References for papers published after 2023 are currently being built.