디지털 휴먼 제작기술 기반 영상미디어 활용사례 및 특성연구 -실시간 / 비실시간 영상을 중심으로-

Abstract

(Background and Purpose) The size of the market utilizing digital humans is growing rapidly. Compared to the past, the production costs of digital humans today have been significantly reduced due to technological development, and the scope of this technology’s utilization has been extended with the emergence of various social media that enables active interaction of digital humans with the public. However, compared to technological development and market expansion, studies on utilization cases of media based on digital human production techniques remain insufficient. Therefore, this study aims to provide practical implications for digital human content production in the future by analyzing the characteristics of utilization cases of video media according to the types of digital human production techniques. (Method) This study first confirmed the theoretical definition of digital humans, further classifying them into three types based on the classification criteria of production methods in previous studies: real image composition, Full 3D, and artificial intelligence (AI). The scope of the study was limited to video media, and it was divided into real-time and non-real-time according to the production method. (Results) This study analyzed utilization cases of video media based on digital human production techniques. The real image composition type had high efficiency due to low production time and production costs; however, it was not suitable for real-time video because the composition looked unnatural depending on the angle or movement of the face if the video continued in real time for a long period. As most full 3D types are produced based on game engines, they produce high-quality results graphically; however, motion capture continues to be unstable to be transmitted in real time; therefore, it is suitable for non-real-time video. AI type is mostly used in the form of restoring real people based on deep learning, which is most suitable for real-time broadcasting as mouth movement looks natural in this case, and it is possible to respond immediately based on the contents learned in advance. (Conclusions) Through the analysis results, it was possible to identify differences and limitations in efficiency, detail, and real-time response by type of digital human production techniques and to determine whether the type was suitable for real-time or non-real-time videos. In the future, it is hoped that this study will provide producers with practical implications for using digital human production techniques in media in a way most suited to each type.