基于语音驱动的风格化数字人关键技术研究与应用示范  

Research and Demonstration of Key Technologies for Voice-driven Stylized Digital Humans

作　　者：郝洺 张翀 冯海亮 施玉海 Hao Ming;Zhang Chong;Feng Hailiang;Shi Yuhai(Academy of Broadcasting Science,NRTA,Beijing 100866,China)

机构地区：[1]国家广播电视总局广播电视科学研究院,北京100866

出　　处：《广播与电视技术》2024年第10期20-23,共4页Radio & TV Broadcast Engineering

基　　金：广科院2024基本科研项目《生成式人工智能应用于视频生成的安全评估技术研究》(24011401),2024年广科院实验室运行维护经费(240305)资助。

摘　　要：近年来,数字人技术的应用正逐渐成为推动媒体融合发展的重要力量。本文提出了一种基于语音驱动的风格化数字人技术框架,旨在为广电领域提供一种新的技术解决方案。该框架通过语音生成、唇形同步、面部风格化等AI技术生成数字人,在保留较高的人物形象真实性的同时降低了制作成本。通过在北京东城会馆的应用示范,验证了该技术在大屏应用中展现出广泛的适应性和可行性,为广电行业的内容创新和形式多样化提供了强有力的技术支持。In recent years,the application of digital human technology has gradually become an important force in promoting media integration development.This paper proposes an innovative framework for voice-driven stylized digital humans,aiming to provide a new technical solution for the broadcasting and television industry.The framework utilizes AI technologies such as voice generation,lip synchronization,and facial stylization to create digital humans,while maintaining a high level of character image authenticity and reducing production costs.Through the application demonstration at the Dongcheng Assembly Hall in Beijing,the wide adaptability and feasibility of this technology in large-screen applications have been verified,providing strong technical support for content innovation and diversification in the broadcasting industry.

关 键 词：语音驱动 风格化 数字人 唇形同步 会馆 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]