Baidu Unveils Digital Human Video Podcast Solution with Advanced Micro-Expression Tech at WAIC

At the WAIC World Artificial Intelligence Conference on July 17th, Baidu's "Yi Jing" (一镜) officially launched its digital human video podcast solution. This innovative offering significantly advances digital human micro-expression technology, enabling natural interjections and professional audiovisual language capabilities. The solution addresses persistent industry challenges such as the high cost of real-person interviews, the stiff demeanor of traditional digital humans, fragmented multi-person dialogues, and distortion in multimodal splicing. It is poised to revolutionize the industrial-scale production of deep-interview content featuring digital humans.

Technologically, the solution is built on three core pillars. The Wenxin large model meticulously plans micro-expressions, moving beyond simple emotion tags. A self-developed specialized digital human model achieves expression control within one second, precisely synchronizing eye movements, body language, and speech. Extensive training on professional audiovisual samples ensures deep unification of multiple modalities, substantially enhancing digital human generation capabilities. The product integrates three AI agents—scriptwriter, director, and editor—along with over 80 skills to automatically generate analytical interview scripts. It supports single-person lectures and two-person debates across finance, law, technology, and humanities sectors. Real-world tests indicate a 74% reduction in production costs, a 785% increase in content expressiveness, a 29% rise in average playback duration, and a comprehensive superiority in character naturalness compared to current market-leading models.