Hu Jing Wen Yu, a subsidiary of Alibaba Group, uses its self-developed AI large model to restore high-definition video from several mainland Chinese and Hong Kong television dramas from the 1980s and 1990s, solving common issues such as frame skipping and image distortion in old films. According to a report by Radio Television Hong Kong (RTHK), Hu Jing Wen Yu’s algorithm expert Cheng Yijia stated that traditional manual restoration takes weeks or even months, while AI can compress the time to a few hours, reducing actual costs by more than 90%. Unlike manual restoration, which can only remove noise, the AI large model, trained on massive resources, can learn light and shadow patterns, facial structures, and texture details to restore and generate more image details. Cheng Yijia pointed out that the current technical challenge lies in ensuring that the emotions conveyed by characters’ micro-expressions are not damaged after restoration. Future goals include dialogue clarity, adding surround sound effects, and simulating film textures from different eras.
Economist Song Qinghui analyzed that in recent years, the homogenization of new film content has led to audience fatigue. The commercial value of classic works restored by AI is more certain than that of new films, with lower capital investment risks. At the same time, IP remodeling can spawn products such as games, cultural tourism, and short videos, giving rise to new industrial tracks like copyright operations and digital archiving. He expects this wave of value release from classics to last at least two to three years, and believes that AI restoration will push the film and television industry to shift from content production to upgrading existing content assets, bringing asset restructuring opportunities to enterprises with rich copyright reserves.