Renowned Chinese director and screenwriter Lu Chuan unveiled a preview of his AIGC short film Lishi Xianchang (History on Site) at the 2026 Apsara Conference in Hangzhou, Zhejiang province on Sept 21, marking his first in-depth use of AI to create film-quality visuals.

The film recreates the 1626 Wanggongchang explosion in Beijing during the Ming Dynasty (1368-1644). Lu said the project grew out of his research into Ming history while preparing for his film Tiangong Kaiwu. He hoped AI could combine accounts from different eyewitnesses, bring the historical event to life, and explore mysteries surrounding the explosion.

Lu began experimenting with AIGC in 2023. Using traditional production methods, recreating Ming-era Beijing, palace scenes, and a large-scale explosion would have taken months and incurred enormous costs. With AI, his team completed the work in just five days.

The team translated descriptions from historical texts, such as smoke clouds resembling black mushrooms, into modern visual concepts. After about four rounds of prompt refinement and references to real explosion footage, Alibaba's video generation model reproduced smoke, debris, and other effects with an accuracy rate of more than 95 percent, according to Lu's expectations.

Lu noted that AI-generated historical scenes can resemble familiar costume dramas because general-purpose models are trained heavily on existing film and television content. An Alibaba technical executive said the company will work with directors, industry experts, and historians to improve historical accuracy and further develop Agentic Video capabilities.