A chance encounter
Chen first heard about the project by chance in 2022, while filming at the Yungang Grottoes in North China's Shanxi province for the second season of Hello AI, her documentary series on artificial intelligence and cultural heritage. It is a follow-up to the 2019 film that marked one of the country's first documentaries dedicated entirely to AI.
On set, she ran into an old collaborator, Huang Xianfeng, a remote sensing expert from Wuhan University. They had previously worked together on digital restoration projects at Dunhuang in Gansu province and the Jiankou section of the Great Wall in Beijing.
She asked him whether his work had ever extended to artifacts that had been looted and taken overseas. That was when Huang mentioned the Longmen reliefs.
Chen started digging. What she found was a detective story spanning continents and decades, with smuggling rings and a paper trail buried in archives from New York to Cambridge, Massachusetts. As shown in the film, a key piece emerged from the Harvard University archives — a letter from Sickman to Alan Priest of the Met, accompanied by a sketched male head. The drawing proved that American curators had been searching for the same specific missing piece, providing crucial historical corroboration for fragments later found in Longmen's storeroom.
For years, Chen had worked with two seemingly separate tool kits. Her early career was focused on cultural documentaries, such as The Forbidden City, Kunqu of Six Centenary and Taipei Palace Museum. Later, she directed A Century with Cars, a 12-part series examining Western industrial civilization through what she describes as "the eyes of a Chinese documentary filmmaker".
That was followed by Hello AI, where she learned to visualize invisible algorithms while tackling the challenge of projecting data streams onto server racks and using crosscutting techniques to keep otherwise static code visually alive on-screen.
"One foot in the humanities, one foot in technology. I'd been waiting for a subject that could bring them together," she recalls.
She realized that the emperor and empress reliefs offered exactly that opportunity.
In 2024, the Longmen Grottoes Research Institute launched a full-scale digital restoration project, led by senior researcher Gao Junping. The first challenge was separating authentic fragments from fakes.
Tian Hengci, a researcher at the Chinese Academy of Sciences who usually analyzes lunar samples, applied nondestructive testing technology to the Longmen fragments. His device can detect trace elements at concentrations as low as one part per million — a geological fingerprint precise enough to distinguish different stone sources.
As documented in the film, of the five candidate pieces, only one cataloged as H05 matched the chemical composition of the original cave wall. The rest were counterfeit.
Yet, finding the correct fragment solved only part of the puzzle. The cave wall stands two meters high and four meters wide. H05 is roughly the size of a palm. Somewhere on that vast, weathered surface, the tiny piece needed to find its home.
The documentary reveals how researchers combined high-precision 3D scanning, AI-assisted surface matching, and what some scholars call "digital reunion": scanning overseas collections and virtually reassembling scattered fragments without physically moving them.
"Academically, 'digital reunion' isn't quite accurate, but it helps people understand what we're trying to do," Chen emphasizes.
Privately, she prefers another phrase: "civilization backup".
"For people 200 years from now, these digital models will be a heritage of their own — a record of how we, in our time, tried to save what we could," she explains.
Xiao sees this fusion of technology and humanity as essential to the documentary's success.
"Hard technology plus soft humanities is the soul of this kind of film, and a growing trend," she says.
"Without technology, Wen Yucheng's regret might have remained just regret. But without humanities, the film becomes a boring engineering log. You need both — technology as the engine, humanities as the heart. Neither works alone," she elaborates.
For Xiao, the idea of "reunion" defines the film.
"It's not only about reassembling broken stones," she says. "It's also about reuniting hearts and reconnecting civilizations." She adds that researchers, scientists, curators and scholars from China and abroad are all participating in an effort to restore a fragmented cultural legacy.
"We're using hands that reached for the moon to touch the missing pieces left behind a thousand years ago," she says.
The film's title How I Miss "Her" comes from a line by early 20th-century poet Liu Bannong. The Chinese character ta (her) was once reserved in Chinese documentary captions for national treasures, a grammatical gesture that imbued stone and bronze with something like soul.
"This film is an innovative work that uses contemporary thinking to question history and technological artistry to empower imagery," says Tan Xue, the film's producer.
In recent years, artificial intelligence has opened new narrative possibilities for cultural heritage, giving rise to many moving stories of "reunion", Tan notes.
The film allows those who have missed "her" to see how cultural heritage researchers, represented by the team at the Longmen Grottoes Research Institute, are using contemporary wisdom to heal the wounds of history, she says.
The title also raises the question: how do you miss something that may never come back?
Chen's answer lies less in sentiment than in action.
"Science and art part ways at the foot of the mountain. But they meet again at the summit," she says.
At the foot of that mountain, a sixth-century artisan carved an emperor's face into stone. At the summit, a laser originally developed for lunar exploration scans a fragment no larger than a hand while algorithms search for patterns invisible to the human eye.
"That's where we are now," Chen says.
"Not because we have all the answers. But because, for the first time, we have a way to ask the right questions."
