How AI and Big Data Drive Casting Decisions in the TV Series “The Longest Day in Chang'an”

Under AI and big‑data calculations, Lei Jiayin’s attractiveness score “outshines” 90% of male celebrities, ranking in the TOP 10. What’s the secret behind this? Youku’s algorithm experts decode it.

The drama that has taken the nation by storm is “The Longest Day in Chang'an”. On Douban, nearly 140,000 users gave it an 8.6 rating. Many viewers praised the strong fit between Lei Jiayin and Yi Yangqianxi with the characters Zhang Xiaojing and Li Bi, which added great value to the series.

Youku told Southern Metropolis (Nan Du) reporters that both Lei Jiayin and Yi Yangqianxi were selected through a big‑data casting system. Lei Jiayin’s attractiveness was scored by AI and big data, placing him in the TOP 10 among male stars. Southern Metropolis interviewed the people behind the data‑driven casting – Alibaba Youku Cognitive Lab algorithm expert Shen Yi and Alibaba Entertainment Group Youku Beidouxing product expert Shi Xiaofei – to further uncover the story.

“Big data found Lei Jiayin through tags like ‘big‑head’ and ‘hard‑man’.”

Nan Du: It’s said Lei Jiayin was chosen by big data for the role of Zhang Xiaojing. What dimensions were evaluated?

Shen Yi: We have a casting model built on two main dimensions – the actor himself and the historical performance of related works. We extract audience‑generated tags from the whole‑web sentiment data and compute the relevance between actor tags and role tags. We also consider current popularity, commercial value, and looks. Additional factors include past roles, screen time, box‑office, streaming numbers, and ratings. All these are aggregated into a candidate list from which we select.

Nan Du: In which product is this model embedded?

Shi Xiaofei: Casting is one of the application scenarios of the Beidouxing product, an AI platform for “pan‑content” (film, TV, variety, comics, artists, creators, etc.). Beidouxing leverages AI + big data to deconstruct content across its entire lifecycle – from investment and production to operation and marketing – providing decision‑making assistance.

Nan Du: Which tags of the character Zhang Xiaojing matched Lei Jiayin?

Shen Yi: Zhang Xiaojing is tagged as “hard‑man”, “lone hero”, “smart”, and “righteous”. The novel also calls him “Big‑Head” because of his large head, a tag that appeared frequently in audience comments about Lei Jiayin. His previous anti‑Japanese war drama roles also contributed tags like “hard‑man”, “righteous”, “handsome‑rebellious”, and “rigorous”.

Lei Jiayin’s matched tag also includes “Big‑Head”.

Nan Du: How did you evaluate that Lei Jiayin’s looks beat 90% of male stars?

Shen Yi: We built an attractiveness model using machine learning. Thousands of real people scored photos of actors and celebrities; we fed those scores to the model, which learned the human scoring logic. The model can then evaluate new photos and assign a score. Using this model, we scored all online actors, and Lei Jiayin ranked in the TOP 10, surpassing 90% of male artists.

“Changing to two episodes per week was a big‑data recommendation.”

Nan Du: Were there other characters in “Chang’an” selected by big data?

Shi Xiaofei: Yi Yangqianxi was also selected through big data. The Beidouxing system provides AI + big‑data decision support, but the final casting decision still rests with the director, who also considers schedule and cost‑effectiveness. The combination of AI, data, and industry experience creates a strong complementary advantage.

Yi Yangqianxi as Li Bi

Nan Du: Was the shift to two episodes per week also driven by big data?

Shi Xiaofei: The Beidouxing system suggested increasing the release frequency. User comments and viewership data indicated the need for faster updates, so the schedule changed from once a week to twice a week. This decision was approved in a project meeting as an AI + big‑data recommendation.

Nan Du: What pain points does big data help solve in casting?

Shen Yi: AI provides assistance and makes results more objective, reducing subjective bias such as personal preferences. Big‑data results are explainable across dimensions and broaden the talent pool beyond what a single director might consider, preventing omissions.

Nan Du: In which stage of the drama is AI + big data most directly applied?

Shi Xiaofei: The broadcasting stage provides the most direct data, as we can capture viewership, interaction, and performance metrics in real time. Our platform monitors everything from IP, script, to final product, quantifying content indicators as much as possible.

Nan Du: What challenges do you face?

Shi Xiaofei: Quantifying content is harder than quantifying a product. For a shoe, dimensions are clear; for a drama like “The Longest Day in Chang’an”, it’s far more complex. Attractiveness scoring required massive training data, and understanding the audience for content demands deep psychological profiling beyond simple demographics.

We will continue to upgrade the product – extending Beidouxing from script analysis to production, providing detailed recommendations for casting, character portrayal, filming, editing, and even promotion and operation, aiming for tighter integration between content and platform.

Report by Southern Metropolis journalist Xu Bingqian