Interview with the non-professional team behind a 12bn-play AI drama
专访《持械入宋》主创：草根团队如何做出120亿品质爆款
B · Useful, limited Entertainment outlet Wenyu Jiazhiguan interviews the three lead creators of the AI short drama Chixie Rusong; a first-hand account of a working AI video workflow, with all production and revenue details self-reported.
Summary based on the full text·Original: about 12 minutes to read
In short
A 8 November 2026 interview by Wenyu Jiazhiguan (文娱价值官), carried by 36Kr, with the producer, planner and director of Chixie Rusong (持械入宋), a ten-season AI-generated historical short drama that topped Hongguo's AI chart. The core team of six has no film training and works from home across China; members include a stay-at-home mother, a ride-hailing driver and a delivery worker. They say they never buy traffic. Script adaptation, not video generation, takes most of their time, and repeated generation ("drawing cards") is the main production cost.
Key numbers
- 12 billionreported plays of the series on the Hongguo platform as of 6 November 2026
- 0.48% of 221,900share of AI dramas that passed 100 million plays, cited by the outlet to show the series is an outlier
- about RMB 1 millioncost figure circulating in media, which the team confirms covers compute only; it rejects the RMB 10 million return estimate
- 6 core / 30 totalcore creators on this drama, and members across several groups working remotely, none with film training
- 9 to 10+reference images per generation, Seedance 2.0 versus 2.5, which helped crowd scenes
As reported by the source.
Chinese·Web page·opens 36kr.com
What it reports
- Launched 21 August 2026; season 4 topped the Hongguo AI new-drama chart; season 10 heat reached 82.06 million; about 4 million people booked unaired episodes.
- Team: 6 core members on this drama, 30 in total across groups making different dramas; each works as a one-person company, so compute is 100% of cost; revenue is split by contribution.
- The circulating RMB 1 million cost covers compute only; the RMB 10 million return is a media estimate the team calls inaccurate.
- Workflow per person: read the novel, break down the script, write prompts, generate video, edit. Consistency comes from a shared asset library of character, scene and prop images and a shared understanding of the adaptation.
- Tools: Seedance 2.0 for the first season, then Seedance 2.5 (from 9 to over 10 reference images); Doubao for script work.
- Scenes with twenty or thirty characters need many regenerations because of consistency errors; long takes (currently 30 seconds) and facial performance remain weak points.
- An earlier drama hit 160 million Douyin plays but was removed from Hongguo after 6 days for breaching review rules.
- Preparation: IP chosen and team selected in about 7 days; first 10 episodes made by two people in a week.
How they got it
A single interview with three creators, plus platform chart figures cited by the outlet.
Limits
- All production, cost and team figures are self-reported; revenue is undisclosed.
- Play counts are platform-reported and not audited.
- The creators credit luck as a large factor; one hit says little about repeatability.
Why it matters for companies
The case shows how far production cost has fallen for video content: a distributed amateur team with subscriptions and a shared asset library produced a ten-season hit, and the bottleneck moved to story adaptation and judgement. For media and marketing teams the operating lessons are a common asset library for consistency, budget for many regenerations in complex scenes, and early knowledge of platform review rules, which removed one of the team's earlier hits.
Key passages in translation
Chinese original
1000万回报是媒体估算的，不准确。具体收入我没有对外讲过，而且我也没收到那么多钱。
The RMB 10 million return is a media estimate and not accurate. I have never disclosed our income, and I did not receive that much money.
Chinese original
技术是服务于内容的。
Technology serves the content.
Chinese original
目前时间成本最高的部分是花在剧本打磨上了。
At present the largest share of our time goes into polishing the script.
Summary based on the full text. Read the original for the full argument and data.Read the original on 36kr.com