Hengdian, do you believe in AI?

Hengdian is now at the crossroads of artificial intelligence technology. On the one hand, short drama producers still adhere to the traditional production methods, quickly and efficiently producing a large amount of content; on the other hand, the introduction of AI technology is changing the investment and distribution of short dramas, bringing new variables to this traditional industry. This article will explore in depth the acceptance of AI in the short drama industry and how it affects the production, distribution and profit model of short dramas.

Hengdian, a place occupied by vertical screen short dramas, is being divided into two halves by AI.

Half of them are skit makers who have little interest in AI.

Among the various shooting recruitment information in Hengdian, projects with a cycle of 3 to 7 days occupy most of the pages. Countless actors, directors, photographers, and makeup artists form a stable production line, combining programmed plots such as slapping, kneeling, and scattering money, and continuously producing earthy short dramas that dominate the screen and Kuaishou.

"99% of the short dramas are real shots," said Chen Yiyi, head of Kuaishou's entertainment business, at an event in July.

The other half are the “short drama promoters” who use AI to lower the price of short drama streaming.

"30% drama, 70% investment." Generally speaking, the investment cost of a short drama accounts for as much as 80% to 90% of the total cost, and it is also the key to determining the life or death of a short drama.

In the past, an editor earned a few hundred yuan a day and could cut out about 30 to 100 pieces of information flow materials; the pitchman was responsible for taking these materials, buying traffic for testing over and over again. A single short drama needed to be tested for 1 to 2 days to find the best way to deliver it.

And now, some people have begun to use AI to complete editing and streaming with just one click.

Not only are traditional streaming companies trying to go AI-based, but new streaming companies based on AI have also begun to cooperate with short drama parties and enter into commercial verification. For example, the AI ​​editing and streaming company Starry Sky Intelligence is cooperating with short drama parties to distribute short dramas in the form of CPS sharing by experts.

The increase in delivery speed means that AI materials can take the lead in occupying traffic and be the first to complete the process from user viewing to user payment.

Why does AI have a different effect on the production and streaming of short dramas? Will AI's changes to streaming make the highly profitable short dramas even more profitable?

1. Side A: The producers produce one drama every seven days and have no time to eat the big cake of AI

While online literature and film and television have developed industrialized production capabilities over decades, short drama production has reached maturity within five years. In the eyes of many producers, the seamless short drama production chain has left no room for AI.

For example, from the starting point of this industry chain, which is also the script link, on the surface, the underlying logic of the short play script is a formulaic routine, which should also be the comfort zone of the big model. At present, there are indeed a large number of AI writing products on the market, such as Frog Writing, WriteWise (an AI writing tool for online articles and novels under Himalaya), etc.

But in the entire short drama business model, the content of the short drama is precisely the least important. The production logic of short dramas has always been to buy enough scripts, shoot enough dramas, and bet on the probability of a hit.

In terms of the quantity of content, short drama platforms like Jiuzhou have their own writing and editing teams, and can stockpile short dramas at a very low cost. The total production cost of a short drama is between 300,000 and 1 million yuan. The general script quotation of large platforms is between 10,000 and 30,000 yuan, and the script quotation of small platforms may only be a few thousand yuan. AI has no price advantage but increases the learning cost.

Scripts are not scarce, what is scarce are good scripts, but good scripts are difficult to define.

The reason why short dramas form a mechanism of betting on probability is precisely because under the algorithm mechanism of Douyin and Kuaishou, there is no universal formula for the methodology of creating popular products.

Compared to delving into the unpredictable user psychology and algorithm black boxes, a more practical approach for short drama producers is to find a screenwriter who often produces hits based on the principle of hard work and bet on a higher probability, or after a TV series, short drama or movie of a certain theme becomes popular, quickly copy the script, shoot and release it, directly imitating the successful path that has just been verified.

For example, during the Spring Festival this year, when Jia Ling's film "Hot and Hot" was a big hit, the short drama company quickly launched "Hot and Hot: Gorgeous Transformation", and "Marriage Riding the Wind and Waves" relied on the same name to "absorb" the traffic of the big market, "The God of War Mom Stops Pretending" was compared to "The God of War" and so on...

Next, in order to bet on the probability, the production capacity of short dramas must be expanded. In the process of shooting short dramas, the most important thing is speed. In this regard, the current AI technology cannot keep up with humans at all.

An obvious comparison is that the 15-minute Shanhai Mirror by Xianren Yikun took half a year to produce, while the shooting cycle of a Hengdian short drama has been compressed from 2 weeks to 7 days, or even 3 days or 1 day. The crew may not remember each other, and the crew will be disbanded after one day of shooting. They have no need and are unwilling to spend time exploring how to use AI to replace traditional shooting.

Even if the AI ​​face-changing short drama that was popular for a while went overseas and crushed the original overseas short dramas in terms of cost and timeliness, it soon fell into obscurity because users did not buy it. This is another characteristic of the short drama industry - it is all about money.

Compared with the nearly $300,000 production cost of hiring overseas actors for filming, the production cost of the entire process of using AI face-changing only costs tens of thousands of yuan, a cost reduction of 5 times.

However, the final effect failed to withstand the test of the market. Due to the large number of micro-expression flaws in AI face-changing technology, the user retention rate and payment rate of the face-changing short drama were very low. The "2024 Short Drama Overseas Marketing White Paper" released by TikTok for Business mentioned that in terms of content repertoire, local original dramas and translated dramas are parallel, but the hit dramas are mainly concentrated in local original dramas.

This also means that although short plays have different acting requirements from TV dramas and movies and can be rough and absurd, they must be direct enough and ensure a fast pace, emotionality, and high information density. However, what AI videos are currently best at are empty shots and slow motion.

Character performance is also a major pain point in current AI production. While exploring AIGC works, Xianren Yikun discovered that "through the method of text-to-image and picture-to-video, the consistency of action and scene in AI videos can be controlled, but the range of action and character performance still need to be improved." Users often watch revenge skits to put themselves in the role, and AI-flavored fake people will only distract people from the show.

In order to avoid the problem of character performance, the AI ​​short dramas that have been launched recently, such as "Mountain and Sea Mirror", "Sanxingdui: Future Revelation", "Fantasy Store", etc., are all fantasy and science fiction themes with non-real pictures, and they emphasize the display of imagination and shocking visual effects, which are suitable for AI generation to display.

From the perspective of each link in the entire production chain, under Hengdian's high-speed operation mechanism, the producers of short dramas that pursue short, flat and fast productions have no time to care about AI at all.

2. Side B: AI editing and streaming, more than 10 times faster than humans

People who make skits have little interest in AI, but people who distribute the skits have already begun to enter the AI ​​exploration stage.

After all, for short dramas, traffic determines life or death. In paid and CPS short dramas, the amount of traffic consumed basically reflects the revenue of the short drama. Therefore, in addition to announcing the recharge amount and the amount of profit sharing, the more commonly used statistical dimension for short dramas is actually consumption, that is, how much money was spent on advertising traffic.

This is because in the process of traffic investment, as long as the preset ROI is achieved, the pitcher will continue to buy traffic to strive for higher payments. Roughly calculated, taking ROI 1.2 as an example, this means that if a short drama wants to earn 120 yuan in user payments, it will have to invest 100 yuan in traffic purchase costs. If the total consumption of a drama is less than 1 million, it means that the income is less than 1.2 million, the profit from traffic investment is less than 200,000, and even the production cost is not covered.

However, short drama investment and streaming, as the most costly link in the short drama industry, is also the biggest pain point in the current industry.

Judging from the macro data, as the number of short dramas increases, the growth rate of short drama traffic has slowed down, the price of buying traffic has become higher and higher, and it has become increasingly difficult to obtain paid traffic.

Data shows that the scale of investment in the market has begun to shrink. On July 10, DataEye Research Institute released a report on the investment data of short dramas in the first half of the year, predicting that the investment scale of short dramas in 2024 will be between 25 billion and 30 billion yuan. At the beginning of the year, this data was estimated to be 42 billion yuan, even lower than the investment scale of 31.2 billion yuan in 2023. The report mentioned that the main reasons are the rapid acceleration of industry competition, the impact of piracy, and stricter supervision.

Short dramas are a traffic business in themselves. However, traffic investment is highly dependent on the algorithm black box of the distribution platform. The traffic investor can only continuously test the advertising materials and put money into the pocket of the platform.

In this process, the editor and the pitcher are the main people who work together. The editor is responsible for continuously producing short drama clips based on the excitement and timing, while the pitcher is responsible for configuring the corresponding delivery plan based on the type of these materials, such as which advertising account to choose, which group to target, and when to deliver. He also adjusts the delivery materials and plans through real-time delivery feedback.

For example, the pitcher first invests 100 yuan and analyzes that the main audience is middle-aged women around 50 years old. Then, based on his own experience, he judges that this group of people likes to watch stories about ethical disputes. He then communicates with the editor, adjusts the focus of the material, and cuts in the scenes that best highlight family relationships such as mother-in-law and daughter-in-law into the material. He then places another round of ads and optimizes the best delivery model based on data feedback.

The entire chain reaction from pitcher to editor to market verification only requires rounds of testing, and there are no high barriers.

Especially because investment requires advance payment, in order to expand the coverage, the short drama producers often hope that as many people as possible will participate in the distribution process. Therefore, there are many large and small investment companies and even individual "retail investors" in the market.

As a result, the launch will become a speed competition, competing to see who can test it earlier and grab paying users faster.

In the race to speed up, the pressure on the production of materials is increasing. At the end of last year, a person in charge of short drama production mentioned that "the profit of the production business itself is very low. Now a piece of material may be wasted in three days. Our editors can produce limited materials every day. If we want to increase the production of materials, we have to hire more people. But relying on manpower will not have a scale effect."

AI can solve all these problems.

According to Sun Zhipeng, an entrepreneur in AI smart editing and delivery, they have now run through the AI ​​editing and delivery process based on a combination of multiple large models, and have begun to cooperate with two short drama parties on distribution.

The logic of AI editing and delivery is to first use AI's multimodal recognition capabilities to analyze the content of the short drama, and batch edit short drama materials of about 1 minute. During the distribution process, by connecting to the API of platforms such as Douyin, AI can re-understand market demand based on real-time delivery effects, adjust the new round of material editing methods, and then continue to cycle until the best material and delivery cooperation model are found.

"Unlike previous algorithms that can only analyze data, today's multimodal models can compare image data, subtitle text data, sound data, and delivery data on a parameter dimension, and can even analyze more dimensions than humans," said Sun Zhipeng.

During this process, the big model can understand and edit skits, analyze data, and autonomously learn how to adjust the material effects, which is equivalent to connecting the pitcher and the editor together.

"Before, one pitcher was assigned to four or five editors, who would hold frequent meetings and do tests in the first two days, taking four to five hours to test one set of content. But with today's logic, AI can automatically test dozens of sets of content in an hour, and then iterate continuously," he said.

Although AI’s generation capabilities on a single task may not be better than humans, it has great advantages when handling labor-intensive, iterative assembly line jobs.

AI has not yet revolutionized the short drama industry, but it has begun to slowly change the rules of the game.