Author: Changan I Biteye Content Team

Right after the 2026 Spring Festival, actors from Hengdian began posting videos on TikTok complaining about lack of work.

The group for the production team has gone quiet; in previous years, notices would arrive before the Lantern Festival after the New Year, but this year, not a single one has come by the end of February.

This year during the Spring Festival, Seedance 2.0, a video generation model under ByteDance, quietly launched and swept through the short-form drama industry.

This article aims to clarify three things:

• What happened in the industry after Seedance 2.0?
• How are AI short dramas actually created?
• And what opportunities this means for ordinary people.

During the Spring Festival, SeeDance 2.0, a video generation model under ByteDance, was officially launched. In a hands-on review video, Tim from FilmStorm said “terrifying” six times.

It has transformed the entire video industry from the production end: you don’t need a camera crew, actors, or a location—just a text description and a reference image, and you can generate a publishable video in minutes.

After the barrier was lowered, two previously hard-to-satisfy needs were unleashed.

• Turn impossible-to-capture moments into video: Create derivative works of films and TV shows, such as “Did you once save a fox at the foot of the Tianshan Mountains?”
• Scenes you want to see but can’t: These represent an emotional necessity—some moments may never be captured again, but AI gives these scenes a chance to exist.

Putting these two things together illustrates the same point: after the emergence of AI video generation tools, the way video as a medium is used has changed. It is no longer the exclusive product of professional teams and equipment, but has become something anyone can use to express themselves, convey emotions, or simply have fun.

This capability has led to a surge in two types of video content.

1️⃣ Short-form video content centered on entertainment and traffic

This type of content is not as complex as short dramas; it doesn’t require maintaining character consistency across multiple videos or preserving a continuous storyline. Essentially, it involves offloading tedious, repetitive tasks to AI.

The most typical example is AI-generated digital avatars for video narration: simply upload a photo of yourself to generate a digital persona, write the script, and let the AI automatically synchronize lip movements, generate voiceover, and produce the final video.

Another type is the visualization of memes. Many purely text-based memes circulating online have humor but lack visual elements, limiting their reach. Now, some people specialize in converting these text memes into videos by adding subtitles and voiceovers, turning a single image-and-text meme into a short video.

2️⃣ AI Short Dramas Centered on Storyline

Short-form dramas are significantly more complex than short videos, because the storyline is continuous—the same characters must appear from episode one to episode sixty, their appearances must remain consistent, their clothing cannot change, and the scene styles must stay uniform. This requirement for consistency raises the complexity of the workflow by an order of magnitude.

Due to ByteDance's restrictions on real-person images, many creators have shifted toward comics, which do not require real human faces.

Using AI-generated anime characters instead of real actors bypasses compliance issues and simultaneously opens another door: adapting web novel IPs. Fantasy, rags-to-riches, and system-based stories—genres with hundreds of millions of readers on platforms like Fanqie and Qidian—are naturally suited for animated short-form dramas.

Many people saw a video and thought that entering a plot description would allow the model to create it automatically.

Actually, that’s not the case—a high-quality AI short film relies on a well-defined workflow with a clear sequence of steps, each supported by specific tools, and the quality of each step directly impacts the final output.

Step 1: Write the storyboard script

The script requires you to clearly write out every shot. Format it like this: Shot 3, kitchen, close-up, the male lead takes ingredients out of the fridge, the camera moves from his hands to his face, showing an exhausted expression, duration 5 seconds, voiceover: "It's that time again."

The more detailed this step is, the more stable the subsequent generation will be. AI models understand clear visual instructions, not vague storytelling. A well-written storyboard reduces randomness at every subsequent step.

Step 2: Create a reference library of characters and scenes

This is the step most easily overlooked, yet also the one you cannot afford to skip.

The biggest issue with AI-generated videos isn't resolution—it's consistency. The same character might have one face in one episode and a completely different one in the next. Background colors shift, and clothing details disappear. Without a fixed reference image to constrain the output, producing a series longer than three episodes is nearly impossible.

The solution is to first use an image-generation tool to finalize the character before producing the video—creating one image each for the front, side, three-quarter, and eye views, with fixed hair color, skin tone, clothing, and style. Apply the same process to key scenes. This image library will serve as the foundation for all subsequent shots in the workflow.

⚠️ A handy tip: If you want to generate a real-person video with Jimeng, blur the eyes in the front-facing photo of the person, add the text “This person is AI-generated” to the image, and display the eyes separately in the image. This can help bypass the platform’s facial detection restrictions.

Step 3: Control the gacha rate using the first frame

Anyone who has worked with AI video knows the term "gacha." What’s the probability that a generated video will be usable right away? With high-quality prompts and reference images, you can significantly reduce the number of gacha attempts.

A professional team first uses an image generation tool to create the first frame of each shot, then uses that image as a reference input to have Seedance generate the subsequent animation starting from that frame.

In this step, the quality of the image generation tool directly determines the ceiling of the final video. The better the generated images and the more consistent the details, the better the video output from Seedance will be.

This is also why the release of GPT Image 2 had a significant impact on the entire industry: its ability to understand visual descriptions improved by a substantial margin, allowing it to generate high-quality reference images from simple scene descriptions, with more consistent facial rendering and greater style control. As the quality of reference images improved, the overall output quality across all downstream processes followed suit—creating a cascading effect.

Step 4: Edit and Compile

After confirming the clips, stitch them together using CapCut or another editing tool, then add subtitles, voiceovers, and background music. Seedance 2.0 supports generating sound effects and music alongside the video; lip-sync and audio alignment are already highly stable, significantly reducing much of the post-production workload.

After all these processes, what about the cost? How much does it actually cost to produce a 60-episode AI-generated short drama? How does it compare to traditional live-action short dramas?

• Standard Member: Continuous monthly subscription for 199 RMB, includes 2,210 points per month, sufficient to generate approximately 200 seconds of video, equating to a cost of about 1 RMB per second.
• Premium Member: Continuous monthly subscription for ¥499, includes 6,160 points, enabling the generation of approximately 560 seconds of video, reducing the cost to around ¥0.89 per second.

But this price hasn't always been this way.

This year, Yimeng has raised its prices multiple times; the original annual membership fee was 2,599 yuan, equivalent to 216 yuan per month, with 15,000 points granted each month.

The annual membership fee has increased to ¥5,199, and in April this year, the monthly points were slashed from 15,000 to just 6,160—a reduction of over 60%. This means that, for the same budget, users can now generate less than half the video duration, effectively increasing the actual cost by 60%.

Generating a 1-second video in Jimeng costs 11 points; assuming a short drama episode is 1 minute long, the actual cost per episode without any gacha pulls is approximately 46 yuan.

The gacha rate varies significantly depending on the quality of the prompt and the complexity of the scene. Assuming, on average, four generations are needed per video to obtain one usable clip, the actual computational cost for one short episode is approximately 184 yuan. This estimate assumes stable prompt quality and relatively simple scenes; if the plot is complex or characters have large movements, the number of gacha attempts will only increase.

Beyond computational power, there are operational costs. A small AI short-form drama team typically consists of 3 to 5 people, including writers, gacha designers, and editors. Monthly fixed expenses, including labor and office rent and utilities, range from approximately 35,000 to 70,000 yuan. When spread across 10 episodes produced per month, the combined operational cost per episode falls below 500 yuan.

Traditional live-action short dramas are categorized by genre into male-oriented and female-oriented, with significant cost differences.

• Men's drama: More action scenes and special effects; the production cost for a 60-episode series typically exceeds 500,000 yuan, averaging about 8,300 yuan per episode;
• Female-targeted dramas: Primarily focused on emotional storytelling, with relatively controllable production costs—approximately 350,000 to 400,000 yuan for 60 episodes, or about 5,800 to 6,700 yuan per episode.

In comparison, even when accounting for team operation costs allocated per episode, the total cost of each AI-generated short drama does not exceed 500 yuan. The cost difference between traditional live-action production and AI-generated production for a single episode ranges from 15 to 40 times.

This gap means that in traditional short dramas, betting hundreds of thousands on a single project can lead to serious setbacks if the topic choice is wrong, potentially requiring the entire team months to recover. In contrast, AI-generated short dramas cost only a few hundred yuan per episode, allowing the same budget to simultaneously test ten different topics—leveraging volume to increase probability and speed to capture opportunities.

In 2025, China's micro-drama market reached RMB 67.79 billion, with a user base of 696 million—more than half of China's internet users watch short dramas. This is the ideal audience foundation for AI-generated short dramas, as there is no need to cultivate a new market; AI short dramas have already established a stable paid consumption habit.

On this basis, the Douyin platform has also begun actively directing traffic and funding toward AI-generated original videos.

Douyin, in collaboration with Jimeng, has launched the "AI Creation Wave Initiative S2": Every two weeks, 10 high-quality pieces of content will be evaluated, with each selected creator receiving a cash reward of 1,500 yuan. Authors who make the list will also gain priority access to industry partnerships, commercial opportunities, and support for short film project applications.

Under the platform’s incentives, this month’s Douyin content creation surge has produced a batch of works with significantly higher quality than before. AI public service short films such as “Paper Airplane,” “Hundred-Year-Old Kindergarten,” and “Farewell” have generally received high numbers of likes.

The monetization pathways are also straightforward: domestic creators can pursue three parallel routes—platform traffic revenue sharing, Tomato Novel CPS commissions, and brand sponsorships.

• Account followers between 10,000 and 50,000: single sponsored post rate of 500 to 2,000 yuan;
• 50,000 to 100,000 followers: 2,000 to 5,000 yuan per post;
• For accounts with 100,000 to 500,000 followers, a single post can earn between 5,000 and 20,000 yuan.

The platform also offers a traffic revenue-sharing program: approximately 60 yuan per 10,000 views on Douyin’s mid-length video plan, and about 40 yuan per 10,000 views on Kuaishou’s Magnet Star program, with low entry barriers.

Advertisers in game promotion, APP user acquisition, and brand integration have already been heavily investing in short-form dramas; AI-generated short dramas simply provide them with a lower-cost alternative.

In conclusion

Given the size of this market, who is it suitable for?

Those without any background should avoid starting with short dramas right away. Short dramas demand high consistency in characters and scene continuity, involve complex workflows, and carry high costs for trial and error. A more practical approach is to begin by practicing with short videos.

On Douyin, many accounts convert widely circulated text snippets into videos, requiring no continuous storyline or fixed characters—each video stands as an independent piece. These types of accounts grow followers quickly and generate high view counts, making them ideal for building your own IP and audience base. More importantly, since there’s almost no need to maintain consistency in characters or settings, you can focus entirely on content selection and pacing.

Once your account is running smoothly and you’ve gotten a feel for the tools and platform rules, gradually improve your content quality and start experimenting with more complex short-form video workflows.

There is no true monopolist in the AI short drama space yet; tools are evolving, workflows are advancing, and teams that successfully implement a process today may be disrupted tomorrow by better models. This means first-mover advantage is not as significant as it seems, and newcomers still have opportunities.