AI News

One Year of the AI Video Boom: Creators Take Orders via Seedance, with Quotes Reaching 1,000 Yuan per Second

TMTPost · AGIUpdated Nov 16 18:41

AI summary

The AI video sector has exploded since the release of Seedance 2.0. Independent creators are monetizing by taking on commissioned commercial projects, with project quotes reaching 1,000 yuan per second, and some have gained 200,000 followers in just one to two months. However, Jimeng membership prices keep rising, with monthly credits for annual plans shrinking from 15,000 to 6,160, making model costs a core expense. Creators generally want to build IPs, but face pressure from a crowded sector and team-based competition.

Body · AI translation

By Xinmei | Written by Feng Ye | Edited by Zhai Wenting

Yi Ye is an AI video creator and also a student in the directing department of a film academy. At the end of April this year, he completed his first original AI video, "The Last Bus," about a bus driver's choice in the era of autonomous driving. After the video was released, the numbers skyrocketed—on Xiaohongshu alone, it got nearly 100,000 likes.

As a result, he also landed his first paid AI video gig: as executive director, he took part in producing the MV for Chopstick Brothers' new song "Xiaoyao Xian." After that, advertising requests from major brands kept coming, with project quotes reaching 1,000 yuan per second.

Yi Ye thought it over. "I did make a bit of money."

"The Last Bus" @Yi Ye

At a gathering of peers, the director of the AI short film "Unruly," "Yizhi Menglao" (hereafter Menglao), met many young people similar to Yi Ye. They are all independent creators who have earned income from AI videos they made themselves, and some have already set up their own "one-person companies." These creators, along with their works, become widely known on content platforms at astonishing speed. "Some people can accumulate 200,000 followers in one to two months," Menglao added. "But in other niches, that scale would take seven to eight months on average."

Industry insiders agree that the AI video track began to explode early this year, that is, after the release of Seedance 2.0. As video models' capabilities grew, professional and amateur creators kept pouring in, and demand from advertising, short dramas and other markets surged.

This group is a unique product of AI. Never before in any era has technological progress been so within reach of ordinary people. Their production and creation do not depend on organizations; they can complete it independently. Internet platforms handle content distribution and dissemination, while a mature industry chain helps them monetize quickly.

They have stepped into a brand-new world, facing unprecedented traffic, money and technology.

Their confusion followed close behind.

01 "The orders are almost more than I can handle"

A cruise ship floats on the sea; the camera moves upward as the wreckage of interstellar war spacecraft falls downward.

This was the first AI video clip Yi Ye made this year. To test the quality of Seedance 2.0, he gave it only a reference image of a cruise ship sailing and two lines of prompts. "For me, the effect was a bit shocking," he said, still remembering the sound effect of the wreckage hitting the sea. "The splash sounded very realistic."

He showed me the opening of a paid gig, eight seconds in total, with brilliant beams of light shuttling from the towers of the city's CBD into an apartment. If those eight seconds were made without AI, first you would need a 100 percent faithful model of the CBD; second, you would have to handle all the details of the buildings' materials, camera movement and so on. Finally, there would be rendering, the most time-consuming step. "High-precision rendering might take two hours to finish a single frame, but there are 24 frames in one second."

That ad was completed by him and a friend. A 90-second finished product took six days to make, and they earned 90,000 yuan.

Director Chen has done many kinds of work. After 20 years in e-commerce, he suddenly ran into the pandemic. Later he began shooting product videos and promotional films for e-commerce bosses, and in doing so he delved into AI text-to-image and video models. For a while, he also did AI top-up services. At the end of April, OpenAI released ChatGPT-Image-2, and the updated image-generation capability improved the quality of images used as video references. "AI's evolution speed is like a high-speed train," he said. He quickly made the decision to do AI video full time.

Director Chen's studio promotional film

"The orders are almost more than I can handle," Director Chen said. "Even while I'm chatting with you, I'm still working on storyboard images for a video." E-commerce product videos are the requests he takes on most, but there are also more "special" tasks.

For example, a client with a patent for retractable electrical cords asked him to make videos for products such as curling irons to show to different factories, so as to reach cooperation through patent transfer or licensing. "I made five or six videos for him in a month. He doesn't need modeling, and he doesn't need shooting. It's a very smart business."

Usually, the base quote for modeling a product ranges from 300 to 1,000 yuan, with sampling and video shooting priced separately, while Director Chen's base quote for making an AI video is 300 yuan per 15 seconds.

Previously, Yi Ye had also taken on traditional advertising shoots. He observed that after AI video models rose, the volume of traditional paid gigs had already dropped sharply by 40 percent. Model capabilities are still improving. "In another six months, that figure may become 60 percent."

In the industry, the form in which Yi Ye and Director Chen take on paid AI gigs is called "contract production," that is, making money by delivering AI videos to clients, in forms that can include ads, short dramas and more. This is currently the most common way for creators to monetize, and also the source from which they have tasted the sweetness of technological innovation.

But that sweetness comes with a hint of bitterness, and also a measure of difficulty.

02 Rising computing power costs, points that can't be saved

Among the creators Xinmei spoke with, the mainstream production method is to first use tools such as ChatGPT-Image for text-to-image, then use Seedance 2.5 to generate video. To ensure output quality, GPT Plus/Pro, Jimeng or memberships on aggregator platforms are almost standard for them. So model costs are the core expense of creators' contract production.

Director Chen bought GPT Pro and the highest-tier Jimeng membership, but when it came to costs, he hesitated a little. "It's hard to calculate. That number may change at any time."

The reason behind this is that Jimeng has been raising prices all along. The membership fee Chen the director currently pays has a list price of 11,699 yuan per quarter; the latest first-recharge discount is 43% off, or 4,999 yuan. But back in June, the first recharge cost only 3,999 yuan. At the start of the year, Yiye bought a Jimeng premium annual membership card for 2,599 yuan, which came with 15,000 points per month. Now the same annual fee grants only 6,160 points per month, a shrinkage of nearly 60%.

Jimeng official website membership pricing system

Other aggregator platforms that have integrated Seedance 2.0/2.5 are raising prices as well. LibTV has issued multiple official price adjustment notices, the most recent at the end of August, announcing that it expects to raise its annual card price by at least 1,000 yuan.

Chen the director said, "Seedance is already dominant, so naturally it has pricing power."

It is worth noting that Seedance, as the industry's absolutely leading video generation model, is also the first model to truly help ByteDance make money. 36Kr once reported that Seedance contributed more than half of the MaaS service revenue for Volcano Engine, to which it belongs.

Chen the director estimates, "I can keep production costs within 100 yuan per 15 seconds." Currently, the total membership fees he pays amount to four to five thousand yuan per month, and under a work mode of constant grinding, "working myself to death, I can make twenty or thirty thousand a month, at most."

Cost control has become a required course for creators. AI video creator ThxFish (hereafter Fish) told Xinmei that he has been in the industry for half a year, published nearly 30 original AI videos, and taken on dozens of commercial orders, but the total cost he has paid to AI video models is only 659 yuan, less than one percent of his income.

He alternates between multiple models when creating. "Kling's trial-and-error cost is relatively low." When choosing 1080P quality on LibTV to generate a 15-second video, Kling 3.0's point consumption is almost 30% of Seedance 2.5's. "And Kling's dialogue-scene performance is not bad, especially when doing empty shots and fixed shots," Fish said.

Even though Kling and Jimeng have both launched a one-time generation length of 30 seconds, he rarely tries it. "I don't go all out on gacha draws. I basically still use a generation length of 10 to 15 seconds. If the 30-second result isn't good, I have to draw again, and the cost will be higher."

Of course, tokens are never something you save your way to. To attract outstanding creators, aggregator platforms such as LibTV and Tapnow have launched various plans and activities that use points as incentives. After Fish published his second video, he was invited to join LibTV's incentive plan, which grants 20,000 points per month. The redemption condition is uploading two original videos longer than one minute on the platform each month.

Even so, Fish still finds that "after Seedance 2.5 came out, the points aren't quite enough." When taking on commercial orders, he negotiates with the client to have them bear the model costs. "If I have to both produce and buy computing power, I'd feel a bit aggrieved."

After Menglao finished the second episode of "Unruly," he reached a partnership with SenseTime AI. "I hardly think about points, and I don't have to work very cautiously." But after carefully studying the market, he found that although aggregator platforms cooperate with "super creators," this is more like marketing work to generate buzz and cannot bring direct revenue to the platform. It is a fact that large-model computing power costs continue to rise, and platforms do the math too. In the future, overall support will gradually decrease and become stratified: "Point incentives will concentrate on the top tier; mid-tier and bottom-tier creators won't get much."

This means creators need to be even more meticulous. Measuring costs like a company and obtaining as much revenue as possible is their daily homework.

03 "No energy, no fun, boring"

In theory, AI can already cover every stage of film and television production: from early-stage scriptwriting, storyboard drawing, and character styling, to mid-stage shooting, and post-production editing, special effects, and sound effects. But like all work that can use AI, what AI does is still the execution part; what truly determines the quality of the work is people.

Chen the director said that many platforms promote "AI storyboards"—"just type a few commands and dozens of images are produced immediately"—but he feels such images "lack aesthetic sense and artistic effect" and are hard to quote a high price for. "My clients are all repeat customers. Because all the storyboards are 'hand-made' by me." Usually he conceives the composition and shooting method of each image, uses PS or hand-drawing as assistance when necessary, and talks out each shot with AI. But this also creates higher labor costs.

His production speed is not fast; a video of about one minute is usually delivered within two days. He has carefully studied pricing, and such videos can be sold in volume on the market for one to two hundred yuan, but considering work hours, his pricing is several times higher. "If a client keeps haggling, I'll refer them to cheaper labor. But I won't take it, and I'm not willing to compete on price."

Content revisions are unavoidable. That ad beginning with "dazzling beams of light" took Yiye and his friends six days from output and revision to final draft. More accurately, it was "six all-nighters." "I've already lost count of how many revisions," Yiye said. He once posted a photo of coffee on WeChat Moments, and in the background, workflows on multiple screens were still running. "I dreamed the King of Hell praised me for being in good health," he mocked himself.

After taking on several months of commercial orders, Yiye's personal model costs came to about 10,000 yuan, with returns roughly ten times that. But he feels that for the first few ads, "it was pretty fun"; over time, all that was left was "boring, pointless, dull." "Money can't buy the joy of creating," he said.

Work takes up too much of Yiye's time: he has to carry a laptop when traveling, he's still color grading on his computer while in a taxi, and he can't fully engage at dinners with important people—"the boss is sitting right there, and I'm still at the table working overtime on my laptop."

Almost all creators want to build an IP, and this is also Yiye's main goal for the coming year. He mentions "The Laid-Off Girl," an AI short drama that has accumulated over 100 million views on Douyin; its AI protagonist Fang Taozi quickly landed endorsements and appeared on the cover of Fashion COSMO. As a content product, an IP has more flexible and diverse monetization methods, and creators have more say.

After Menglao released "Unruly," he quickly received many commercial inquiries, and he chose a robot vacuum placement that could tie in more closely with the plot. He describes that once an IP has buzz, it can not only take ads but also do image licensing, digital human livestreaming, and more—"now there are even AI variety shows, can you imagine?"

English text is being prepared.

However, this is not easy.

Menglao analyzes that at present, the main battlefield for AI video is still Douyin, and an IP needs to conform to traffic logic; strong conflict and tense plot progression are necessary. "Even MCNs may not be able to do this well—hit projects are the very few that internal teams grind out. Creators working alone will find it even harder."

Second, IP content with monetization potential requires upfront business planning.

English text is being prepared.

04 A one-person company—does that mean no need to collaborate?

In mid-October, Fish's short film was selected for the World AI Film Festival. Since April, almost all of his WeChat Moments posts have been related to AI video. He once intensively attended creative events in Seoul, Hangzhou, Pingyao, and other places within a single month; in his posts, the emojis he uses most are "sun" and "spinning circle." This matches how creation makes him feel: elated and delighted.

Fish is not formally trained. But like all creators, he loves movies and likes Lee Chang-dong, Yasujiro Ozu, and Fellini; his bio reads "world-renowned moviegoer." As an undergraduate, he tried to cross-discipline into a film graduate program but failed. After graduation, he wanted to find work in the film and television industry, but it was "not very satisfactory."

Until AI video models emerged, he could finally express his ideas completely independently. He had almost no live-action filmmaking experience; when making his first original AI video, he drew on the theoretical materials and film-watching experience he had accumulated while preparing for exams, and the final result left him "extremely excited." Traffic, applause, and commercial orders followed one after another.

Fish likes contract production work. "I think I'm good at this, and it doesn't feel tiring to do."

Short film "Dream," selected for the World AI Film Festival @Thxfish

Sometimes he also feels anxious, and this anxiety is growing. He knows he is among the first batch of people to use AI video for commercial projects, but as more people enter the field and model capabilities improve, the barrier to AI video creation is lowering. "I worry that monetization ability will also weaken."

This confusion contains a hidden worry about "individual limitations." Fish is clear that "a person's time and energy are ultimately limited."

"The people entering now are not just individuals, but also MCNs and bloggers transitioning from traditional short video. Their capabilities are crushing-level, and team operations are more efficient," Menglao analyzes. The track is becoming crowded.

Paul Jarvis, author of the book "Company of One," believes that "the key to a company of one is to stay small and not pursue unnecessary growth. This does not conflict with seeking collaboration." He also says that hiring freelancers as outsourced help is his largest expense each year.

This means that if independent creators want more sustained, long-term development, collaboration is almost inevitable.

Fish decides to dig deeper in the direction of a studio. He brought in a friend who is good at script creativity; he himself handles image generation and post-production, and all business negotiations are handed to the other person—Fish considers himself relatively "i." He once wanted to focus on festivals and competitions, but on second thought, "after winning an award, so what? In the end, you still have to make money."

Menglao, meanwhile, has a new identity: head of the AIGC label "Romantic Front." "When you hear the name, you know I'm a bit idealistic," Menglao laughs. "I hope to gather creators through the label so everyone can collaborate with each other. At this stage I want to be purer and think more about what I can provide everyone, such as distributing some useful AI tools, computing power credit packages, and so on." "Unruly" will still be updated, but Menglao says frankly, "it may appear more under the label's image going forward."

English text is being prepared.

"People definitely do whatever makes money," he says.

Source

TMTPost · AGI