Text | 阑西
We still don't know why Kuaishou — and not other big players — pioneered the development of a Sora-like video model.
The date of Keling's initial release, just half a year after the amazing birth of Sora, in this half year, the AI industry is still iterating at a rapid pace, but the progress of the video model is still difficult, in addition to being a "futures" Sora is still hanging the market appetite, and there are almost no products that can be compared with it.
Until Kuaishou's self-developed Keling suddenly landed, whether in terms of picture smoothness or overlapping image fit, it completely threw off those peers who stayed at the level of "picture dynamics", and it was "spot", as long as the application was queuing, the qualification could be issued to any user.
Only a month after Keling's debut, Kuaishou produced China's first AIGC original fantasy micro-short drama "Mountains and Seas Mirror: Splitting Waves", the whole film does not have any live actors and physical props, and the in-depth technical support of Keling completes the film.
Kuaishou even held an offline screening meeting for this short drama, in order to benchmark the ritual sense of the traditional film and television industry, ringing the crisp bell that the change has arrived, and the story that happened at the same time is that the starry short drama - Kuaishou's own head content label - launched a creator incubation plan with Keling large model, to give enough traffic and cash resource support for AI short dramas.
To a small extent, this may just be Kuaishou's business strategy, which is to make full use of Keling's first-mover advantage to add value to specific businesses, and to put it in a big way, China's AIGC industry is trying to find a highly pragmatic route, and even has the opportunity to reproduce the historical laws of the consumer Internet, and the breakthrough of the technology layer can be paid for by the prosperity of the application layer.
· · ·
Chen Kun, the director of "The Mirror of Mountains and Seas: Splitting Waves", actually uploaded the trailer "Strange Mirrors of Mountains and Seas" generated by him with AI on major platforms as early as January this year, which can be regarded as the predecessor of "Strange Mirrors of Mountains and Seas".
If you look back at the concept film that is less than 2 minutes long, it is not difficult to find the obvious generation difference in technology: it is essentially a PPT-style film, without lens language and transition screens, and it belongs to stacked visual animations.
Later, through the connection of Kuaishou short dramas, Chen Kun had in-depth exchanges with the Keling team, and then the version of the IP of "Mountains and Seas Mirror" was upgraded, until the finished film of "Mountains and Seas Mirror: Splitting Waves" was released, and the sound and painting generated by AI began to have a soul - or the presence of the director's intention - the performance of camera movement, action, and narrative has reached the ceiling level of current AI videos.
"Mountains and Seas Mirror" before Keling's launch
Keling's "Mountains and Seas Mirror: Cleaving Waves"
Chen Kun also said that in the process of making this short drama, the most common emotion experienced is dissatisfaction, the creative team is not satisfied, Kuaishou is not satisfied, and even when it comes to the stage of the test version, it cannot be said that it has reached complete satisfaction.
Creators who use AI intensively should be familiar with this state, and the industry also calls the experience of AI-generated pictures/videos "card gacha", which is a small probability event to be able to get content that fully meets the needs, but "card gacha" also means that the threshold for creation is greatly reduced, and anyone can dispatch AI to realize their imagination and narrative.
Just like each shot of "The Mirror of Mountains and Seas: Chopping Waves", it may have undergone hundreds of generations and corrections, but compared to the spectacle it finally presents - the steaming and backward sea, the Kunpeng swallowing clouds and mist, and the co-workers pushing the mountains - the actual cost of computing power is completely bearable.
You must know that the annual cost of traditional film and television post-production companies on renderers can easily exceed hundreds of millions of yuan, and the labor is even higher, and the special effects of sensory blockbusters such as "Avatar" and "Transformers" can often account for more than 60% of the total production cost.
Knowing the cost of burning money on such a large scale, it is also possible to understand why Sora has taken a different path of penetration from ChatGPT, keeping a distance from the geeks in Silicon Valley on the one hand, and continuing to courtship Hollywood studios on the other.
Chen Kun worked with Zhang Yimou in a variety show in his early years, coinciding with the premiere of Zhang Yimou's Chinese New Year film "The Great Wall", he asked Zhang Yimou why he didn't make gluttony more intelligent, and the answer was that the budget was limited and there was no condition to support more detailed displays, which made Chen Kun still remember it to this day.
It is said that AI is an enabling tool, but in fact, the premise of empowerment is empowerment, giving human beings the right to dare to think and shoot, and not to let the desire to create be lost in the impossible wilderness.
In addition to demonstrating the correct use of the video model, it has also become an invitation to all creators and film and television organizations: Now, immediately, immediately, enter the AI short drama.
· · ·
For a long time, the embarrassment of AIGC was that the industry was telling AI stories, not using AI to tell stories.
Telling AI stories is very simple, and Keling can also do it, for example, swiping the screen on social media to make photos move, which is a lot for Keling and even AI technology, and everyone can feel that AI is human, not naturally alienated from ordinary life.
It is no longer a short-lived communication material such as shooting the same model or creating a hot spot, but truly using AI as a productivity tool, empowering this "high-precision" technology to the PUGC creator group and creating an emerging content market.
After all, new AI products have been launched almost every day in the past few years, but only a handful of them can find landing scenarios like ChatGPT, Midjourney, and Adobe Firefly, and most of them are idling in an atmosphere full of subversiveness.
Kuaishou's daily active users are almost 400 million, of which 270 million - equivalent to 7% of daily active users - consume short drama content every day, in other words, the form of short drama has deeply penetrated into Kuaishou's content pool, and exerts pressure on the production capacity of 100,000 short drama creators on the supply side.
Therefore, Kuaishou has to take the initiative to go down the market and use its own label and its own model to accelerate the "population expansion" of AI creators, which is different from the solicitation activities organized by other AI manufacturers in order to promote products, Keling itself does not exist independently, it serves Kuaishou's content business, and acts as an infrastructure.
As early as March this year, Kuaishou disclosed the layout of short dramas + AI, relying on a sufficiently rich short drama ecology - according to public data, Kuaishou should be the platform with the most daily active users of short dramas in China - creators usually do not ignore signals from the official, and the supply and demand ecology of short dramas is already hot, which also provides the best time for AI to enter the game.
According to the statistics of iiMedia Consulting, the market size of China's online short dramas last year has reached 37.39 billion yuan, you know, the cumulative box office of Chinese films in the same year was only 54.91 billion yuan, and hundreds of millions of users unlocked and paid for one dollar or two yuan in the short video App, supporting 70% of the output value of the entire film industry, and all this has grown in just a few years.
In the incubation plan proposed by Kuaishou, the economic account of the creator is very calculated: all the costs only depend on the manpower and flexibility of the creative side, and the benefits include the investment of Xingmang Theater in key projects, the sharing of natural broadcast volume, the platform-led investment promotion services, and the promotion and distribution of the whole network.
It can only be said that under such conditions, all that can be lost are talents.
On the other hand, just when the Internet has been filled with too many Erchuang activities of Mama Rong eating hamburgers, Wu Jing and Donnie Yen kissing, the imagination space of AI in deep processing is actually underestimated.
Perhaps, in Chen Kun's hands, Ke Ling will be his artifact to reproduce hundreds of strange beasts in the "Classic of Mountains and Seas", but in the rural live-action "One Piece" team, which is known for its five-cent special effects, Ke Ling does not need to replace any actor, but its video continuation ability can more vividly realize the rubber special effect of arm elongation for Rizhao Luffy.
On top of that, AI can finally stop being an expensive toy.
· · ·
Many people may not know that 3D cinema was invented as early as the fifties of the last century, and Hitchcock made "Murder by Telephone" in 1954 through stereoscopic film technology, and was nominated for Best Director at the United States Directors Guild of Directors Guild Awards in 1954.
However, it was not until more than half a century later that Cameron's "Avatar" really popularized 3D movies to the field of mass consumption, and since then it has created a wave of mainstream 3D movies.
In the middle of a few decades, it is a long climbing period in the technology diffusion curve, when the equivalent change is insufficient to cause qualitative change, we can only hope that the qualitative change will cause both sides, and the "Avatar" that exploded at the box office is the qualitative change that the film industry has been waiting for for many years.
"Mountains and Seas Mirror: Splitting Waves" may not belong to Keling's "Avatar", but what Kuaishou is promoting is to make hundreds of "Mountains and Seas Wonders Mirror's Splitting Waves" grow on the platform, with the sustainable ecology of short drama + AI, there may not be a need for "Avatar" at all, and Keling's infrastructure and inclusiveness have been completed.
So back to the original question, why Kuaishou?
The answer may be close at hand: only by truly understanding the source of content and understanding what problems creators are trapped in can we know what kind of AI to develop and let AI serve real needs.