Is the world ready for a giant worm going to work on the subway? Korean director Kim Sung-soo thinks so and is making a film on this very subject, Worm Train, inspired by Franz Kafka’s short story Metamorphosis.
Kim, whose credits include major Korean blockbusters such as Asura: The City Of Madness and 12.12: The Day, talked about how he’s using Kling AI tools to make the short film on a panel at Busan’s Asian Contents & Film Market (ACFM) today.
Kling AI, which recently released its 4.0 image generation model, also invited Chinese animation filmmaker Li Wei (Big Fish & Begonia); Harry Hyun, a director and CEO of Korea’s MooAm Co; and Lee Sang-wook, a producer and team manager at Korea’s MBC C&I AI Contents Lab, to present recent projects made using Kling AI tools, as well as discuss the latest trends in this new mode of filmmaking.
After presenting a teaser for Worm Train, Kim said AI has changed the way he thinks about filmmaking, but that “one thing never changes, which is what makes a film cinematic”.
“Now everyone has similar tools and anyone can access Kling, but while some will make a masterpiece, others will make a film that is mediocre,” Kim said. “Initially, I thought AI would destroy the industry, but I don’t think so any more, because technology does not replace creativity. The value of images still comes from human intervention.”
Hyun, who has made AI films including Crypto Man and Genfluencer, expanded on this topic. “I don’t think AI production compared to traditional production is really that different – the difference is efficiency and simplicity – but if you don’t have experienced filmmakers on set, everything can end up looking the same.
“That’s why AI images may look impressive at first, but not later, because everyone is using the same lighting, same everything,” Hyun continued. “You need an art director who can do the fine tuning, a director and cinematographer who have really studied and learned when to use a close shot, wide shot, how to frame an image. These are long discussions we have on set, they’re not random decisions. Only when you’re working with experts will you be able to create something different with AI – not just cookie cutter deliverables.”
Li Wei, who is making animated feature Born Of The Tide with Kling AI tools, said one useful aspect of AI filmmaking is that results can be seen right away, so it’s easier to get feedback from multiple people on the crew. “There’s going to be a lot of trial and error in AI filmmaking, but effective communication can be used between the various production members, so everyone has a chance to feed back into the production,” Li said.
Lee presented a teaser of upcoming series Rendering Love, produced by MBC C&I and Barunson E&A, which combines live-action scenes filmed the traditional way with AI generated characters to tell the story of a virtual K-pop group. He observed that different roles and ways of working are emerging in AI production crews.
“All staff members working on a project must fully understand the technology. That’s the only way to have high quality content. If you’re updating to a new model, the team leaders need to analyse the new features and share that understanding so all staff members are on the same page. The technology is changing constantly and if anyone is lagging behind, overall quality can decrease.”
The speakers also discussed whether there was a need to introduce industry standards in AI filmmaking, but Kim suggested it was still too early, at least in South Korea. “Creators in Korea have their own ideas and imagination and want to unleash that using AI tools. We need to see where all this is heading before we start talking about regulations and standards.
“But one thing I’m very aware of as a filmmaker is thinking about whether you’re infringing somebody’s rights. That’s something we need to be conscious of from the early stages.”
The panel, entitled New Cinematic Possibilities: Reimagining Filmmaking with Kling AI, took place as part of the InnoAsia track at Asian Contents & Market (ACFM), which is running until October 13.
