Forty-five-year-old film director Huang Lei and Wu Hankun, a young actor and AI creator, formed the team Lightcone. Huang confessed that he was pressured to get involved with AI because he has had colleagues who have had difficulty getting their film projects made in the past few years.
"When you don't have adequate funding or the resources to make a feature film, at least AI can help you construct a preliminary outlook of what you want to show. That's very useful in the fundraising stage of movie production.
"Very often we found AI actors' performances stiff and unrealistic, so we decided to take advantage of Hankun's background as a professional actor," Huang added. "We shot his performances first, then used AI to transform his appearance for different characters, while retaining the authentic details of human acting."
But that didn't completely solve the problem. "AI tends to smooth out performance imperfections, such as micro-expressions and vocal cracks," Wu said. "But these are exactly what make your acting moving, personal, and real."
Also, purely AI-generated performances suffer from inconsistency and require massive amounts of regeneration, he added.
On the other hand, the integration of human performance with AI also helps navigate portrait rights issues, said renowned film producer Ren Ning, highlighting the need for better industry-level digital asset protection.
The current tools cannot completely avoid infringing portrait rights, which presents an important challenge for filmmakers working with AI all over the world. "Once we get all these industry-level setups done, we can let AI empower us better and make the path for industry development smoother."
On June 19, an academic report was released on the AI Backlot initiative, written by the observation team from the School of Theater, Film and Television at the Communication University of China.
The report concluded that while AI can assist, humans still remain the ultimate creators. AI can't replace people's judgment on camera work, storytelling and audience experience. When AI struggles with handling long takes, a director's value shows in how they break down, reorganize and use editing to create a movie scene.
