Russian filmmakers are actively using neural networks to generate backgrounds, crowds, and secondary characters, as well as to automate routine operations and speed up film production processes. At the same time, the question of whether AI will replace directors, artists, and screenwriters is no longer on the table. Artificial intelligence-based solutions are merely tools. Making films is a job for humans. This was the conclusion reached by the participants of the open session at the 3rd Moscow International Film Week, which took place at the Khudozhestvenny cinema as part of the annual "Cinema Night" event.
"Soyuzmultfilm uses neural networks to create short videos featuring characters from popular animated films, including Cheburushka, because performing such tasks manually is time-consuming and expensive. AI also helps the studio's animators generate character prototypes, costumes, locations, and props—objects that the characters interact with," said Yulia Osetinskaya, General Producer of the Soyuzmultfilm studio.
The expert also noted that the studio uses neural networks to generate crowds in animated films and to convert content into 3D format.
"We created a wonderful children's project, 'Umka', in 2D. A total of 104 episodes are already done. Now we are running an experiment to convert them into 3D. It seems to be turning out well. Perhaps we will even be able to restore the 'Soyuzmultfilm' golden collection of animation and make it more modern," Osetinskaya noted.
According to director, screenwriter, and general producer of AP Entertainment Dmitry Rudovsky, the use of Unreal Engine, LED screens, and generative content can significantly reduce production costs for large-scale historical projects.
Generative models are most effective for creating short formats, such as trailers, promo videos, and vertical series, believes Sergey Ryabov, CEO of AI Lab.
"What can be done easily and effortlessly using neural networks? Trailers. There are no long and complex director's scenes, dialogues, or emotions. There is only an action-packed montage, which AI handles very well," he said.
The session's experts agreed that a neural network is merely a tool. The success of a project still depends on screenwriters, directors, artists, and producers.
"Generative cinema must be made by people who know how to make films. Because a neural network is just a tool," Sergey Ryabov believes.
Cinema cannot be left entirely to neural networks, agreed Yana Shmaylova and Ekaterina Kononenko, co-founders of the Forge Cinema AI platform. They also spoke about the difficulties of using neural networks in filmmaking. According to them, one needs to generate a thousand minutes of video just to select a single minute that will make it into an action scene. This requires massive resources.
The session participants identified the combination of traditional filming and generative technologies as one of the most promising areas. Neural networks help optimize costs and speed up the film production process, but they should not be abused. "In the series *Kaifsk. City of Dwarfs* (produced by 'Mesto Sily' for Okko), we decided to go the route of combining live-action footage with generation," stated Sergey Vasilyev, the series' director.
The series' creative producer, Ilya Shutov, added that although the town of Kaifsk—located inside a St. Petersburg apartment—was generated, dozens of artists worked on its sketches.
Magic Studio co-founders Yulia Lipatova and Makar Kozhukhov shared their experience of optimizing processes in filmmaking using new technologies. "The machine takes care of the main routine. But our guys—neuro-operators, directors, screenwriters, and storyboard artists—are actively monitoring everything," Yulia Lipatova said.
