Russian filmmakers are actively integrating neural networks into movie production across scriptwriting, pre-production, filming, and editing, but artificial intelligence cannot yet fully replace human professionals, according to a study published by news outlet RT.
While artificial intelligence tools can generate striking visual sequences, the study found that neural networks struggle significantly with dramaturgy, narrative depth, and convincing acting performances.
Industry experts estimated that artificial intelligence allows production timelines for a feature film to be compressed from three years down to just six months. A primary example cited in the study was Hell Grind, marketed as the first neural film, which was screened at the Cannes Film Festival in May 2026. However, critics dismissed the project as a technological publicity stunt rather than a true work of art, pointing to a flat screenplay, visual artifacts, and an absence of human emotion.
Scriptwriting and pre-production tools
Nikolai Shishkin, a screenwriter known for his work on the superhero film Major Grom: Game, noted that text-based artificial intelligence can accelerate research and fact-checking, but remains incapable of writing nuanced dialogue with emotional subtext.
Pre-production has emerged as one of the most effective areas for AI deployment. Dmitry Averkiyev, creative director of animation company Yarko, confirmed that neural networks prove particularly useful during early development for establishing visual styles, exploring character variations, and rapidly converting two-dimensional sketches into volumetric images.
Yarko is a Russian animation studio that produces animated series and media content, while major film franchises have increasingly adopted digital asset workflows to shorten early production cycles.
Technical leaps and digital actor costs
The study highlighted a major technical leap on set over the past 1.5 years, with the maximum length of a coherent generated shot increasing from four to eight seconds up to 30 seconds in a single pass. Filmmakers are using models such as Google Nano Banana, which can maintain up to 50 reference points to preserve character consistency across shots.
However, technical obstacles persist during editing. Yegor Ershov, head of a laboratory at the Moscow Institute of Physics and Technology, warned that after ten editing cuts, no existing model can guarantee the complete preservation of a character's face, clothing, or overall texture.
Digital actors remain the most contentious issue in the industry. Real actors demand payment equivalent to full shooting days for their digital likenesses, while directors report that although a neural network can depict an emotion, it lacks any comprehension of an actor's dramatic objective.
In hybrid production, artificial intelligence is increasingly replacing traditional visual effects by generating scenes that would be impossible or cost-prohibitive to shoot live. The study noted that a scene involving semi-trucks, which would have cost 150,000 rubles to shoot live, was completed for 15,000 rubles using AI generation.
Post-production workflows and critic adoption
Artificial intelligence is also being used in post-production to organize raw footage and assist with editing, though experts noted that its speed and output quality still lag behind human editors. While generating a one-minute video clip requires only a single button press, producing one minute of finished, review-ready footage takes anywhere from three days to several weeks.
The findings follow a separate investigation by culture publication Afisha Daily into AI adoption among film critics. That survey revealed that 52 percent of film critics utilize artificial intelligence tools, primarily to handle routine administrative tasks rather than to generate written film reviews.
