AI-Generated Script for the Sidorov Series: Production and Perspective

No time to read?
Get a summary

The Sidorov series was crafted with the help of an AI TV Script Generator neural network created by Vladimir Larkin and Vadim Povolotsky, researchers affiliated with the Skillbox education platform’s Data Scientist PRO program. They fed the system scripts from popular projects such as Rodkom, Voronins, Dyldy, The Eighties, and The Oligarch’s Wife to fine-tune the model. Through thousands of pages of text, the algorithms were adjusted to resemble authentic movie scripting more closely, enabling the network to learn from a wide range of narrative patterns.

In the project, the neural network absorbed 355,892 words across 88 episodes and seven TV programs. To maximize quality, twenty experimental runs tested various parameters and data feeding methods. By inputting a small text fragment, the network could suddenly generate an infinite variety of scenarios in less than a hundredth of a second, as explained by Vladimir Larkin.

Over time, the AI learned to autonomously determine the where and when of action, character names, copies, descriptions, and other fundamental elements of the script.

The project director, Sergei Znamensky, previously worked on comedy hits such as Ivanov-Ivanov and The Family. He emphasized an aim to minimize manual edits to preserve the original neural network idea:

“We take the situation generated by the neural net and try to justify it through acting and directing, even if the logic isn’t initially clear. We search for it, add it, and refine it. If everything were filmed exactly as written by a neural network still in training, it would feel like a theater of moving doodles with text.”

The AI-created cast featured an ordinary family: husband Georgy Dronov, wife Alina Alekseeva, son Sasha Novikov, grandmother Tatiana Orlova, and grandfather Sergei Stepin. They leave their home, hear off-screen laughter, and stumble into humorous, anecdotal situations. When the Sidorovs realize they exist as characters in a sitcom, they attempt to break free from the setup.

“This story resembles Groundhog Day for our heroes. The family mansion setting stays constant for the 385 episodes. Yet a moment comes when the main character asks why there is nothing but cabbage soup to eat and why the grandfather goes fishing every day, sparking a desire to escape the landscape,” Znamensky noted.

Casting established the on-screen team, with auditions spanning several months. Short reels for secondary roles were sourced from the VK Clips platform. The winner of the casting was Maxim Sushko, a 31-year-old dancer, actor, and model from Murmansk who plays the bumbling neighbor. Tatiana Orlova, who previously collaborated with Znamensky on The Family, was swiftly approved for her role.

Orlova suggested that producing the Sidorovs with a neural-network scenario did not feel markedly different from standard projects. Since she did not know the screenwriter behind the AI-driven script, she approached the role as if the writer were a figure who could be a bit cheeky with the text and characters.

Responding to this, Dronov highlighted a clear distinction between AI-generated and human-authored storytelling. He described the neural network as still young and acknowledged a desire to modify considerable portions of the outcome.

“We were given a mandate to justify as much of the neural network’s output as possible, but there is a lot I would change,” he added.

For the actor, integrating an understanding of creativity and the comedy genre’s rules into a neural model remains challenging. The sitcom format relies on fixed character masks during the performance, with each character bearing specific traits: one is cautious and kind, another is foolish yet brave, and a third is treacherous. The overall arc unfolds within these dynamics. It seems the AI might push drama faster than comedy, which would elevate the genre yet remain among its hardest forms.”

Developing a script with a neural network was described as a bold experiment by the cast. Dronov joked about the possibility of robots replacing human actors, noting that producers could simply run the program and continue working. Meanwhile, AI might not yet consistently match the performance quality of human performers. The Sidorovs’ apartment set was built in the same Amedia pavilion that housed the Voronins for a decade. Dronov observed a noticeable shift in the neural network’s script toward that familiar sitcom rhythm. The project team draws from past work on Ivanov-Ivanov and The Family to shape the new series.

— This creative approach reflects ongoing explorations into how artificial intelligence can contribute to serialized storytelling, while still relying on human direction, acting nuance, and editorial insight to shape a coherent, engaging narrative. [Citation: AI TV Script Generator project, research notes and project brief].

No time to read?
Get a summary
Previous Article

Coconut Kitty and the Diana Deets Story: A Cautionary Look at Online Personas

Next Article

A proposal to issue one-million-ruble certificates for family car purchases