Box office success thanks to artificial intelligence

- EN - FR
Sami Arpa,cofounder of the spin-off Largo Films © 2020 Alain Herzog
Sami Arpa,cofounder of the spin-off Largo Films © 2020 Alain Herzog

Largo Films, an EPFL spin-off, unveiled its data-assisted movie-making software to producers and other movie-industry specialists today at the Berlin Film Festival. Largo’s program uses artificial intelligence to generate recommendations for just about every stage of the filmmaking process - from putting together a script and soundtrack to selecting the actors and emotional register - all in just a few minutes.

Movie production is a risky enterprise, even for major film studios, because the chances of a flop are just as high as those of a blockbuster. And which of these two categories a movie falls into is often unpredictable. However, new software developed by Largo Films aims to help producers better predict whether a movie will be a hit by taking an analytical approach. The software uses artificial intelligence to analyze a movie’s script, estimate its box-office revenue based on how famous its actors are, and suggest what scenes could be changed to appeal to a given target audience. All that can save producers considerable time and money. The software, called LargoAI , was unveiled today at the Berlin Film Festival. By going step-by-step through the decisions involved in making a film - from finalizing the script and selecting the actors to creating a soundtrack and identifying what post-production edits to make - LargoAI can provide valuable insight into which elements to adjust for maximum impact.

200,000 actor profiles

To develop the software’s artificial intelligence algorithms, Largo’s founders - graduates of EPFL’s Image and Visual Representation Lab - first identified hundreds of variables to consider and incorporated data from numerous databases, including one with 200,000 actor profiles. Then they ran nearly 40,000 films through their program to train the algorithms. The result is a deep-tech system where users can easily adjust key variables in just a few clicks. The program’s findings are generated in just a few minutes and displayed in colorful, easy-to-read charts and diagrams.

Largo’s founders tested their program on several movies, including some that had already been released. They found that their algorithms were highly accurate. "For example, for Venom - a US superhero film that came out in 2018 - our program estimated $201 million in box-office revenue, and the film actually took in $213 million," says Sami Arpa, one of Largo’s founders. For Domani č un altro giorno - an Italian feature-length film by Simone Spada - LargoAI estimated box-office revenue of between ¤1.6 million and ¤3.9 million based on the film’s script; this range proved to be spot on. And for The Gentlemen, released in the US on 24 January, the program estimated proceeds of $10.6 million in its first week - which was exactly how much the film earned.

Analytics for independent filmmakers, too

Independent filmmakers have long shunned the use of algorithms for making cinematic decisions. Even the major studios and streaming companies have only recently begun admitting to using this kind of technology. Warner Bros announced in early January that it had signed a deal with an AI analytics firm to help "estimate movies’ profitability and develop production budgets," but that the AI software would not be used "in the creative process." Largo views this as the first step towards a larger opportunity. "Compared with what’s already on the market, our program is designed for use earlier in the production process and encompasses a wider range of criteria," says Arpa, himself a film buff who makes movies in his spare time. "It can assess a movie’s potential and enable producers to make adjustments accordingly, such as by getting rid of a specific scene, changing the soundtrack or finding an actor who better fits a given character. And it can calculate the financial consequences of different options."

LargoAI is designed not just for major studios but also for independent filmmakers, with a variety of features targeted to them specifically. "For example, it can suggest which festivals directors should show their films at to get the most out of their time and money," says Arpa.

Filmmaking will always be an artistic endeavor

While some producers have tried using artificial intelligence to automate certain steps of the filmmaking process, creating a good film will always remain an art. A successful film depends on a variety of factors: a compelling script, strong emotions, excellent actors and a healthy dose of talent - which is something a supercomputer can never replicate. Arpa doesn’t see his program as being a foolproof way to create a great film, but rather as a tool for helping producers make data-driven decisions. That’s also because LargoAI has the same drawback as all other AI-based technology - it relies on historical data. But the movie industry is continuously evolving with new story lines to invent and new generations of actors to discover. And it’s not unheard of for low-budget films with unknown actors to end up being big box-office hits. What’s more, viewer reactions can change in response to shifts in societal values - something even the best artificial intelligence programs can’t predict. Yet another reason why the artistic touch is so important.