Quilty: the AI that promises to predict box office hits

Quilty, a technology company, claims it can predict a film's box office success simply by analyzing its script. However, early users of this tool are skeptical. Indeed, Quilty incorrectly predicted that the script for Christy, a commercial failure, would outperform that of Sinners, an Oscar-winning blockbuster.

Quilty's founders, Simon Horsman and Daniel Wood, hope their technology can "democratize" the film industry by providing emerging creators with assistance tools. A good Quilty score could open doors with producers, while a low score might indicate the need for revisions. Yet, Quilty appears to be a compilation of existing AI systems, and its effectiveness in identifying a future hit remains to be proven.

Quilty uses AI to analyze scripts and generate detailed reports on a project's chances of success. When an unproduced script is submitted, the technology assigns a score from 0 to 100, reflecting the quality of the storytelling, its commercial viability, audience appeal, and the likely production cost. The platform promises users a glimpse into the future as they seek to validate their films.

In a recent meeting, Horsman and Wood emphasized the importance of "keeping humans in the loop" rather than fully automating the pre-production process. They sought feedback from other creatives who are concerned about the potential impact of AI on jobs.

Horsman stated, "We agree with a lot of the negative sentiments towards AI, but what we're trying to do is enable human creativity." Quilty aims to provide users—whether they are screenwriters, producers, buyers, financiers, or studio executives—with the maximum information to make informed decisions.

Instead of offering a single AI model for script feedback, Quilty combines several widely available AI tools to provide different types of analyses. Users simply upload their scripts, and a few minutes later, they receive a report detailing elements such as an estimated budget, summaries of key story moments, and character analyses. The service costs $50 per individual analysis, with discounted rates for multiple analyses.

The idea for this type of analytical workflow came to Wood, who is also Quilty's CTO, a few years ago when he was involved in a real estate matter. Rather than spending money on a lawyer, he used ChatGPT, which quickly informed him that it was not a lawyer.

Wood then tried Gemini, which worked better at the time. His experience allowed him to better understand how different AI models can excel at various tasks. Quilty uses Gemini to generate breakdowns—documents that summarize all production elements of a film or series in comprehensive lists. For financial modeling, the company uses a DeepSeek model hosted in the United States, and for narrative and character analysis, it combines Claude and ChatGPT.

Wood explained that the company relies on context prompting—a process where additional contextual data is provided—to generate quality results. Quilty does not personally train the models it uses to create reports on films, but Wood insisted that this is a strength, as it facilitates the incorporation of improved new models into its workflow.

While the modularity of Quilty's technology may make it more agile for updates, it also complicates understanding how the platform analyzes a script and generates metrics intended to measure intangible elements, such as audience reaction to a film that does not yet exist. Prediction has always been an essential part of film development, but this work has traditionally been done by humans with a nuanced understanding of audiences.

No AI company has succeeded in developing a model capable of truly replicating human thought processes or the imprecise way we form our opinions about art. However, Quilty's founders believe their "sentiment engine" is the best alternative for evaluating scripts, thanks to tools like VADER—an open-source software that measures the degree of positivity or negativity in a text.

Horsman and Wood are confident that Quilty can accurately determine how a project "responds to the cultural moment" and provide reliable box office projections. They cited Revenge of the Nerds as an example of an older film that would receive a lower Quilty score due to its portrayal of sexual assault in a comedic light, which modern viewers would find in poor taste.

When I asked Horsman and Wood why Quilty assigned a higher score to Christy (which ultimately grossed about $2 million) than to Sinners (which grossed $370 million), they insisted that the platform's judgment was based on Sydney Sweeney's popularity. They explained that, on paper, Sweeney's star power, coupled with the fact that biographical dramas about boxing are less expensive to produce than fantasy/action films like Sinners, made Christy a safer bet. However, this situation highlights the limited reliability of Quilty's logic.

Horsman and Wood admitted that there are situations where Quilty cannot anticipate factors that could influence a film's financial performance or how the audience receives it. For example, Quilty would not have been able to foresee that Magazine Dreams by Elijah Bynum (produced by Horsman) would be disrupted by the downfall of actor Jonathan Majors in 2023. Similarly, nothing in the script of a Minecraft Movie could have indicated to the platform that the Chicken Jockey phenomenon would become a key element of its success.

For all its promises, what Quilty offers is indirect access to a set of language models that are solicited to predict the future regarding unproduced works of art. It would be truly remarkable if these AI tools functioned as Quilty claims. But most of them are just sophisticated pattern recognition machines, still far from understanding what humans find entertaining.