AI Startup Claims It Can Predict Script Success for Blockbuster Films
When Quilty made its debut in the industry earlier this year, the AI startup claimed its tool could accurately forecast a film’s success simply by analyzing the script. However, those who tried out Quilty’s product were left questioning its reliability. Despite having access to extensive data, it predicted that the script for Christy, which ultimately flopped at the box office, would outperform Sinners, which went on to win an Oscar.
Quilty’s founders, like many in the AI space, believe their technology can “democratize” the industry by providing emerging creatives with valuable assistive tools. A high Quilty score could potentially open doors with producers, while a low score might indicate the need for further revisions. Yet, at this stage, Quilty appears to be a disorganized blend of existing AI systems, lacking the ability to demonstrate that it can accurately identify future hits.
Founded by film producers Simon Horsman and Daniel Wood, Quilty employs AI to analyze scripts and generate comprehensive reports on a project’s likelihood of success. After inputting an unproduced script, Quilty assigns a score from 0 to 100, reflecting the narrative quality, commercial potential, audience appeal, and estimated production costs. The platform aims to provide users with insights as they seek to get their films greenlit, with Horsman and Wood envisioning Quilty as a vital tool for traditional production studios.
In a recent conversation, Horsman and Wood emphasized their commitment to keeping “humans in the loop” rather than fully automating the pre-production process. During the company’s early days, they sought feedback from various creatives who expressed concerns about AI’s potential to disrupt jobs and diminish human skills.
“We share many of the concerns about AI, but our goal is to enhance human creativity,” Horsman explained. “Quilty is about development and equipping users—whether they are writers, producers, buyers, financiers, or studio executives—with the information they need to make informed greenlight decisions.”
Instead of providing access to a single, custom AI model for script feedback, Quilty integrates multiple widely available AI tools to offer diverse analyses. Users simply upload their scripts, and within minutes, they receive a report detailing estimated budgets, key story beats, and character analyses. Each individual analysis costs $50, with discounts available for bulk purchases.
The concept for this analytical approach originated with Wood, Quilty’s CTO, during a legal dispute over real estate. Rather than hiring a lawyer, he turned to ChatGPT, which promptly advised him to seek professional help. He then tried Gemini, which worked better for a time due to its larger context window. His exploration of various AI models informed Quilty’s strategy for assessing a script’s potential success. For instance, Quilty uses Gemini for structural breakdowns and DeepSeek for financial modeling, while narrative and character analysis is handled by a combination of Claude and ChatGPT.
Wood explained that Quilty employs context prompting to enhance output quality and minimize inaccuracies. The company does not train the models it uses, but Wood views this as a strength, allowing for easier integration of new and improved models as they become available.
“When Claude Mythos is released and proves to be a superior LLM, my entire software improves,” Wood noted, referencing a powerful new model currently limited to select organizations for cybersecurity. “If a new model from another region outperforms US models, why wouldn’t I switch to it?”
While Quilty’s modular tech stack offers agility for updates, it also complicates understanding how the platform generates various metrics that aim to gauge audience reactions to scripts for films that haven’t been produced yet. Predicting audience response has been a part of film development since Hollywood’s inception, traditionally relying on human insight.
No AI has yet managed to replicate the nuanced thought processes humans use to form opinions about art. However, Quilty’s founders believe their “sentiment engine” is a close alternative, utilizing tools like VADER, which assesses the positivity or negativity of text.
Horsman and Wood are confident that Quilty can accurately evaluate how a project resonates with current cultural trends and provide reliable box office forecasts. They cited Revenge of the Nerds as an example of a classic film that would receive a lower Quilty score due to its problematic portrayal of sexual assault, which modern audiences would likely find offensive.
When I inquired about Quilty’s higher score for Christy, which grossed around $2 million, compared to Sinners, which earned $370 million, they attributed it to Sydney Sweeney’s popularity. They argued that Sweeney’s star power, combined with the lower production costs of biographical dramas, made Christy seem like a safer bet. However, this situation underscores the unreliability of Quilty’s logic, as the founders acknowledged that there are factors influencing a film’s financial success that the platform cannot predict.
For instance, Quilty could not have foreseen that Elijah Bynum’s Magazine Dreams, which Horsman produced, would be affected by actor Jonathan Majors’ public downfall in 2023. Similarly, the script for A Minecraft Movie did not indicate that the Chicken Jockey phenomenon would contribute to its success. Horsman and Wood expressed a desire for Quilty to eventually anticipate such factors, but it’s challenging to envision how that could be achieved.
Despite its ambitious claims, Quilty essentially offers indirect access to a collection of large language models tasked with predicting the future of unproduced artistic works. It would be remarkable if these AI tools functioned as Quilty suggests. However, most are advanced pattern recognition systems that still have a long way to go before they can truly grasp what captivates human audiences.
