When Quilty appeared in industry publications earlier this year, the AI startup claimed its system could forecast a movie's performance simply by reviewing the script. Practical tests, however, left observers unconvinced. Despite access to broad datasets, the tool rated the script for Christy higher than the one for Sinners, even though the former underperformed commercially while the latter became a major award-winning success.

Quilty's leaders echo other AI executives in arguing that the platform can open up the sector by supplying emerging creators with helpful resources. A favorable rating might secure producer interest, whereas a lower one could prompt additional script work. At present, though, the service functions mainly as an uneven collection of existing AI platforms and has not yet shown it can reliably spot upcoming hits or established successes.

Established by producers Simon Horsman and Daniel Wood, Quilty applies AI to examine scripts and produce reports on likely outcomes. Once an unproduced script is submitted, the system assigns a score from 0 to 100 covering narrative strength, commercial prospects, audience appeal, and probable production expenses. The service markets itself as offering a preview of results during the greenlight process, with the founders expecting it to integrate into standard studio workflows.

In recent discussions, Horsman and Wood stressed the importance of retaining human oversight instead of automating pre-production entirely. During early development, they gathered input from other creatives who raised worries about generative AI displacing roles and reducing skills.

"We share many concerns about AI, yet our goal is to support human creativity," Horsman explained. "Quilty focuses on development and supplying writers, producers, buyers, financiers, and studio executives with details needed for sound decisions."

Rather than relying on one custom model, Quilty merges several widely used AI tools for varied analyses. Users upload scripts, and the platform returns reports including budget estimates, story outlines, and character breakdowns. Individual reports cost $50, with discounts available for multiple purchases.

Wood, who serves as CTO, first explored this approach during a legal matter when he tested consumer AI models after initial refusals from one system. He found varying performance across platforms and applied those lessons to Quilty's design.

"When a new model like Claude Mythos appears, the entire system improves," Wood noted, highlighting the benefit of swapping in stronger tools as they emerge.

This modular method allows quick adoption of better models, whether domestic or international. Gemini assists with structure, a US-hosted DeepSeek instance handles financial projections, and Claude combined with ChatGPT supports narrative and character reviews. Context prompting helps limit errors, and the company avoids training its own models.

Although the setup enables flexibility, it complicates full transparency on how scores are derived for subjective elements such as potential audience reactions. Traditional prediction relied on experienced professionals with deep audience knowledge.

No AI has replicated human judgment on art, yet the founders view their sentiment engine as a strong alternative. It incorporates tools like VADER to assess positive or negative tone in text.

Quilty cannot account for every external influence on reception. Horsman and Wood maintain the system can evaluate cultural relevance and project earnings. They cited Revenge of the Nerds as an older title that would score lower today due to outdated handling of sensitive topics.

Regarding the higher score given to Christy over Sinners, the founders attributed it to Sydney Sweeney's popularity and lower production costs for biographical dramas versus action films. They acknowledged limits in foreseeing events such as casting issues or unexpected cultural moments that affected other releases like Magazine Dreams and A Minecraft Movie.

Ultimately, Quilty provides indirect access to large language models asked to predict outcomes for undeveloped creative works. While impressive if accurate, these systems remain pattern-based tools far from fully grasping what audiences value.