Lionsgate’s ambitious push to use generative AI for filmmaking has hit a wall, forcing the studio to scale back its vision and rethink how to deploy the technology.

In 2025, Lionsgate partnered with AI startup Runway to build a model trained on the studio’s library of films and shows, touting it as a foundation for next‑generation, AI-assisted productions.¹ The goal was to generate footage robust enough for a traditional, feature-length movie.

By last fall, however, the project was reportedly running into serious “technical issues.” The AI system struggled to create any footage that could realistically be turned into a full-on movie, despite being trained on Lionsgate’s catalog.² According to reporting cited in coverage of the partnership, the underlying problem was scale: the “Lionsgate catalog isn’t enough to create those kinds of large-scale projects,” limiting what the model could learn and produce.²

As these problems mounted, Lionsgate began to change course. Instead of aiming first at AI-assisted feature films, the studio is now “putting more energy into producing ‘short episodic series using existing Lionsgate IP,’” using AI to generate shorter-form content that’s easier to control and refine.¹

From the studio’s perspective, this pivot keeps the AI initiative alive while reducing risk: short, IP-based episodes require less flawless footage and can be iterated quickly. From a technology standpoint, it reflects a broader industry lesson that even major studio libraries may be too small and too narrow to power fully automated, blockbuster-scale productions.

Observers now see Lionsgate’s experiment as an early test case for Hollywood’s AI ambitions: a high-profile partnership that reveals both the promise of AI-generated media and the constraints imposed by limited training data and current model capabilities.²