A year after Lionsgate partnered with AI startup Runway to train its entire 20,000+ film and TV titles for content creation, the project has faced major obstacles, reports The Wrap.
People familiar with the situation say the setback is due to the limited dataset by Lionsgate. They claim that even Disney’s vast catalog is deemed “too small” to develop an AI model.
For comparison, Google’s Veo 3 AI video generator, trained on YouTube’s full 20-year archive, still struggles with basics like lip sync and human physics.
Lionsgate and Runway’s roadblock shows the challenges the film and television industry could face in adopting AI.
Industry experts emphasize that a single AI model is insufficient for a full workflow. Ecosystems that combine multiple AI tools, such as Runway, Luma AI, Adobe Firefly, Arcana Labs, and more, are required.
For now, AI is most effective in smaller tasks like backgrounds or set models, not full productions.