Creator makes feature-length film using LLM, calls it a failure

The creator, who runs a YouTube channel focused on AI-generated music and music videos, decided to test whether a large language model could independently produce a feature-length film. They chose to adapt William Hope Hodgson's science fiction novel The House on the Borderland, and used Claude Fable 5 to generate the work. The experiment succeeded in the literal sense: a feature-length movie was created and uploaded to the channel. However, the creator explicitly frames the outcome as a failure, despite being willing to attach their name to it. They clarify that this judgment is relative—the film meets the standards of small YouTube channels but falls short when compared to movies as a broader category. The creator notes that Claude Fable 5 demonstrated strong agency and understanding throughout the process; when given permission, the model knew what to do and executed well. This suggests that the reasons for the shortfall lie elsewhere than in the LLM's ability to plan or understand the task. The creator does not elaborate in the excerpt on what specifically caused the failure, but the implication is that while an LLM can be tasked with and execute a long creative project, the result does not meet the standard of professional filmmaking.

The creator's experiment sits at the intersection of capability and execution. They began with a straightforward technical question—can an LLM generate a feature-length film?—and found that the bottleneck was not the LLM's understanding or agency. Claude Fable 5 demonstrated clear purpose and competence, suggesting that the LLM itself was not the limiting factor. Instead, the creator identifies other reasons the result fell short, though the article body does not detail those reasons in full. This reframing is significant: it implies that the failure is not a failure of intelligence or planning by the model, but rather a failure in other dimensions—perhaps narrative coherence, visual direction, or some other aspect of filmmaking that an LLM struggles to sustain or coordinate across a full-length project. The creator's own assessment—that they would not release it under professional film standards, only YouTube standards—underscores that the gap is real but perhaps not in the dimension one might initially assume.

No comments yet. Be the first to share your thoughts!