Can AI Agents Make a Film Automatically?
A filmmaker tests every major AI filmmaking agent (Luma, Kling, LTX, Hedra) and finds they're all "basically unusable for professional context"—creating more tedious work than traditional workflows despite the hype.
Read Original Summary used for search
TLDR
• Tested Luma's AI agent with a full screenplay—it generated storyboards with seven-fingered hands, terrible continuity, jump cuts, and characters whose heads touch the ceiling
• Kling unprompted generated Elon Musk and randomly changed a wife character to a male character in an intimate scene
• User-facing AI tools (like Redfin's interior redesign) show more promise than professional filmmaking agents
• Anthropic report: 80% of arts/media jobs will be AI-affected, but we're only 25% through the adoption cycle
• NYT quiz reveals people prefer human writing only 60% of the time at sentence level, but this breaks down at macro storytelling scale
In Detail
This is a hands-on reality check of AI filmmaking agent hype. The creator tested multiple AI agents that promise to automate the entire filmmaking workflow—from script to storyboard to final video—and found them all severely lacking. Luma's agent, fed a complete screenplay, generated storyboards with absurd continuity errors: the same camera angle repeated (creating jump cuts), a character with seven fingers, lighting that "is honestly just terrible," and a man whose head nearly touches the ceiling. Kling's agent unprompted generated Elon Musk as a character and inexplicably changed the wife character to a male character. LTX created characters holding refrigerator interiors with one hand. The core argument: these tools create more frustrating, tedious work than traditional manual workflows.
The more interesting finding is the divergence between professional and consumer-facing AI tools. Redfin's interior redesign feature (likely using tools like Flux) shows genuine promise for set design work, and user-facing tools like Suno are significantly more popular than professional tools like Runway according to A16Z data. An Anthropic report reveals that over 80% of arts and media jobs will be dramatically changed by AI, but we're only 25% through the adoption cycle—meaning three-quarters of the disruption is still ahead.
The NYT created a quiz comparing AI-rewritten text to classic literature at the sentence level. People preferred human writing only 60% of the time, but this doesn't scale to macro-level storytelling. AI excels at making individual sentences clearer but fails at emotionally resonant narrative structure across multiple pages. The takeaway for filmmakers: ignore the agent hype, stick with proven manual techniques for character continuity and shot breakdown, and expect the real AI transformation to come from unexpected directions.