Creativity can't be automated

Maybe it is a skill issue. But for months I've been building AI tools for filmmakers: story development, screenplay analysis, image generation, storyboards. I've written careful prompts. I've used the most powerful models available. And I keep hitting the same wall.

AI can generate. It cannot choose.

Ask it for ten story concepts and you'll get ten competent options. Ask it which one is actually good and it'll confidently pick whichever you seem to prefer. It has no taste. It doesn't know what makes a story matter, only what makes a story statistically plausible.

The features that taught me this

Filmvision has a story development chat. You bring an idea, maybe a premise, a character, a scene stuck in your head. AI helps you build it into a structured story. It can identify which of 40+ story structures best fits your concept. Hero's Journey, Save the Cat, Kishotenketsu, whatever. It'll generate a full screenplay draft in minutes.

As an unblocker, it's remarkable. When you're staring at a blank page, having something to react to is invaluable. It gets you moving.

But here's what I've learned after months of using it: not a single time has the output felt right.

The structure? Usually solid. The beats land where they should. But the pacing is off. The dialogues are functional but flat. The screenplay reads like a screenplay. Technically correct, emotionally hollow. It passes every mechanical test and fails the only one that matters: does this feel like something worth making?

There's another feature that falls into the same bucket: screenplay analysis. Upload a script and AI extracts characters, identifies locations, breaks down scenes, even suggests shot compositions. For pre-production, it's genuinely useful. Hours of breakdown work compressed into minutes.

But the shot suggestions? They're competent. Safe. The AI knows that a close-up works for emotional beats and a wide shot establishes location. It doesn't know when to break that rule. When a distant wide shot of a character crying says more than any close-up could.

Great to get started. Not where you finish.

Why suggestions, not solutions

So I made a decision. These features would live in the "suggestion" landscape. AI proposes, humans dispose.

The tools offer structure options. They draft scenes. They identify patterns and extract data. But every output is framed as a starting point, not a finished product. A human always needs to review, tweak, rewrite. That's where the actual value lives.

Not in the generation. In the curation.

Taste is the thing you can't automate

Taste is the thing that says "cut this scene" when the scene works fine. It's knowing when to break the rules on purpose. It's the judgment that comes from years of consuming, making, failing, and refining. You can't prompt for it. You can't fine-tune it in.

Art collectors say the same thing. Not expertise. Intimacy. "How a work made me feel, what it asked of me." That comes from years of looking, being wrong, learning what actually moves you. There's no training data for that.

Here's the twist: this doesn't make AI less valuable. It makes taste more valuable.

AI is a taste amplifier. Give it to someone with sharp instincts and they'll produce more, faster, better. Give it to someone without taste and you get polished mediocrity at scale.

The bottleneck was never generation. It was always curation. AI just made that obvious.

The human in the loop isn't a bug

I keep seeing tools that promise to "fully automate" creative workflows. End-to-end screenplay generation. Automated storyboards. One-click films.

I don't believe in it. Not because AI isn't powerful. It is. But the value of creative work isn't in the artifact. It's in the choices that shaped it. The decision to linger here, cut there, say less, imply more.

Those choices require taste. And taste requires a human.

So no, creativity can't be automated. The part that decides what's worth making? That's still yours.