Develop a film idea · Part 1 of 6
Loglines and What If’s
Get from a hunch to a premise you can actually develop.
Do not ask the model for “movie ideas.” Give it your hunch, force it into a testable sentence, then change one variable at a time until only one version survives.
What is a logline?
A logline is your whole film in one sentence: who it is about, what they want, what stands in the way, and what it costs if they fail. It is not a tagline, and it is not a plot summary. It is the quickest test of whether you have a story yet. If you cannot say it in one sentence, you are still circling the idea.
Star Wars: A New Hope
After a droid carrying stolen battle station plans lands on his desert planet, a restless farm boy must get past the Empire’s fleet to deliver them to the Rebellion, or a weapon that can destroy whole planets will end the fight for good.
The Godfather
After rival gangsters shoot his father, a war hero who swore off the family business must outmaneuver New York’s crime families to protect the people he loves, or lose them one by one to a war he never wanted.
What to start with
- The raw hunch in your own words (messy is fine)
- Three comps: not “it’s Inception meets Knives Out,” but what you are stealing from each (the con structure, the family pressure, the last-ten-minutes turn)
- The feeling the ending must leave (dread, release, shame, awe)
- Hard limits: runtime, locations you can shoot or generate, budget, no magic, period, language, rating
If you skip limits, you get a generic studio logline. Treat the model as a development exec with a job, constraints, and a required output shape — not as a brainstorm intern. Nainsi Dwivedi
Force a formula, then break it
Make the first pass mechanical:
After CATALYST,What happens HEROWho must overcome OBSTACLEWhat blocks them to achieve GOALWhat they want or else STAKES.What it costs
That sentence is a diagnostic. If you cannot fill a slot, the idea is still a vibe. Academic Save the Cat + AI experiments start from this same skeleton (logline + setting + comps + protagonist + antagonist) before any beat sheet. Kenyon study
Then break the formula on purpose. The interesting version is usually the one where the goal is ugly, the hero is the wrong person, or the stakes are personal instead of apocalyptic.
Fork one variable at a time
Do not generate ten “totally different movies.” Generate ten mutations of this movie. Change only one thing per row:
- Protagonist (who has the problem)
- Antagonist (who benefits if they fail)
- Arena (where this is expensive or embarrassing)
- Tone (same events, different genre contract)
- Who holds the secret
- Ending irony (they get the goal and lose the need, or the reverse)
This is the same move people use when they ask Claude to turn one idea into multiple angles or run “what-if” scenarios — applied to story, not marketing. Tech_Rose
Worked fork (keep the engine, swap one part)
Base: After a pop star’s bodyguard witnesses a murder, he must keep her alive long enough to testify, or the syndicate erases them both.
- Protagonist swap: the pop star is the one who witnessed it; the bodyguard is the liability.
- Arena swap: same plot, all inside a 48-hour press tour. No hideout, no safe house.
- Irony swap: they survive by becoming more famous, which is the thing that will kill them next.
You should be able to point at the cell you changed. If you cannot, you started a new pile of ideas and threw away the original.
Kill test
For each surviving logline, make the model argue against it:
- What existing movie is this already?
- Where does Act 2 become travel + conversations?
- What would a buyer say in the first thirty seconds?
- If we remove the twist, is there still a movie?
Keep the one you still want after the attack. If you cannot defend it out loud in one breath, it is not locked.
Using AI as a writing partner
Start a conversation with an AI agent, give it your hunch and this brief, and develop the logline together one fork at a time.
Paste into your AI chat