Develop a film idea with an AI writing partner

Give the conversation a direction

A request for “an original short film idea” leaves almost every creative decision open. You still have to decide what matters, what feels familiar, and what you can actually make. Give your AI writing partner a smaller problem to explore.

Our suggested approach is to bring a specific observation, ask for different actions that could grow from it, then develop the one that interests you. The model supplies alternatives. You decide which have a story worth pursuing.

This guide uses a 30-second film as a practical constraint. Six options and five beats are working limits to keep the exercise manageable, not a formula for originality.

1. Start with something you noticed

Choose an image, habit, or contradiction you have an opinion about. “Loneliness” is broad. “The person who cleans a playground never gets to play in it” gives you a person, an activity, and a tension to explore.

Add the feeling you want to leave behind and a few production limits. For our example: a moment of quiet delight, one night cleaner, an empty school corridor, five shots, no dialogue. These limits make it easier to compare ideas and plan a first film.

Keep one thing open: what the character does when their routine is interrupted. That choice is where we will look for the story.

2. Ask for different actions

Paste the prompt below into your AI chat app and fill in the brief. Each option should change what happens. A new costume or a different color palette does not give you a different story.

Help me find a story for a short film.

My brief:
- Starting observation or image: [something I noticed or care about]
- Intended feeling: [what the audience should feel at the end]
- Runtime: 30 seconds
- Cast and location: one character, one location
- Production limits: five shots, little or no dialogue, no readable text required
- Avoid: [familiar plots, devices, or imagery I do not want]

Give me six different story possibilities. Change the action and the character's choice in each, not just the setting or visual style.

For each, give me:
1. What the character is trying to do.
2. What interrupts that action.
3. The choice they make in response.
4. The final image that lets us see the consequence.

Keep each possibility under 70 words. Use things we can see or hear. Flag any idea that needs backstory, extra characters, or difficult continuous action. Do not score the ideas or write a screenplay yet.

For the cleaner, possible directions include secretly taking a turn on the hopscotch grid, repairing a missing square for tomorrow’s children, or treating the cleaning routine itself as a game. Each gives the character a different action and motive.

These are illustrative ideas written for this guide, not results from a model comparison. Asking for alternatives creates material to judge; it does not establish that an idea is new.

3. Choose the moment you want to watch

Read the options as scenes. Can you picture what the character does? Does their choice change anything? Can the final image carry the feeling without someone explaining it?

Cross out ideas that depend on hidden backstory or that you cannot yet stage. Of the remaining options, choose the moment you are most curious to see. You do not need a numerical ranking from the model.

We will choose the cleaner taking a turn at hopscotch. It expresses the tension through a physical action. “A cleaner discovers the magic of childhood” is still too vague to shoot. “A cleaner starts erasing a chalk game, tries one hop, then leaves the rest intact” gives us a visible change of mind.

4. Fix one weakness at a time

Tell the model what to preserve and exactly what is missing. For our cleaner: “Keep the hop. The ending needs to show that this moment changes the cleaner’s behavior. Give me three possible final actions.”

This keeps the revision focused. You can compare the endings without reconsidering the entire film. Use the same method if the opening is unclear or the character’s choice feels unearned.

Develop option [number].

Keep: [the action, choice, or image that interests me].
Change: [the specific part that feels predictable or unclear].
My reason: [what I want the viewer to understand or feel].

Give me three different ways to fix that one weakness. Keep the same character, location, and runtime. For each, state what changes on screen and what the viewer could infer. Do not add a new subplot.

Then stop so I can choose. After I choose, turn that version into five visual beats totaling about 30 seconds. For each beat, give the visible action, approximate duration, and the new information it gives the audience. Flag any continuity or generation challenge. Finish with a one-sentence premise. Do not write video-generation prompts yet.

If the suggestions keep repeating the same weakness, change the brief or supply a direction yourself. Another request to “make it more creative” gives the conversation little new information.

5. See whether it holds in five beats

Choose the revision, then sketch the action across the runtime. Our ending is simple: the cleaner steers the cart around the remaining chalk. Here is an illustrative beat sheet for that version.

1. 0–5s

   A night cleaner mops a school corridor, pushing toward a chalk hopscotch grid.

   The job and the obstacle.

2. 5–10s

   The mop wipes away the nearest square. The cleaner pauses at the next one.

   The routine is interrupted.

3. 10–16s

   Looking down, the cleaner lifts one foot over a surviving square.

   An impulse to play.

4. 16–23s

   A close shot catches a single hop into the next square. Hold on the landing.

   The cleaner acts on that impulse.

5. 23–30s

   The cleaner pushes the cart around the remaining grid, leaving it intact.

   Play has changed how the job gets done.

Read it aloud with a timer or rough it out with still frames. The timings are a starting point. Give the pause and the final choice enough time to register.

Plan for the hard parts: the chalk layout must match between shots, and the hop needs a convincing landing. A close shot of one hop keeps the movement brief. Reference images and a simple continuity sketch can help you plan the grid before generating clips.

The premise now fits in one sentence: A night cleaner about to erase a chalk hopscotch game tries one hop and decides to leave room for play.

Stop when you can make the next decision

You are ready to move into scripting and shot planning when you have a clear premise, a visible choice, an ending you care about, and beats that fit your limits. Save those together with the details you want to preserve.

There may be a better idea in the next batch. There is also a film you can start making now.