Verified against ChatGPT · 2026-08-11
Turn a loose feature idea into a working three-act beat sheet with turning points named and justified
Converts a premise, protagonist, and ending into a structured beat sheet — major turning points, midpoint, and act breaks laid out as a table with a one-line justification for why each beat happens where it does.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Build a three-act beat sheet for a feature-length screenplay from the premise below. I don't want prose or scene-by-scene detail yet — I want the structural skeleton: the beats that have to exist for the story to function as a feature, with a reason attached to each one. LOGLINE A disgraced weather forecaster gets one more shot at credibility when he's the only person who believes an approaching storm is far worse than official models predict. PROTAGONIST AND WANT Mara, 42, wants her old network job back and, more specifically, wants to be right in public one more time after being wrong in public cost her everything. ENDING STATE She's right, the storm hits as she predicted, but she doesn't get the job back — she gets something she didn't know she wanted more: her estranged daughter's trust. GENRE AND RUNTIME TARGET Disaster drama, roughly 105 minutes. RULES Produce the following beats at minimum, each tied to an approximate page number based on a standard one-page-per-minute runtime: Opening Image, Inciting Incident, Break into Act Two (the protagonist commits to pursuing the want, past the point of easy retreat), Midpoint (a false victory or false defeat that raises the stakes rather than just marking the halfway point), All Is Lost / Low Point, Break into Act Three, Climax, and Final Image. For every beat, write one sentence describing what happens and a second sentence justifying why it has to occur at that point in the structure rather than earlier or later — a beat sheet that just lists events without structural reasoning is a plot summary, not a beat sheet. The Midpoint must genuinely raise the stakes or change the nature of the protagonist's problem, not simply be the scene that happens to fall at the halfway page count — if the premise as given doesn't naturally produce a stakes-raising midpoint, say so explicitly and propose one rather than forcing a weak beat into that slot. The Final Image should meaningfully invert, mirror, or answer the Opening Image so the two beats read as a pair when placed side by side. WHAT NOT TO DO Do not write dialogue or scene description — every beat stays at the level of "what happens and why," not "how it's shot or said." Do not pad the sheet with minor character beats that aren't load-bearing for the three-act structure; if a beat doesn't change the protagonist's situation or the audience's understanding of the stakes, it doesn't belong in a structural outline. OUTPUT FORMAT A table with columns: Beat Name | Approx. Page | What Happens | Why It Belongs Here. Follow the table with one paragraph flagging any beat where the given premise created genuine structural tension (for example, an ending state that doesn't obviously follow from the stated want) and how you resolved it.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A beat sheet is only useful if each beat is load-bearing and placed for a structural reason, and the single most common failure when a model generates one unprompted is to produce a plot summary chopped into named sections — an Opening Image and a Midpoint label slapped onto whatever events happen to fall near those page counts, without the events themselves doing the structural job those labels imply. Requiring a justification sentence alongside every beat forces GPT-5.1 to check its own placement rather than just naming beats in sequence, which surfaces the difference between a Midpoint that merely occurs at the halfway page and one that actually raises stakes — a distinction the model will not reliably self-enforce unless asked to argue for it. The explicit instruction to flag when the given premise doesn't naturally produce a strong midpoint or a clean final-image inversion matters because loglines supplied by a human are frequently underspecified for structure, and a model that silently forces a weak beat into a slot to satisfy the template produces an outline that looks complete but would fall apart the moment someone tried to write scenes from it — surfacing the tension instead lets the actual writer make the creative call rather than inheriting an invisible compromise. Pinning approximate page numbers to a one-page-per-minute convention gives the runtime target real teeth instead of leaving "105 minutes" as a number that never touches the actual beat placement, which is what happens when a runtime is stated but never operationalized against where events land. Banning dialogue and scene-level detail keeps the deliverable at the correct altitude — a beat sheet mixed with prose scenes is neither a clean structural document nor a usable draft, and separating the two lets each later stage of development work from the right level of abstraction.
What you get back
Midpoint (approx. p.55): Mara's storm model is confirmed correct by a rival network hours before her old bosses would admit it — but the confirmation comes from her daughter's amateur weather account going viral, which means Mara is now competing with the one person she was trying to reconnect with. Why here: this converts her external professional goal into an internal relational cost right at the point the story needs to complicate rather than simply confirm her original want.
Verified against
ChatGPT GPT-5.1 · 2026-08-11
Changelog
- 2026-08-11 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
