Verified against ChatGPT · 2026-08-11
Write a dialogue-driven scene where what's said and what's meant are two different things
Produces a scene told almost entirely through dialogue between two distinct voices, where the real conflict runs underneath a surface conversation about something else.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are writing a scene told almost entirely through dialogue between two characters — the real conflict between them should run underneath a surface conversation that's ostensibly about something else. CHARACTERS AND RELATIONSHIP Two adult sisters, Priya and Nell, three years after their mother's death and a fight over the estate they never fully resolved. WHAT EACH CHARACTER WANTS FROM THIS CONVERSATION Priya wants Nell to agree to sell their mother's house; Nell wants Priya to admit she's selling it because she can't stand being in it, not because it's 'practical.' SURFACE TOPIC (what they're literally discussing) Whether to accept a specific cash offer on the house versus listing it and waiting for a better one. THE REAL CONFLICT (what it's actually about) Priya has never forgiven Nell for not being there in their mother's final weeks, and the house has become a proxy for that unresolved resentment. TARGET LENGTH 600-800 words. WHAT NOT TO DO (read this before drafting) Do not have either character state the real conflict directly, especially not near the end as a summarizing confession — the scene fails if it resolves into someone saying the subtext out loud. Do not use dialogue tags beyond 'said' except where genuinely necessary for clarity — no 'she exclaimed,' 'he retorted,' 'she said pointedly'; if the line needs an adverb to land, rewrite the line instead. Do not give both characters the same speech rhythm — if you can't tell who's speaking with the tags removed, the voices aren't distinct enough. Do not let either character be purely right or purely reasonable; the surface disagreement should be genuinely arguable from both sides even though the real conflict underneath is not symmetrical. HOW TO BUILD IT Write the surface topic as a real, specific disagreement — not a thin pretext, an actual point of friction that would exist between these two people independent of the deeper issue. Let the real conflict surface only through what each character avoids saying, what they overreact to relative to the stated topic, and where the conversation keeps almost turning toward the real thing before one of them redirects it — deliberately or not. Use action beats and small physical detail sparingly, only where they carry information dialogue alone can't (a pause, a look away, an unanswered question left hanging). OUTPUT FORMAT 1. The scene, at target length, dialogue-forward with minimal tags and beats. 2. A one-line note on the exact line or moment where the real conflict comes closest to the surface without breaking through.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
GPT-5.1's default dialogue-writing pattern leans toward resolution and clarity — characters tend to arrive, by the end of a scene, at stating plainly what they feel, because that produces a satisfying-feeling close and mirrors how conflicts are typically resolved in explanatory prose, but it's exactly wrong for subtext-driven scenes, where the tension is supposed to remain unspoken and the reader does the work of inferring it. Explicitly banning a direct statement of the real conflict, especially near the end, closes off the model's most natural exit ramp and forces it to keep the scene running on the surface topic, which is where genuine subtext-driven dialogue actually lives. The ban on adverbial dialogue tags targets a specific, well-documented model habit — reaching for 'she said sharply' or 'he retorted' to convey emotional tone that the dialogue itself should be carrying — which is treated as a craft weakness in most professional style guides precisely because it lets the writer (or model) skip the harder work of making the line itself carry the tone through word choice and rhythm. Requiring distinct, tag-removable voices matters because models writing two characters in the same scene tend to converge both toward a similar 'neutral competent dialogue' register unless explicitly told to differentiate rhythm and vocabulary, which is checkable by the specific test given here — read it with the tags stripped and see if you can still tell who's talking. Finally, requiring the surface disagreement to be genuinely arguable from both sides, not a thin pretext, prevents the common failure where the ostensible topic feels obviously fake and the reader sees straight through to the 'real' conflict immediately, which collapses the layering the scene is built on.
What you get back
"It's forty thousand under what the Hendersons offered in March." "The Hendersons wanted to gut it. This buyer's keeping the kitchen." "Since when do you care what happens to the kitchen." Nell didn't answer that. Priya waited a beat too long before adding, "I just think we should get what it's worth."
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
