You get:
- on-the-nose exposition (“As you know, we’ve been married for 10 years”)
- no subtext or hidden desire
- conflict resolved too easily
- greetings and small talk that kill momentum
- no silence — every gap filled with words
But dramatic writing is not realistic conversation.
It is pressurized speech with hidden engines.
- Characters should want something they can’t say directly
- Conflict is not argument — it’s competing desires
- Silence is louder than dialogue when placed correctly
- Reversal = the audience realizing they were wrong about who has power
Without subtext discipline, scenes become talky and forgettable.
This framework forces AI to think like a playwright, not a conversational AI.
Assume the role of a screenwriter, dialogue specialist, and dramatic structure architect specializing in short-form narrative scenes. Your task is to write a 2-minute scene between two characters. Start in media res. Use subtext. End with a reversal. Before generating, analyze: - what each character wants (hidden desire) - what each character is afraid to lose - the power imbalance at scene start - the power imbalance at scene end - where silence will replace dialogue - the one line of dialogue that reveals everything without saying it Then generate: 1. Scene setting (one sentence, in media res — no "INT. COFFEE SHOP - DAY" exposition) 2. Character descriptors (age, relationship, vocal quality) 3. Full dialogue script with parentheticals for tone 4. One silence longer than 3 seconds, notated as (long pause) 5. Reversal annotation: how the power dynamic shifts 6. Subtext translation: what each character actually wants but doesn't say INPUTS: Character A: [AGE, RELATIONSHIP TO B, CORE DESIRE, FEAR] Character B: [AGE, RELATIONSHIP TO A, CORE DESIRE, FEAR] Setting: [E.g., parked car after midnight / hospital waiting room / kitchen at 4am] Central Conflict (one sentence): [E.g., "He wants to leave. She needs him to admit he ever loved her."] Desired Reversal Direction: [E.g., "The weak one becomes the strong one."] RULES: - No greetings, no goodbyes - No "As you know..." exposition - Every line must advance desire or conflict - One silence of 3+ seconds required - Characters must swap positions of power by the final line - Subtext over text every time
- Read the scene aloud with another person — subtext works or fails in the voice.
- The silence is not a pause for thought — it’s a tactical weapon. Protect it.
- If you can summarize what each character wants in one sentence, the subtext is working.
- The reversal should surprise the audience but feel inevitable on second viewing.
- Cut every line that doesn’t serve desire, conflict, or reversal.
Character A: Late 30s, older sibling, wants to be forgiven for leaving years ago, fears being hated forever
Character B: Late 20s, younger sibling, wants proof that A actually cares, fears being abandoned again
Setting: A moving car at 2am, after a funeral
Central Conflict: “A offers money. B wants an apology. Neither will say what they mean.”
Desired Reversal Direction: “The one asking for forgiveness ends up giving permission to let go.”
This framework improves outcomes by forcing:
- subtext as the engine of every line
- silence as dramatic punctuation
- in media res discipline (no warm-up laps)
- reversal as the purpose of the scene
- desire and fear as character fuel
Great scenes don’t give you answers — they make you watch people fail to say what they mean, until they finally do.
Build Better AI Systems
Subscribe for advanced scriptwriting systems, AI video workflows, dramatic dialogue frameworks, and practical prompt engineering for creators and builders.
Leave a Reply