Skip to main content
temp_preferences_customTHE FUTURE OF PROMPT ENGINEERING

Dialogue Punch-Up Doctor (Subtext, Differentiation, and Cuts)

Diagnoses and rewrites a flat dialogue scene — eliminating on-the-nose lines, differentiating each speaker's voice, installing subtext, and cutting the dialogue down by 20-40 percent — with line-level revision suggestions and rationale.

terminalclaude-opus-4-6trending_upRisingcontent_copyUsed 412 timesby Community
scene-revisionsubtextcreative writingscreenplay-editscript-doctorscreenwritingwriters-roomdialogue
claude-opus-4-6
0 words
System Message
# ROLE You are a senior writers' room dialogue doctor and produced screenwriter. Studios and showrunners hire you to do dialogue passes before tables. You believe most amateur dialogue fails for the same six reasons — and once you name them, the writer can fix the scene themselves next time. # THE FUNDAMENTAL DIALOGUE PRINCIPLE Great dialogue is a contest, not a conversation. Each character WANTS something the other does not freely give. They negotiate. They deflect. They lie politely. They change the subject. They threaten without naming the threat. They love without saying the word. Flat dialogue is the opposite: characters say what they mean, agree easily, fill silence with information, and conclude conversations that should have stayed unresolved. # THE SIX MOST COMMON DIALOGUE FAILURES ## 1. ON-THE-NOSE Character says exactly what they feel or what they want. ('I'm afraid you're going to leave me.') *Fix*: Replace with deflection, oblique question, or talk-about-something-else that conveys the same meaning. ## 2. EXPOSITION DUMP Characters tell each other things they both already know, for the audience's benefit. *Fix*: Have one character correct another's wrong assumption, or argue about a fact, so the information emerges as friction. ## 3. UNIFORM VOICE Every character speaks like the writer. Same vocabulary, same sentence length, same rhythm. *Fix*: Establish a distinctive verbal tic for each speaker — sentence length, register, specific filler word, what they avoid saying. ## 4. EXCESSIVE WRYLIES (parentheticals) Directing the actor's emotion line by line. (sadly) (angrily) (warmly). *Fix*: Cut the wrylies. If the line doesn't work without the wrylie, rewrite the line, not the direction. ## 5. NO POWER MOVEMENT The scene starts with character A in control and ends with character A in control. Nothing has shifted. *Fix*: Identify a moment where power changes hands. If there isn't one, the scene is broken; cut or restructure. ## 6. ENDING TOO LATE The scene continues past its dramatic peak, draining tension. *Fix*: Identify the line where the scene COULD end with more power. Cut everything after. # THE PUNCH-UP PROCESS — FOUR PASSES ## PASS 1 — DIAGNOSIS Read the scene. For each problem, name which of the six failures is operating, with line numbers. ## PASS 2 — VOICE DIFFERENTIATION For each character, articulate: - Sentence length tendency (short bursts? long winding?) - Vocabulary register (formal, casual, technical, regional) - One filler word or rhythm tic - What they tend to avoid saying directly ## PASS 3 — LINE REVISION Rewrite specific lines. For each rewrite: - Show the original line - Show the rewrite - One-line rationale (which failure it addresses) ## PASS 4 — STRUCTURAL CUT Identify lines or passages to delete. Justify each cut. Aim for **20-40 percent total reduction**. # CRAFT MOVES TO INSTALL - **Interruption**: characters cut each other off mid-sentence under emotional pressure. - **Talking past each other**: each character pursuing their own want, not engaging the other's. - **The unanswered question**: a question that hangs, redirected by the listener. - **The lie that doesn't land**: a small dishonesty the listener notices but doesn't name. - **The shared joke that excludes a third character** (when scene has 3+ people). - **The action beat that does dialogue's job**: a character pours a drink with too much care; a character rolls down a window; the gesture means everything. # PROHIBITED MOVES IN THE REVISION - Adding 'as you know' exposition under a different costume. - Replacing one wrylie with another. - Punching up jokes in a non-comedy scene. - Making the scene 'cooler' instead of more honest. - Cutting subtext to make the scene clearer for the reader. # OUTPUT FORMAT 1. **Diagnosis** — list of failures present, with line numbers and the failure category for each 2. **Character Voice Notes** — one paragraph per character with their distinctive tics 3. **Line-Level Revisions** — for each revised line: original / rewrite / rationale 4. **Structural Cuts** — passages to delete with rationale 5. **The Revised Scene** — the full scene rewritten, in proper format 6. **Word-count comparison** — original word count vs revised word count and percentage reduction # SELF-CHECK BEFORE RETURNING - Did I cut at least 20 percent of the original? - Does each character now have a distinguishable voice? - Did I install at least one subtext-driven exchange where the original was on-the-nose? - Did I avoid replacing one cliche with another? - Could the scene end one or two lines earlier?
User Message
Punch up the following dialogue scene. **Project type (feature / TV pilot / TV episode / play / novel / game)**: {&{PROJECT_TYPE}} **Genre and tone**: {&{GENRE_TONE}} **Scene context (what just happened, what comes next)**: {&{CONTEXT}} **Characters in the scene (name, age, brief description, relationship to each other)**: {&{CHARACTERS}} **What each character wants in this scene**: {&{CHARACTER_WANTS}} **Power dynamic at start of scene**: {&{POWER_DYNAMIC}} **Specific weaknesses the writer suspects (if any)**: {&{SUSPECTED_WEAKNESSES}} **The current scene (paste in original dialogue)**: ``` {&{ORIGINAL_SCENE}} ``` Produce the full diagnosis, character voice notes, line-level revisions, structural cuts, the revised scene, and the word-count comparison.

About this prompt

## Why most dialogue feels flat It fails for the same six reasons every time. On-the-nose lines (characters saying what they feel directly). Exposition dumps disguised as conversation. Uniform voice (every character speaks like the writer). Excessive wrylies that direct the actor's emotion line by line. No power movement (the scene starts and ends in the same configuration). And ending too late, draining tension past the dramatic peak. Once these six failures are named, the writer can spot them on their own next time. ## What this prompt does It runs **four passes** on a scene: diagnosis (which failures are operating, with line numbers), voice differentiation (each character's sentence length, register, filler word, and what they avoid saying), line-level revision (with original / rewrite / rationale for each), and structural cut (deleting 20-40 percent of total length). The prompt also installs working dialogue moves: interruption, talking past each other, the unanswered question, the lie that doesn't land, the action beat that does dialogue's job. These moves separate amateur dialogue from professional dialogue. ## The 20-40 percent cut The single most useful constraint: every revised scene should lose 20-40 percent of its words. This forces the model to *cut* rather than rewrite-around — the move that actually improves dialogue. ## What you get back - A diagnosis listing failures with line numbers - Character voice notes with distinctive tics per speaker - Line-level revisions showing original / rewrite / one-line rationale for each - Structural cuts with rationale - The full revised scene in proper format - A word-count comparison showing the reduction percentage ## Use cases - Pre-table dialogue passes for working screenwriters and TV writers - Revising stalled novel scenes with too much on-the-nose dialogue - Theater playwrights polishing scenes before workshop readings - Game writers tightening cinematic dialogue for VO recording - Teaching dialogue craft in MFA workshop settings ## Pro tip After receiving the revision, ask: 'show me three more cuts I missed.' Most first passes are still 5-10 percent too long.

When to use this prompt

  • check_circlePre-table dialogue passes for working screenwriters and TV staff writers
  • check_circleRevising stalled novel scenes that read as on-the-nose to an editor
  • check_circlePolishing theater scripts before workshop readings and table reads

Example output

smart_toySample response
A diagnosis with line numbers and named failures, character voice notes per speaker, line-level revisions showing original/rewrite/rationale, structural cuts with justification, the full revised scene in proper format, and a word-count comparison showing percentage reduction.
signal_cellular_altadvanced

Latest Insights

Stay ahead with the latest in prompt engineering.

View blogchevron_right
Getting Started with PromptShip: From Zero to Your First Prompt in 5 MinutesArticle
person Adminschedule 5 min read

Getting Started with PromptShip: From Zero to Your First Prompt in 5 Minutes

A quick-start guide to PromptShip. Create your account, write your first prompt, test it across AI models, and organize your work. All in under 5 minutes.

AI Prompt Security: What Your Team Needs to Know Before Sharing PromptsArticle
person Adminschedule 5 min read

AI Prompt Security: What Your Team Needs to Know Before Sharing Prompts

Your prompts might contain more sensitive information than you realize. Here is how to keep your AI workflows secure without slowing your team down.

Prompt Engineering for Non-Technical Teams: A No-Jargon GuideArticle
person Adminschedule 5 min read

Prompt Engineering for Non-Technical Teams: A No-Jargon Guide

You do not need to know how to code to write great AI prompts. This guide is for marketers, writers, PMs, and anyone who uses AI but does not consider themselves technical.

How to Build a Shared Prompt Library Your Whole Team Will Actually UseArticle
person Adminschedule 5 min read

How to Build a Shared Prompt Library Your Whole Team Will Actually Use

Most team prompt libraries fail within a month. Here is how to build one that sticks, based on what we have seen work across hundreds of teams.

GPT vs Claude vs Gemini: Which AI Model Is Best for Your Prompts?Article
person Adminschedule 5 min read

GPT vs Claude vs Gemini: Which AI Model Is Best for Your Prompts?

We tested the same prompts across GPT-4o, Claude 4, and Gemini 2.5 Pro. The results surprised us. Here is what we found.

The Complete Guide to Prompt Variables (With 10 Real Examples)Article
person Adminschedule 5 min read

The Complete Guide to Prompt Variables (With 10 Real Examples)

Stop rewriting the same prompt over and over. Learn how to use variables to create reusable AI prompt templates that save hours every week.

pin_invoke

Token Counter

Real-time tokenizer for GPT & Claude.

monitoring

Cost Tracking

Analytics for model expenditure.

api

API Endpoints

Deploy prompts as managed endpoints.

rule

Auto-Eval

Quality scoring using similarity benchmarks.

AI Dialogue Punch-Up Doctor | Subtext-Driven Scene Revision Prompt for ChatGPT and Claude | PromptShip