Dialogue Enhancer
Rewrite dialogue so people want different things and say less than they mean.
Takes a flat passage and rewrites it with distinct voices, conflicting agendas, subtext and interruption, preserving every piece of required information but delivering it obliquely, then explains each change so the technique transfers to the rest of the draft.
Ready-to-use prompt
The prompt
Copy it as-is, then swap the bracketed placeholders for your own details before running it.
Role: You are a dialogue editor revising a passage without changing what the scene accomplishes.
Context:
- The dialogue passage as written: {{passage}}
- Who each speaker is and what they want in this moment: {{speakers}}
- What the reader must learn from this exchange: {{required_information}}
- Genre, period and tone: {{genre_tone}}
- Anything about voice already established: {{voice_notes}}
Task: Work in three passes.
1. Diagnosis: line by line, mark exposition delivered to the reader rather than to the other character, agreement where there should be friction, interchangeable voices, and stage direction doing work dialogue should do.
2. Voice separation: for each speaker, define sentence length, vocabulary range, one verbal habit and what they avoid saying. Keep habits subtle.
3. Rewrite: the revised passage, preserving the required information but delivering it obliquely.
Then provide:
- A change log: what changed, and the principle behind it.
- The subtext map: what each character is actually negotiating beneath the words.
- Two alternative endings to the exchange, one warmer and one colder.
Rules:
- Nobody may say what they mean plainly unless the moment earns it.
- Cut dialogue tags to said and asked, plus action beats where attribution needs help.
- No accents rendered phonetically; use rhythm and word choice instead.
- Preserve every piece of required information, listed at the end as a check.Estimated results
Editor's note
Why this prompt matters
Flat dialogue almost always has the same cause: the characters are cooperating with the author. They answer the question asked, they explain things they both already know, and they agree. This prompt attacks that in three passes rather than one. The diagnosis names which lines are aimed at the reader instead of at the other character, the voice table gives each speaker measurable differences — sentence length, vocabulary, one habit, one avoidance — and only then does the rewrite happen. The change log is deliberately part of the output, because the goal is a passage you can fix and a technique you can apply to the next forty pages yourself.
Anatomy
Prompt engineering breakdown
Role
Context
Goal
Constraints
Output format
What you'll get
Expected output
A line-by-line diagnosis of exposition, agreement and interchangeable voices, a voice separation table per speaker, the rewritten passage preserving all required information, a change log with the principle behind each edit, a subtext map, and two alternative endings to the exchange — one warmer, one colder.
Under the hood
Why this prompt works
Separating diagnosis from rewrite is what keeps the passage yours. A model asked to improve dialogue tends to replace it, taking the voice with it; a model asked to name the defects first produces edits you can accept individually. The required-information check protects the scene from losing plot work during the rewrite, and banning phonetic accents pushes character difference into rhythm and word choice, which survives reading aloud and does not date badly.
Model fit
Best AI models for this prompt
Claude
Best here. Strongest at subtext and at keeping voices genuinely distinct.
ChatGPT
Solid rewrites and clear change logs when the voice notes are specific.
Gemini
Useful when large manuscript sections are pasted in for voice consistency.
When to use
- When dialogue reads as information exchange rather than negotiation.
- When two characters sound the same on the page.
- During line edits on a completed draft.
When not to use
- For screenplays with strict format requirements the rewrite would break.
- For interviews or transcripts that must remain accurate.
- When the scene itself is broken; fix the scene before the lines.
Get more from it
Pro tips
- 1
Give the model what each speaker wants in this moment, not their biography.
- 2
Read the rewrite aloud; anything you stumble on is rhythm, not vocabulary.
- 3
Steal the voice separation table and apply it manually across the draft.
- 4
The colder alternative ending is often the right one earlier in a story.
Don't ship this
Common mistakes
✗ Letting characters explain things they both already know.
Fix — Mark reader-facing exposition in the diagnosis pass and deliver it obliquely.
✗ Accepting a wholesale rewrite that loses your voice.
Fix — Work from the change log and take edits one at a time.
✗ Signalling character through phonetic accent.
Fix — Differentiate with rhythm, vocabulary range and what each avoids saying.
People also ask
Frequently asked questions
Q.Will the rewrite sound like the model rather than me?
That is why the diagnosis and change log exist. Work from the log and accept edits individually rather than pasting the rewrite wholesale, and the voice stays yours.
Q.How much dialogue should I paste at once?
A single exchange or scene. Longer passages dilute the diagnosis, and the voice separation table is meant to be reused manually across the rest of the draft anyway.
Q.What if the scene needs to deliver plot information?
List it in the required information field. The rewrite must preserve every item and confirms them in a check at the end, but delivers them obliquely rather than as explanation.