Editor Mode: Polishing the Prose
“The first draft is just the clay. Now we sculpt.”
Editor Mode is the final phase of the Writers Factory workflow. While Director Mode prioritizes generation and structure, Editor Mode prioritizes refinement, consistency, and voice alignment.
This is where your draft becomes a manuscript.
The Problem with “Just Edit”
Traditional AI editing assistance works like this:
- You: “Make this scene better.”
- AI: Rewrites everything, loses your voice
- You: “No, keep my style but fix the pacing…”
- AI: Rewrites again, still generic
- Repeat until you give up and edit manually.
The fundamental problem? The AI doesn’t know what “better” means for your novel. It has no reference for your voice, your character’s psychology, or your world’s rules.
The Editor Mode Solution
Editor Mode uses your Voice Bundle as the gold standard. Every edit is measured against:
- Your calibrated voice (POV, tense, vocabulary, rhythm)
- Your anti-patterns (constructions you specifically want to avoid)
- Your Story Bible (character psychology, world rules)
The result: edits that enhance your prose without losing what makes it yours.
The Editing Pipeline
Editor Mode runs a structured pipeline, not random rewrites.
Pass 1: Voice Alignment Check
Before changing anything, the system scans for voice drift.
What It Checks:
| Element | Gold Standard | Detection |
|---|---|---|
| POV Consistency | Your calibrated POV (1st, 3rd limited, etc.) | Head-hopping, unauthorized omniscience |
| Tense Consistency | Your calibrated tense | Unexpected shifts to past/present |
| Vocabulary | Your approved word choices | Generic AI vocabulary intrusion |
| Rhythm | Your sentence length patterns | Mechanical uniformity |
Output: A Voice Alignment Report showing:
- Overall alignment score (0-100)
- Specific passages flagged for drift
- Suggested corrections
Pass 2: Anti-Pattern Scrub
Your Voice Bundle includes anti-patterns—constructions you never want in your prose.
Default Anti-Patterns (All Writers):
| Pattern | Why It’s Banned |
|---|---|
| “It’s important to note that…” | Lecture voice, breaks immersion |
| “Shivers down spine” | AI cliché, physiological lazy writing |
| “Couldn’t help but…” | Weak action, removes agency |
| “With practiced precision” | Empty modifier, tells instead of shows |
| “A mix of X and Y” | Vague emotional description |
POV-Specific Anti-Patterns:
| POV | Anti-Pattern | Why |
|---|---|---|
| First Person | “I couldn’t see his face, but I knew he was angry” | Can’t know what you can’t see |
| Third Limited | “Meanwhile, across town…” | Breaks limited perspective |
| Third Limited | “She didn’t know that he was planning…” | Unauthorized knowledge |
| Omniscient | None specific | Most flexible POV |
Output: Anti-pattern violations highlighted with one-click replacements.
Pass 3: Consistency Check
The Knowledge Graph powers long-range consistency validation.
What It Tracks:
| Continuity Type | Example Check |
|---|---|
| Character State | “Marcus wore the blue jacket” → Later: “He pulled off his red coat” |
| Timeline | “Three days later” → Does the calendar math add up? |
| Object Tracking | Chekhov’s gun: Was it fired? |
| World Rules | Magic costs fatigue → Did the mage feel tired after the spell? |
| Relationship Status | Are they enemies or allies in this chapter? |
Output: Consistency violations flagged with Knowledge Graph evidence.
Pass 4: Character Voice Validation
Each character should sound distinct. Editor Mode validates dialogue and internal monologue against character profiles.
Checks:
- Does Marcus’s dialogue match his established vocabulary?
- Does his internal monologue reflect his Fatal Flaw?
- Is his voice distinct from other characters?
Common Issues:
| Issue | Detection |
|---|---|
| All characters sound the same | Vocabulary overlap analysis |
| Character breaks psychology | Behavior doesn’t match flaw |
| Dialogue too on-the-nose | Subtext analysis |
Pass 5: Prose Polish
The final refinement pass focuses on craft:
| Polish Type | What Happens |
|---|---|
| Pacing | Identify slow passages, suggest cuts or compression |
| Sensory Detail | Flag abstract descriptions, suggest concrete alternatives |
| Metaphor Health | Check for oversaturation in any domain |
| Sentence Variety | Detect mechanical patterns, vary rhythm |
| Word Economy | Flag redundant phrases, suggest tighter alternatives |
Working with the Foreman in EDITOR Mode
The Foreman shifts behavior in Editor Mode. Instead of generating, it analyzes and refines.
Foreman Capabilities
| Capability | How to Use It |
|---|---|
| Run Full Pipeline | “Edit this scene” → Runs all 5 passes |
| Single Pass | “Check voice alignment only” |
| Surgical Fix | “Tighten the dialogue in lines 45-80” |
| Compare Versions | “Show me what changed” |
| Explain Flagged Issue | “Why did you flag line 23?” |
| Preserve Original | “Keep my metaphor, just fix the grammar” |
Example Conversation
You: Edit Chapter 3, Scene 2. Focus on pacing.
Foreman: Running Editor Pipeline on Chapter 3, Scene 2...
Voice Alignment: 94/100 (2 minor drifts)
Anti-Patterns: 1 violation found
Consistency: Clean
Character Voice: Clean
Pacing Analysis: 3 passages flagged as slow
Recommendations:
1. Lines 45-52: Internal monologue drags. Compress to 3 sentences.
2. Lines 78-91: Description overload. Cut 40% while keeping sensory anchors.
3. Line 112: "couldn't help but" → Replace with direct action.
Apply all recommendations? [Yes / Review Each / Skip]
The Monaco Editor Integration
Editor Mode provides a full-featured code editor for your prose.
Features
| Feature | Purpose |
|---|---|
| Diff View | See exactly what changed between versions |
| Version History | Every edit is saved; revert anytime |
| Line Blame | See which agent wrote which line |
| Find & Replace | Regex-powered search across manuscript |
| Split View | Compare original and edited side-by-side |
Treating Prose Like Code
Writers Factory treats your manuscript like a codebase:
- Commits: Each editing session creates a version
- Branches: Test radical edits without losing originals
- Merge: Combine the best parts of different versions
Enhancement vs. Regeneration
Not every scene needs the same level of intervention.
| Score | Recommendation | What Happens |
|---|---|---|
| 90+ | Accept | Scene is ready; minor polish only |
| 80-89 | Enhance | Run the 5-pass pipeline |
| 70-79 | Deep Edit | Multiple enhancement rounds |
| <70 | Regenerate | Return to Director Mode with feedback |
The system recommends based on the Voice Alignment score from Director Mode. Low-scoring scenes may need regeneration, not editing.
The Voice Bundle in Action
Your Voice Bundle (calibrated in Voice Mode) drives every editing decision.
What Gets Injected
When you enter Editor Mode, the system loads:
pov: third_person_limited
tense: past
voice_type: introspective_noir
metaphor_domains: ["gambling", "architecture", "decay"]
rhythm: varied_with_long_passages
vocabulary_level: elevated_but_accessible
anti_patterns:
- "couldn't help but"
- "it's important to note"
- "shivers down spine"
- "with practiced precision"
pov_constraints:
- "No head-hopping"
- "Can only know what POV character perceives"
- "Internal thoughts in italics"
How It’s Used
- Every flagged issue is checked against your Bundle
- Suggested replacements match your vocabulary level
- Pacing recommendations respect your rhythm preference
- Metaphor suggestions draw from your approved domains
API Endpoints (Technical Reference)
For Architect Track students:
| Endpoint | Method | Purpose |
|---|---|---|
/foreman/mode/editor |
POST | Enter Editor mode |
/editor/analyze |
POST | Run analysis without changes |
/editor/enhance |
POST | Run full enhancement pipeline |
/editor/voice-check |
POST | Voice alignment only |
/editor/anti-patterns |
POST | Anti-pattern scrub only |
/editor/consistency |
POST | Knowledge Graph consistency check |
/editor/diff |
GET | Get diff between versions |
/editor/revert |
POST | Revert to previous version |
Data Flow
Draft from Director Mode
│
▼
┌──────────────────────┐
│ Voice Bundle │ ← From Voice Calibration
│ (Gold Standard) │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ 5-Pass Pipeline │
│ 1. Voice Alignment │
│ 2. Anti-Patterns │
│ 3. Consistency │
│ 4. Character Voice │
│ 5. Prose Polish │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ Enhanced Draft │
│ + Change Report │
└──────────────────────┘
Common Editing Workflows
“Quick Polish”
For scenes that scored well in Director Mode:
- Enter Editor Mode
- “Polish this scene, focus on prose”
- Review suggestions
- Accept or tweak
Time: 5-10 minutes per scene
“Deep Edit”
For scenes that need significant work:
- Enter Editor Mode
- “Run full pipeline on this scene”
- Address voice alignment issues first
- Fix anti-pattern violations
- Review consistency flags
- Polish prose
- Compare diff, accept changes
Time: 20-30 minutes per scene
“Final Pass”
For the complete manuscript before export:
- Enter Editor Mode
- “Run consistency check on full manuscript”
- Address any cross-chapter issues
- “Run anti-pattern scrub on full manuscript”
- Review and accept
- Export
Time: 2-4 hours for full novel
Next Steps
Director Mode
Where drafts come from
Voice Calibration
The standard we edit against
GraphRAG
How consistency checking works
Editor Mode: Where good drafts become great manuscripts.