Narrative Story Engine
LOOM

Weave stories with five expert voices.

Story Architect · Narrative Writer · Story Editor · Character Voices · Style Council

LOOM transforms a simple prompt into a polished, multi-layered narrative — five specialized agents collaborate through editorial CCV loops, character voice reviews, and a Style Debate-Judge council to craft stories with depth, authenticity, and style.

5
Specialist roles
8
Pipeline phases
2
Core algorithms

No credit card · 3 free sessions · Results in minutes

World Building

Story Architect researches genre conventions and builds rich character profiles with backstories, motivations, and internal conflicts

Prose Crafting

Narrative Writer transforms architecture into vivid prose — dialogue, atmosphere, action, and inner monologue

CCV Editorial Loops

Story Editor critiques across 5 quality dimensions, Writer revises, Editor verifies — up to 3 rounds of refinement

Character Voices

Each character reviews their own lines for authenticity — are the words true to their personality and motivation?

Style Debate-Judge

Two Style Experts with opposing approaches debate prose style. Story Architect judges the winning direction

6-Axis Quality Score

Plot coherence, character depth, dialogue authenticity, emotional impact, prose quality, and overall engagement

What You Get

Not a rough draft. A story woven by five specialists.

Every narrative passes through an architect, a writer, an editor, character voices, and a style council. What emerges has the depth of collaborative fiction — crafted by AI agents that actually listen to each other.

Rich Character Profiles

Full character dossiers with personality traits, speaking style, backstory, motivation, and internal conflict — the DNA of authentic characters.

Structured Scene Outlines

Each scene outlined with location, characters present, purpose, emotional arc, dialogue goals, and state transitions — a narrative blueprint.

CCV Editorial Cycles

The Critique-Correct-Verify loop from FilmAgent: Editor critiques across 5 dimensions, Writer revises, Editor verifies. Real editorial rigor.

Character Voice Feedback

Each character agent reviews their own dialogue for authenticity — catching moments where words don't match personality or motivation.

Style Council Verdict

Two Style Experts debate opposing prose approaches (immersive vs economical), then the Architect judges — ensuring intentional stylistic choices.

Narrative Quality Scorecard

Six-axis evaluation: plot coherence, character depth, dialogue authenticity, emotional impact, prose quality, and overall engagement.

How It Works

FilmAgent's multi-agent pipeline, adapted for narrative.

An 8-phase pipeline with two Critique-Correct-Verify loops, character voice reviews, and a Style Debate-Judge council. Five specialist roles collaborate to weave a story that's architecturally sound and emotionally resonant.

1
1

Story Architect

Research genre & theme, build character profiles with backstories & motivations, outline scenes with emotional arcs

2
2

Narrative Writer

Transform the architecture into a full prose draft — dialogue, action, atmosphere, and inner monologue

3
3

Editor CCV Loop

Story Editor critiques → Writer revises → Editor verifies. Up to 3 rounds of Critique-Correct-Verify

4
4

Character Voices

Each character reviews their own dialogue for authenticity — voice, motivation, and emotional truth

5
5

Architect CCV Loop

Architect aggregates character feedback → Writer applies → Architect verifies. Up to 3 CCV rounds

6
6

Style Council

Two Style Experts with opposing approaches debate prose style → Story Architect judges and decides direction

7
7

Quality & Synthesis

Six-axis narrative quality evaluation. Synthesizer weaves all feedback into the final polished story

8
8

Pipeline Review

Quality gate — if below threshold, the entire pipeline loops (up to 2 iterations) with specific guidance

Two Core Algorithms from FilmAgent (arXiv:2501.12909)

🔄 Critique-Correct-Verify (CCV)

Agent P generates a draft → Agent Q critiques it → P corrects → Q verifies. Max 3 iterations until approved. Used in Phases 3 & 5.

⚖️ Debate-Judge

Two peer agents independently propose styles → debate their differences → a Judge agent decides the winning direction. Used in Phase 6.

Adapted from FilmAgent's multi-agent film production pipeline · 5 specialized narrative roles · 8-phase story weaving · arXiv:2501.12909

Real Examples

What people weave on the loom.

Literary Fiction

"A retired lighthouse keeper discovers letters from the future"

Evolved into a 6-scene mystery with 3 fully voiced characters, lyrical-sparse hybrid style, and a 8.7/10 quality score

Drama

"Two rival chefs must collaborate to save a failing restaurant"

Character voices caught 12 dialogue inconsistencies; Style Council chose sharp wit for kitchen scenes, warmth for quiet moments

Fantasy

"A child's imaginary friend turns out to be real — but only she can see them"

Editor CCV cycle tightened the pacing from 6/10 to 9/10 across 3 rounds; character voice for the imaginary friend created a distinct ethereal register

Science Fiction

"The last human librarian in a world run by AI"

Style debate produced a retrofuturist tone blending Bradbury warmth with Ishiguro restraint; 4 character voices shaped complex human-AI dynamics

Mystery

"A detective solves crimes by reading people's abandoned journals"

Story architect built 5 interconnected journal narratives; the detective's internal monologue became the story's emotional core

Mythology

"Write a folktale about why the moon changes shape"

Character voices created distinct registers for Moon, Sun, and Stars; Style Council chose oral-tradition cadence with modern emotional depth

Sample Output

What LOOM delivers.

A complete narrative journey — from architecture through editorial cycles, character voice reviews, style deliberation, and the final polished story.

LOOM Narrative Report
Sample Output

Prompt: “A retired lighthouse keeper discovers letters from the future washed up on the shore”

Chapter 1 — Story Architecture

Setting: Abandoned coastal town, 2040s post-storm. Characters: Elias (72, retired keeper — stoic, haunted by wife's death), Maren (the letter writer, revealed in Act 3), Theo (local teenager who helps Elias). 6 scenes outlined with emotional arcs from isolation → curiosity → obsession → revelation → acceptance.

Chapter 2 — First Draft (Narrative Writer)

"The seventh letter arrived on a Tuesday, sealed in a glass bottle that should not have survived the crossing. Elias held it to the morning light. The handwriting was his wife's — or close enough to make his hands shake. 'Dearest E,' it began, 'the lighthouse still stands where you left it. I can see it from here. From when I am....'"

Chapter 3 — Editor CCV Cycle 2 (REVISE → Approved)

Critique: Plot coherence 8/10, but tension drops in Scene 4 when Theo's subplot overshadows the mystery. Pacing: 7/10 — Act 2 middle sags. Dialogue: 9/10 — Elias's voice is excellent. Revision focus: tighten Scene 4, weave Theo deeper into the central mystery. Verified after 2nd pass.

Chapter 4 — Character Voice Review: Elias

"My speech patterns are right — short sentences, nautical metaphors. But in Scene 5, I wouldn't say 'I'm scared.' I'd say 'The tide's turning' and let silence do the rest. Also, my reaction to the final letter needs more restraint — 40 years of grief doesn't break in one paragraph."

Chapter 6 — Style Council Verdict

Immersive Stylist proposed: rich atmospheric descriptions, extended metaphors linking sea and time. Economical Stylist proposed: spare Hemingway-esque prose, let subtext carry emotion. Judge verdict: HYBRID — immersive for the letter-reading scenes (the story's emotional peaks), economical for action and dialogue.

Chapter 7 — Narrative Quality Scorecard

Plot Coherence: 9/10 · Character Depth: 9/10 · Dialogue Authenticity: 8/10 · Emotional Impact: 9/10 · Prose Quality: 8/10 · Overall Engagement: 9/10 · Composite: 8.7 — Excellent. Two CCV cycles completed, 4 character voices consulted, style council resolved.

+ Full character profiles · Scene outlines · CCV editorial history · Style debate transcript · Quality evolution across iterations
LOOM vs. Other Agents

Story crafting, not just idea generation.

LOOM is a narrative engine. It writes, edits, and refines stories with character-level authenticity — not a brainstorming tool or a debate engine.

Dimension
🧵 LOOM
💬 Discussion
🔥 CRUCIBLE
Purpose
Craft polished narrative stories
Generate diverse creative ideas
Refine ideas through dialectic
Structure
8-phase pipeline with 2 CCV loops + Style Council
Open multi-round roundtable
Thesis → antithesis → synthesis
Roles
5 specialists: Architect, Writer, Editor, Character Voice, Style Expert
6 creative personas
1 thesis holder + 1 opposition + MAMV voters
Quality Control
Two CCV loops + Debate-Judge + 6-axis scorecard
Keyword convergence check
MAMV validity + novelty voting
Unique Feature
Characters review their own dialogue for authenticity
Cross-visible free discussion
Dynamic annealing temperature
Output
Complete narrative with character depth & style
Consolidated creative list
One deeply evolved idea + trace
Best For
"Write me a compelling story"
"Give me many ideas"
"Make this idea stronger"
Use when “Weave me a story...”
Use when “Give me many ideas...”
Use when “Refine & deepen this...”
Who Uses LOOM

For anyone who wants to tell a better story.

✍️

Fiction Writers

Develop narratives with editorial-grade feedback loops and character voice validation before publishing

🎮

Game Designers

Craft branching narratives with consistent character voices and emotionally resonant dialogue

📚

Content Creators

Produce polished long-form stories, serial fiction, or branded narrative content at scale

🎓

Educators

Teach narrative craft by showing how editorial feedback, character voice, and style choices shape stories

Ready to weave your story?

Give LOOM a prompt and watch five specialist agents architect, write, edit, voice, and polish a narrative that's more than the sum of its parts.

No credit card · 3 free sessions · Based on arXiv:2501.12909