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.
No credit card · 3 free sessions · Results in minutes
Story Architect researches genre conventions and builds rich character profiles with backstories, motivations, and internal conflicts
Narrative Writer transforms architecture into vivid prose — dialogue, atmosphere, action, and inner monologue
Story Editor critiques across 5 quality dimensions, Writer revises, Editor verifies — up to 3 rounds of refinement
Each character reviews their own lines for authenticity — are the words true to their personality and motivation?
Two Style Experts with opposing approaches debate prose style. Story Architect judges the winning direction
Plot coherence, character depth, dialogue authenticity, emotional impact, prose quality, and overall engagement
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.
Full character dossiers with personality traits, speaking style, backstory, motivation, and internal conflict — the DNA of authentic characters.
Each scene outlined with location, characters present, purpose, emotional arc, dialogue goals, and state transitions — a narrative blueprint.
The Critique-Correct-Verify loop from FilmAgent: Editor critiques across 5 dimensions, Writer revises, Editor verifies. Real editorial rigor.
Each character agent reviews their own dialogue for authenticity — catching moments where words don't match personality or motivation.
Two Style Experts debate opposing prose approaches (immersive vs economical), then the Architect judges — ensuring intentional stylistic choices.
Six-axis evaluation: plot coherence, character depth, dialogue authenticity, emotional impact, prose quality, and overall engagement.
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.
Research genre & theme, build character profiles with backstories & motivations, outline scenes with emotional arcs
Transform the architecture into a full prose draft — dialogue, action, atmosphere, and inner monologue
Story Editor critiques → Writer revises → Editor verifies. Up to 3 rounds of Critique-Correct-Verify
Each character reviews their own dialogue for authenticity — voice, motivation, and emotional truth
Architect aggregates character feedback → Writer applies → Architect verifies. Up to 3 CCV rounds
Two Style Experts with opposing approaches debate prose style → Story Architect judges and decides direction
Six-axis narrative quality evaluation. Synthesizer weaves all feedback into the final polished story
Quality gate — if below threshold, the entire pipeline loops (up to 2 iterations) with specific guidance
🔄 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
"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
"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
"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
"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
"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
"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
A complete narrative journey — from architecture through editorial cycles, character voice reviews, style deliberation, and the final polished story.
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.
LOOM is a narrative engine. It writes, edits, and refines stories with character-level authenticity — not a brainstorming tool or a debate engine.
Develop narratives with editorial-grade feedback loops and character voice validation before publishing
Craft branching narratives with consistent character voices and emotionally resonant dialogue
Produce polished long-form stories, serial fiction, or branded narrative content at scale
Teach narrative craft by showing how editorial feedback, character voice, and style choices shape stories
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