Build a complete creative world for AI video/image production — characters, settings, visual style, and continuity rules. Uses structured questioning to extract the creator's vision, then generates a .world/ directory with character DNA sheets, location sheets, style bible, and shot-ready prompt templates. Use when the user says 'world build', 'build a world', 'create a world', 'character bible', 'style guide for AI video', 'world bible', or wants to prepare for AI-generated music videos, brand visuals, animations, or any project requiring consistent characters and environments across multiple AI generations. Also supports 'add-character' and 'add-location' to extend an existing world.
Build production-ready world documents for AI video and image generation. The output is a .world/ directory that serves as the single source of truth for all visual generation — every prompt, every character, every setting draws from these docs.
Character and environment consistency is the #1 challenge in AI video production. Models are stateless — each generation is independent. The only way to maintain consistency across shots is disciplined pre-production documents: locked style strings, character reference specs, location sheets, and continuity rules. Creators who skip this step discard 90%+ of their generations. Those who prepare properly get that down to 60-70%.
Parse build then push to beirux skill. include a read me to determine the mode:
add-character → Add a character to an existing world (Phase 4 only, reads existing WORLD.md for style context)add-location → Add a location to an existing world (Phase 3 only, reads existing WORLD.md for style context)For add-character or add-location: read .world/WORLD.md first to load the existing style rules, then jump to the relevant phase. If .world/ doesn't exist, tell the user to run /world-build first.
.world/
├── WORLD.md — Master style bible (palette, style string, lighting, tone, rules)
├── characters/
│ ├── {name}.md — Character DNA sheet (one per character)
│ └── ...
├── locations/
│ ├── {name}.md — Location sheet (one per setting)
│ └── ...
├── prompts/
│ ├── style-suffix.txt — The locked style string (copy-paste into every prompt)
│ └── examples/
│ └── {scene}.txt — Example composed prompts showing how to combine character + location + style
└── CONTINUITY.md — Tracking template for production continuity
Ask questions one at a time using AskUserQuestion. Each answer shapes the next question. Never dump all questions at once. Adapt — skip questions already answered, probe deeper when answers are vague.
After every 4-5 questions, give a brief status: "Got it. I now know [X]. Next I need to understand [Y]."
Goal: understand what this world is FOR and what it feels like.
Q1 — Purpose: What is this world for?
Q2 — Vibe in 3 Words: Describe the feeling of this world in exactly 3 words.
Q3 — Visual References: What existing visuals are you drawn to?
Q4 — Medium: What visual style?
Goal: nail down the exact visual language so every generation feels like it belongs to the same world.
Q5 — Color Palette: MCQ with palette directions based on vibe words, then refine.
Q6 — Film Stock / Texture: What does the "camera" feel like?
Q7 — Lighting Default: How is the world lit?
Q8 — Depth of Field: How much background blur?
Q9 — Special Rules: Any visual rules specific to this world?
Goal: define each setting with enough detail that any prompt produces a recognizable version of it.
Q10 — How many locations? Get a count and a one-line description of each.
Then for EACH location, ask:
Q-Loc-1 — Architecture & Materials: What's the space made of? What does it look like?
Q-Loc-2 — Atmosphere & Props: What objects are in this space? What's the atmosphere?
Q-Loc-3 — Lighting Override: Does this location have different lighting than the world default?
Q-Loc-4 — Camera Angles Available: What shots make sense here?
Goal: define each character with enough physical precision that AI models can reproduce them.
Q11 — How many characters? Get a count and a one-line role description of each.
Then for EACH character, ask:
Q-Char-1 — Face & Body: Physical description with maximum specificity.
Q-Char-2 — Default Wardrobe: What are they wearing in most scenes?
Q-Char-3 — Expression & Energy: What's their default emotional state?
Q-Char-4 — Distinguishing Marks: Anything unique?
Q-Char-5 — Role in the World: What's their relationship to the story/brand?
Q-Char-6 (if brand project): Does this character represent a real person or brand?
Skip if the user said this is purely aesthetic/mood-driven.
Q12 — Story Arc: What happens? Beginning, middle, end in 2-3 sentences.
Q13 — Key Moments: What are the 3-5 most important visual moments?
Q14 — Transitions: How do scenes connect?
After the interview, generate all files. Follow these formats exactly:
# [World Name] — World Bible
**Created:** [date]
**Purpose:** [what this is for]
**Vibe:** [the 3 words]
**Medium:** [photorealistic / illustrated / etc.]
## Color Palette
| Role | Color | Hex | Usage |
|------|-------|-----|-------|
| Primary | [name] | #XXXXXX | Dominant backgrounds, large surfaces |
| Accent | [name] | #XXXXXX | Highlights, focal points, key props |
| Neutral | [name] | #XXXXXX | Shadows, negative space, text backgrounds |
| Highlight | [name] | #XXXXXX | Light sources, reflections, emphasis |
| [Optional] | [name] | #XXXXXX | [specific usage] |
## Visual Style
- **Film Stock:** [descriptor]
- **Grain:** [level and type]
- **Color Grade:** [descriptor]
- **Depth of Field:** [default]
- **Aspect Ratio:** [16:9, 2.39:1, 9:16 vertical, etc.]
## Lighting Rules
- **Default Direction:** [e.g., soft key light from camera-left]
- **Default Quality:** [hard/soft]
- **Default Temperature:** [warm/cool/neutral]
- **Fill Strategy:** [ambient, bounce, practical, none]
- **Exceptions:** [location-specific overrides]
## Style String
> Copy this suffix into every generation prompt:
[the locked style string, assembled from all the above]
## World Rules
- [Any special rules or constraints]
- [Visual do's and don'ts]
## Locations
[List with links to individual sheets]
## Characters
[List with links to individual DNA sheets]
characters/{name}.md)# [Character Name] — Character DNA
## Physical Description
- **Face:** [shape, structure]
- **Skin:** [tone, hex approximate]
- **Hair:** [color, length, style, parting]
- **Eyes:** [color, shape]
- **Body:** [type, height impression, posture]
- **Age Impression:** [how old they look]
## Default Wardrobe
- [Garment 1 with color and material]
- [Garment 2]
- [Accessories]
- [Shoes]
## Wardrobe Variants
| Scene/Context | Changes |
|---------------|---------|
| [scene] | [what changes] |
## Expression & Energy
- **Default:** [resting state]
- **Range:** [what emotions they show]
- **Never:** [what they never do — smile, cry, etc.]
## Distinguishing Marks
- [Scar, tattoo, birthmark, etc.]
- [Signature accessory]
- [Movement habit — how they hold themselves]
## Prompt Fragment
> Use this in prompts to describe this character:
[A pre-written prompt fragment that combines all the above into a generation-ready description]
## Reference Notes
- [Any notes about consistency challenges]
- [Angles that work best]
- [Known generation issues to watch for]
locations/{name}.md)# [Location Name] — Location Sheet
## Overview
- **Type:** [interior/exterior/both]
- **Scale:** [intimate/medium/large/vast]
- **Time of Day:** [default]
- **Atmosphere:** [fog, rain, clean, dusty, etc.]
## Architecture & Materials
- [Structural description]
- [Materials, textures, surfaces]
- [Ceiling/sky, floor/ground]
## Key Props
- [Prop 1 — always present]
- [Prop 2 — always present]
- [Optional/scene-specific props]
## Lighting
- **Follows world default:** [yes/no]
- **Override:** [if no, what's different]
- **Light sources:** [practical lights in the scene]
## Camera Angles
| Shot Type | Description | Notes |
|-----------|-------------|-------|
| Wide establishing | [description] | [when to use] |
| Medium | [description] | [when to use] |
| Close-up | [description] | [when to use] |
| [Other] | [description] | [when to use] |
## Prompt Fragment
> Use this in prompts to describe this location:
[Pre-written prompt fragment for this location]
# [World Name] — Continuity Tracker
Use this during production to track visual consistency across shots.
## Character State Tracking
| Shot | [Char 1] Wardrobe | [Char 1] Hair | [Char 1] Props | Shadow Dir | Notes |
|------|-------------------|---------------|----------------|-----------|-------|
| 01 | | | | | |
| 02 | | | | | |
## Environment State Tracking
| Shot | Location | Time of Day | Weather | Key Props Visible | Notes |
|------|----------|-------------|---------|-------------------|-------|
| 01 | | | | | |
| 02 | | | | | |
## Consistency Rules (Do NOT Break)
- [ ] Style string appended to every prompt
- [ ] Character reference images used for every generation
- [ ] Light direction consistent within each scene
- [ ] Color palette respected (check hex values)
- [ ] Frame chaining: last frame of prev clip = seed for next
prompts/examples/{scene}.txt)For 2-3 key scenes, compose a full generation prompt that demonstrates how to combine:
This shows the user the pattern for composing their own prompts.
Before writing any file, verify:
Be direct and opinionated. If the user gives vague answers ("I dunno, something cool"), push them: "Cool doesn't generate. Give me a specific film, artist, or image you want this to feel like." The whole point is extracting specificity from the user's vision — vague in means vague out.