world-labs-text-prompt

name	world-labs-text-prompt
description	Text-to-world generation best practices, prompt structure, style descriptors
allowed-tools	["Bash","WebFetch"]

World Labs Text Prompting

Text prompts are the simplest way to generate 3D worlds. Describe a location in natural language, and the model creates an immersive environment.

Quick Reference

• Max length: 2,000 characters
• Cost: 1,580 credits (plus) / 230 credits (mini)
• Generation time: ~5 min (plus) / ~30-45 sec (mini)

Official Example

From World Labs documentation:

A warm, rustic cabin living room with a glowing stone fireplace, cozy leather sofa, wooden beams, and large windows overlooking a snowy forest.

Prompt Structure

Effective prompts describe a location with:

[Setting/Location] + [Materials/Textures] + [Lighting/Atmosphere] + [Environmental Context]

Components

1. Setting: The primary location type (cabin, temple, alley, forest)
2. Materials: Specific textures (rustic wood, polished marble, weathered stone)
3. Lighting: Light sources and quality (glowing, warm, flickering, dusk)
4. Context: Surrounding environment (overlooking forest, narrow corridor)

Style Descriptors

Lighting Keywords

Keyword	Effect
glowing	Soft emissive light sources
warm	Cozy, orange-tinted lighting
flickering	Dynamic light suggestion
dusk / dawn	Transitional daylight
neon-lit	Vibrant artificial lighting
candlelit	Intimate, warm glow
moonlit	Cool, ethereal nighttime

Atmosphere Keywords

Keyword	Effect
cozy	Intimate, inviting feel
atmospheric	Mood-heavy, immersive
misty / foggy	Depth, mystery
dusty	Age, particles
puddles	Wet surfaces, reflections
snowy	Winter atmosphere

Material Keywords

Keyword	Effect
rustic	Natural, aged wood
polished	Reflective, clean surfaces
weathered	Worn, aged textures
mossy	Organic growth
metallic	Industrial surfaces
leather	Furniture textures

Style Keywords

Keyword	Effect
sci-fi	Futuristic elements
fantasy	Magical, otherworldly
modern	Contemporary design
vintage	Period aesthetics
immersive	Full environmental feel

Best Practices

DO

✅ Focus on location: "A cabin living room" not "a picture of a cabin" ✅ Be specific about materials: "glowing stone fireplace" not just "fireplace" ✅ Include spatial context: "large windows overlooking a snowy forest" ✅ Add atmosphere: "warm, cozy, flickering light" ✅ Use concrete details: "wooden beams, leather sofa"

DON'T

❌ Describe people or animals as main subjects ❌ Use vague adjectives: "nice", "beautiful", "amazing" ❌ Request camera angles: handled in Studio, not generation ❌ Contradict styles: "futuristic medieval" confuses the model ❌ Exceed 2,000 characters: prompts are truncated

Example Prompts

Cyberpunk Alley

A cozy sci-fi alley at dusk, flickering neon signs, puddles reflecting colorful lights, narrow corridor between tall buildings, steam rising from vents, cables overhead

Hidden Temple

A mossy ancient temple hidden in a jungle, neon lanterns casting warm pools of light, stone pillars covered in vines, shallow reflecting pool in the center, mysterious and serene atmosphere

Cozy Interior

A warm coffee shop interior on a rainy day, exposed brick walls, wooden tables, rain-streaked windows overlooking a city street, Edison bulbs casting warm light, bookshelves along one wall, leather seating

Natural Landscape

A dramatic coastal cliff at golden hour, crashing waves against ancient rock formations, wild grass swaying in the wind, salt spray catching sunlight, moody clouds on the horizon

API Usage

{
  "model": "Marble 0.1-plus",
  "world_prompt": {
    "type": "text",
    "text_prompt": "A warm, rustic cabin living room with a glowing stone fireplace, cozy leather sofa, wooden beams, and large windows overlooking a snowy forest.",
    "disable_recaption": false
  }
}

Parameters

Parameter	Type	Description
text_prompt	string	Your location description (max 2,000 chars)
disable_recaption	boolean	If true, uses your prompt exactly; if false, model may enhance/clarify

Iteration Tips

1. Start with mini: Use Marble 0.1-mini for quick iterations (230 credits, ~30 sec)
2. Refine prompts: Adjust based on results
3. Final generation: Use Marble 0.1-plus for production quality (1,580 credits, ~5 min)

Combining with Images

Text prompts can accompany image inputs to guide interpretation:

{
  "model": "Marble 0.1-plus",
  "world_prompt": {
    "type": "image",
    "image_prompt": {
      "source": "media_asset",
      "media_asset_id": "your-media-asset-id"
    },
    "text_prompt": "Transform into a mystical nighttime scene with aurora borealis"
  }
}

When combined with images:

• Text guides the model's interpretation of ambiguous areas
• Can shift lighting, atmosphere, or style
• If omitted, model auto-generates a caption

Related Skills

• world-labs-api - API integration details
• world-labs-image-prompt - Single image input
• world-labs-studio - Camera animation and recording