Prompt Engineering

Write effective prompts for better AI rendering results. Learn the essentials of prompt writing.

Prompt Engineering

Better prompts = better results. Learn how to write prompts that get you the renders you want.

The Basics

A prompt is a text description that tells the AI what to generate. Structure it like this:

[Subject] + [Style] + [Details] + [Environment]

Example:

Modern minimalist house with floor-to-ceiling windows, Scandinavian style, white walls and natural wood accents, surrounded by pine trees, sunny day with soft shadows

Key Components

1. Subject

Be specific about what you're creating:

Good: "Contemporary glass office building" ❌ Poor: "Building"

Include:

  • Building type (house, office, restaurant)
  • Architectural style (modern, traditional, contemporary)
  • Key features (windows, materials, layout)

2. Style

Specify the visual style:

  • Architectural: Modern, Art Deco, Brutalist
  • Rendering: Photorealistic, sketch, wireframe
  • Artistic: Minimalist, ornate, industrial

Examples:

  • "Brutalist concrete architecture"
  • "Minimalist Scandinavian design"

3. Materials

Describe materials explicitly:

  • Exterior: Concrete, glass, wood, metal, stone
  • Interior: Marble, hardwood, fabric, leather
  • Textures: Smooth, rough, polished, weathered

Examples:

  • "Weathered cedar wood siding"
  • "Polished marble floors"

4. Environment

Set the scene:

  • Location: Urban, rural, coastal, mountain
  • Surroundings: Trees, water, other buildings
  • Time of Day: Morning, noon, sunset, night
  • Weather: Sunny, cloudy, rainy, foggy

Examples:

  • "Urban setting with city skyline in background"
  • "Coastal location with ocean view"

5. Lighting

Describe lighting:

  • Natural: Direct sun, diffused, golden hour
  • Artificial: Warm, cool, accent lighting
  • Shadows: Soft, hard, dramatic

Examples:

  • "Golden hour sunlight casting long shadows"
  • "Soft diffused daylight through large windows"

Advanced Techniques

Negative Prompts

Specify what to avoid:

  • "No people, no vehicles, no text"
  • "Avoid cluttered backgrounds"

Style References

Reference specific styles:

  • "In the style of Frank Lloyd Wright"
  • "Bauhaus-inspired design"

Technical Details

Include architectural specifics:

  • "Cantilevered structure"
  • "Post-and-beam construction"

Common Mistakes

Too vague: "Nice building" ✅ Better: "Modern glass office building with geometric facade"

Too complex: "A building with everything" ✅ Better: Focus on key elements

Missing context: "House" ✅ Better: "Contemporary single-family house in suburban setting"

Inconsistent style: Mixing conflicting styles ✅ Better: Choose one style and be consistent

Iterative Refinement

Start broad: "Modern house"

Add details: "Modern minimalist house with large windows and flat roof"

Include context: "Modern minimalist house with large windows and flat roof, surrounded by mature trees, sunny day"

Specify style: "Modern minimalist house with large windows and flat roof, surrounded by mature trees, sunny day, photorealistic architectural photography"

Using References

Image references:

  • Upload clear reference images
  • Describe what you want from the reference
  • Combine multiple references

Previous renders:

  • Reference successful renders with @v1, @v2, @latest
  • Build on what worked
  • Iterate on concepts

Examples

Good Prompt 1

Contemporary glass and steel office building, minimalist design with clean lines, curtain wall facade with bronze tinted glass, located in downtown urban setting, late afternoon golden hour lighting, aerial perspective, photorealistic architectural photography

Good Prompt 2

Scandinavian-style single-family house, white painted wood siding with black window frames, pitched roof with dark shingles, surrounded by birch trees and wildflowers, soft morning light with gentle shadows, eye-level street view, photorealistic, warm and inviting atmosphere

Practice Tips

  1. Start simple: Begin with basic prompts
  2. Iterate: Refine based on results
  3. Experiment: Try different approaches
  4. Learn from results: See what works
  5. Build library: Save successful prompts

Next Steps

Prompt Engineering