3.1.4: Building Your Style Database

• Time to Complete: 20 minutes
• Prerequisites: Consistency & Style (Module 3.1.3)

Start this module in Cursor: Run /start-3-1-4 to begin the interactive experience.

Overview

Module 3.1.4 teaches you the meta-skill of image generation: building a personal creative toolkit that grows more valuable with every project. Instead of starting from scratch each time, you’ll build a library of proven styles you can apply instantly.

Key takeaway: This isn’t just about generating images - it’s about building a compounding system. Every style you save makes future work faster and better.

The Meta-Skill

Most people use AI image tools as one-off novelties:

• Generate something
• Use it once
• Forget about it
• Start over next time

Power users do something different:

• Generate something
• Save the winning prompt
• Reuse it across projects
• Build a library over time

The difference compounds dramatically. After a few months, you have a library of proven styles for every situation - personas, diagrams, presentations, marketing. One command: “use style #12” and you’re done.

Your Style Library

The Nano Banana module includes style-library.html - a local HTML file that visualizes your style collection.

What’s in the Library

Each style entry contains:

Using the Library

To apply a style:

Generate [your subject] using style #5

The AI looks up the style, retrieves the prompt, and applies it to your subject.

To browse styles: Open style-library.html in your browser. Use the filter buttons to explore categories.

Categories

The library organizes styles into categories:

Three Ways to Grow Your Library

Method 1: Save As You Go

The simplest approach - whenever you create something you like, save it.

Workflow:

1. Generate an image you’re happy with
2. Tell the AI: “Add this style to my library”
3. The AI updates style-library.html with the new entry
4. The style is now available for future use

Example:

You: Generate a minimalist journey map with soft colors

[The AI generates image]

You: I love this style. Add it to my library.

AI: Done! Added as style #15 "Minimalist Journey Map"
in the Flow category.

Best for: Capturing your own successful experiments.

Method 2: Collect From Online

Great prompts are everywhere - social media, tutorials, prompt databases. When you find one, test it, then save it.

Workflow:

1. Find a promising prompt online
2. Test it: “Generate [subject] using this prompt: [paste prompt]”
3. If it works well, tell the AI to save it
4. Now it’s part of your permanent library

Example:

You: I found this prompt online - test it with a coffee shop:
"Make a photo that is perfectly isometric. It is not a miniature,
it is a captured photo that just happened to be perfectly isometric."

[The AI generates isometric coffee shop]

You: Great! Save this as "Perfect Isometric Photo"

AI: Added as style #16 in 3D/Isometric category.

Best for: Curating the best prompts from the community.

Method 3: Extract From Any Image

This is the power move. See an image you love? Extract its style as text, then recreate it anytime.

How it works:

1. Provide an image (file path or URL)
2. The AI runs the style extraction module
3. The module analyzes: colors, lighting, composition, textures, mood, effects
4. Returns a detailed natural language description
5. You can now recreate that style from pure text

Workflow:

You: Extract the style from this image: [path to image]

AI: [Runs style_extract.py]

Here's the extracted style:
"Warm analog photography aesthetic with slight film grain.
Golden hour lighting creates soft shadows. Muted earth tones
with desaturated greens and warm highlights. Candid composition
with slight motion blur suggesting authentic moments..."

You: Generate a portrait using that extracted style

[The AI applies the style to a new subject]

You: Perfect. Save this to my library.

Best for:

• Capturing styles from Pinterest, Instagram, or design inspiration
• Recreating looks from competitors or brands you admire
• Building a style from a single reference image

The Style Extraction Module

The style_extract.py module uses Gemini’s image understanding to analyze any image and describe its visual style.

What it captures:

• Color palette and temperature
• Lighting quality and direction
• Composition and framing
• Textures and materials
• Mood and atmosphere
• Typography (if present)
• Special effects or treatments

Why it works: Gemini 3 Pro has exceptional image understanding. It can articulate visual qualities that would be hard to describe manually.

Managing Your Library

Adding Styles

When you tell the AI to save a style, it:

1. Assigns the next available ID number
2. Creates a descriptive name
3. Chooses an appropriate category
4. Adds relevant tags
5. Saves a thumbnail image
6. Updates style-library.html

Customizing the Library

The style library is just an HTML file - you can customize it:

• Add columns: “Ask the AI to add a ‘difficulty’ column”
• Change layout: “Reorganize the library by category”
• Add search: “Add a search bar to the library”
• Modify design: “Make the library dark mode”

It’s YOUR toolkit. Shape it to fit your workflow.

Tags and Filtering

Tags help you find styles quickly. The canonical tags by category:

Use the filter buttons in style-library.html to browse by category.

The Compounding Effect

Here’s why this matters:

Week 1: You have 5 styles

• Basic coverage of common needs

Month 1: You have 15 styles

• Patterns emerging for your work
• Starting to have “go-to” styles

Month 3: You have 30+ styles

• Comprehensive coverage
• Rarely need to create from scratch
• “Use style #23” becomes your workflow

Year 1: You have 50+ battle-tested styles

• Professional-grade library
• Each style proven across multiple projects
• Minutes to produce what used to take hours

This is the compounding effect. Every project makes your library more valuable.

Workflow Examples

Example 1: Consistent Brand Visuals

Scenario: Your company has a specific visual style you need to maintain.

Solution:

1. Collect 3-5 on-brand images
2. Extract the style from your best example
3. Save as “Company Brand Style”
4. Apply to all future graphics

Generate a feature announcement using style #8 (Company Brand)

Example 2: Rapid Exploration

Scenario: You need 10 user persona portraits, all consistent.

Solution:

1. Browse your library for persona styles
2. Pick one that fits the project tone
3. Generate all 10 using the same style number

Generate Sarah Chen portrait using style #12
Generate Marcus Johnson portrait using style #12
Generate Priya Patel portrait using style #12
...

Example 3: Recreating Inspiration

Scenario: Your CEO shares a competitor’s diagram they love.

Solution:

1. Extract the style from the competitor’s image
2. Save it to your library (with appropriate name)
3. Apply it to your own content

Extract the style from this diagram: [competitor's image]
Generate our product roadmap using that style
Save this as "Executive Diagram Style"

Best Practices

Do:

• Save styles frequently - if you like it, save it
• Use descriptive names - “Warm Portrait” beats “Style 7”
• Tag thoroughly - helps future you find things
• Extract aggressively - see something good? grab it
• Review periodically - prune unused styles

Don’t:

• Don’t hoard mediocre styles - quality over quantity
• Don’t forget to save - capture wins immediately
• Don’t over-categorize - keep it simple
• Don’t ignore extracted styles - they’re often the best

Troubleshooting

Style library won’t open

• Make sure you’re opening style-library.html in a browser (not a text editor)
• Try right-click → Open With → your browser

New style doesn’t appear

• Refresh the browser (Cmd+R / Ctrl+R)
• Check that the AI confirmed the save

Extracted style doesn’t match original

• Some visual effects are hard to describe in text
• Try extracting from a simpler example
• Combine with a reference image for better results

Can’t find the right style

• Use the category filters
• Try searching by tag
• Ask the AI: “Which styles work for [your use case]?”

Quick Reference

Three Methods to Grow Your Library:
  1. Save as you go → Good work? Save it.
  2. Collect from online → Test, then save.
  3. Extract from images → The power move.

Using Styles:
  "Generate [subject] using style #X"
  "Apply style #Y to this"
  "Which styles work for journey maps?"

Managing Styles:
  "Add this to my library"
  "Save this as [name]"
  "Show me all persona styles"

What’s Next?

You’ve completed Module 3.1 - Introduction to Image Generation. You understand the mechanics, the prompting principles, and how to build a compounding system.

Module 3.2 puts these skills to work on real PM tasks. You’ll create personas, journey maps, wireframes, architecture diagrams, and marketing assets.

Interactive track: Type /start-3-2-1

Resources

• Dribbble - Design inspiration, find styles to add to your library
• Behance - Creative portfolios, extract styles from professional work
• Pinterest - Visual discovery, curate style inspiration boards

About This Course

Created by Carl Vellotti. Check out The Full Stack PM for more PM builder content.

Source Repository: github.com/carlvellotti/claude-code-pm-course