How to Specify Brushwork Style for Illustration & Art in GPT-Image2

To specify brushwork style for Illustration & Art in GPT-Image2, pass the brushMaterial and brushStrokes parameters alongside the illustration style ID, which the scripts/generate-style-skill.mjs utility merges into the final prompt template defined in data/style-library.json.

The freestylefly/awesome-gpt-image-2 repository provides a structured style library that enables granular control over generated artwork. When working with the Illustration & Art category, you can specify precise brushwork characteristics—such as oil, watercolor, or ink techniques—by leveraging specific prompt arguments that the system injects into the base template before calling the image generation API.

Understanding the Style Library Architecture

The style library (data/style-library.json) serves as the central catalog that drives the visual language of every generated image. For the Illustration & Art category, the engine targets the Illustration style entry (ID: illustration).

According to the source code, brushwork control relies on four key prompt arguments:

• brushMaterial — Defines the medium the virtual brush uses, such as oil, watercolor, ink, dry-brush, or gouache.
• brushStrokes — Characterizes the stroke behavior using descriptors like vigorous, delicate, freehand, or calligraphic.
• palette — Optional color scheme qualifier (e.g., warm, cool) that influences the color selection.
• mood — Optional atmospheric quality (e.g., serene, heroic) that affects the overall emotional tone.

These fields are merged into the base prompt template defined in agents/skills/gpt-image-2-style-library/references/style-library.md. The merging logic executes in scripts/generate-style-skill.mjs, which reads the JSON library, selects the Illustration entry, and constructs the final prompt for the api/generate-image.js handler.

The Brushwork Specification Workflow

The process follows five distinct stages:

1. Select the category — Use cat-illustration for Illustration & Art.
2. Pick the style — Specify the style ID "illustration".
3. Supply brush arguments — Include brushMaterial, brushStrokes, and optionally palette and mood.
4. Generate the prompt — The skill generator assembles text like: illustration, oil brush, vigorous brushstrokes, warm palette, serene mood, <subject>.
5. Execute generation — The assembled prompt routes through api/generate-image.js to the underlying GPT-Image model.

Because brushwork is expressed as plain-text tokens, the model interprets these descriptors and renders appropriate textures—such as visible wet-brush effects for watercolor or bold ink lines for calligraphic styles.

Practical Implementation Examples

Direct API Call via Node.js

// demo-brushwork.js
import fetch from 'node-fetch';

const payload = {
  style: 'illustration',
  subject: 'a futuristic city skyline at sunset',
  brushMaterial: 'watercolor',
  brushStrokes: 'delicate',
  palette: 'cool',
  mood: 'dreamy'
};

fetch('https://your-deployment.com/api/generate-image', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify(payload)
})
  .then(r => r.json())
  .then(data => console.log('Image URL:', data.imageUrl))
  .catch(console.error);

The api/generate-image.js handler parses this JSON body, combines the fields with the Illustration template, and forwards the final prompt to the GPT-Image backend.

CLI Skill Generator

npm run generate-style-skill \
  -- --style illustration \
  --subject "a tranquil bamboo forest" \
  --brushMaterial ink \
  --brushStrokes "freehand" \
  --palette "green" \
  --mood "peaceful"

The scripts/generate-style-skill.mjs script reads these arguments, looks up the Illustration entry in data/style-library.json, and assembles the complete prompt string.

Inline Raw Prompt Syntax

If you prefer manual prompt construction, use this syntax:

illustration, brush material: oil, brush strokes: vigorous, palette: warm, mood: heroic, a heroic samurai on a stormy battlefield

Passing this string to the image generation endpoint yields oil-paint style output with bold, dynamic strokes.

Key Source Files

• data/style-library.json — Central catalog containing the Illustration entry and template keywords.
• agents/skills/gpt-image-2-style-library/references/style-library.md — Human-readable reference documenting prompt composition including brush-related fields.
• scripts/generate-style-skill.mjs — CLI tool that merges user arguments with library templates.
• api/generate-image.js — Server-side handler that builds prompts and interfaces with the GPT-Image model.

Summary

• The Illustration & Art category uses the illustration style ID from data/style-library.json.
• Control brushwork through brushMaterial (medium) and brushStrokes (technique) parameters.
• Optional palette and mood arguments refine color schemes and atmosphere.
• The scripts/generate-style-skill.mjs utility automates prompt assembly by merging your inputs with the base template.
• Final prompt generation occurs in api/generate-image.js, which forwards structured text to the GPT-Image model.

Frequently Asked Questions

What values are valid for brushMaterial and brushStrokes?

Valid values include standard artistic media such as oil, watercolor, ink, dry-brush, and gouache for brushMaterial, and descriptive terms like vigorous, delicate, freehand, or calligraphic for brushStrokes. The model interprets these as plain-text tokens to render appropriate textures.

Can I use brushwork parameters with other categories besides Illustration?

The brushMaterial and brushStrokes fields are specifically designed for the Illustration style within the Illustration & Art category. Other categories in data/style-library.json utilize different parameter sets optimized for their respective visual languages.

Do I need to specify palette and mood, or are they optional?

Both palette and mood are optional arguments. If omitted, the GPT-Image model selects color schemes and atmospheric qualities based on the subject matter and brushwork style you have defined.

How does the system handle conflicting style parameters?

The scripts/generate-style-skill.mjs script performs a structured merge of user-provided arguments with the base Illustration template. Explicit user values override defaults, while undefined optional parameters are excluded from the final prompt sent to api/generate-image.js.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}