Generate Images

🖼 Generate and Inject Chapter Images

Updated for:

• ✅ Default paths for prompt file and output directory

• ✅ Chapter-based prompts

• ✅ Character profile injection

• ✅ Style override support

• ✅ CLI overrides (e.g., API key)

• ✅ Auto-skip existing images

---

📌 Overview

Task	File / Script	Description
🧠 Generate images	scripts/generate_images.py	Creates images using DeepAI based on structured JSON prompt files
✍️ Inject into chapters	scripts/inject_images.py	Inserts  tags into Markdown files
🧰 Character descriptions	scripts/data/character_profiles.json	Stores reusable character descriptions for image prompts
🧰 Prompt JSON file	scripts/data/image_prompts.json	Default prompt template with chapters, characters, and prompts

---

📁 Directory Structure (Example)

project-root/
├── manuscript/
│   └── chapters/
│       ├── 01-holiday-in-ohio.md
│       └── ...
├── assets/
│   ├── illustrations/                     ← 🖼 Output images (default)
│   └── prompts/
│       └── holiday_ch1_prompts.json       ← Alternative prompt file (optional)
├── scripts/
│   ├── generate_images.py                 ← 🧠 Image generation logic
│   └── inject_images.py                   ← ✍️  Markdown image injection
│
├── scripts/data/
│   ├── character_profiles.json            ← Character descriptions
│   └── image_prompts.json                 ← Default prompt file used by script
│
├── .env                                   ← Set your API key here

---

⚙️ Setup

1. Install dependencies

poetry install

2. Set your DeepAI API key in .env

DEEPAI_API_KEY=your_deepai_key_here

3. Optional: Register CLI shortcuts

[tool.poetry.scripts]
generate-images = "scripts.generate_images:main"
inject-images = "scripts.inject_images:main"

---

🧠 Step 1: Create Prompt JSON File

The file must define a style and a list of chapters. Each chapter contains image prompts with optional character references.

Example (scripts/data/image_prompts.json):

{
  "style": "comic book, colorful, cinematic lighting",
  "chapters": [
    {
      "title": "Chapter 1 – Arrival",
      "prompts": [
        {
          "character": ["Chris", "Das Auto"],
          "prompt": "driving toward a city sign that says 'Welcome to Ohio', excited faces",
          "filename": "holiday_ch1_01_arrival.png"
        },
        {
          "character": "Hotelbesitzer",
          "prompt": "lurking behind the hotel desk, eyes glowing",
          "filename": "holiday_ch1_04_owner.png"
        }
      ]
    }
  ]
}

---

👥 Step 2: Create Character Profiles

Stored in scripts/data/character_profiles.json:

{
  "Chris": "a cartoon boy around 13 years old, clever, casual travel clothes, comic style",
  "Das Auto": "a shiny red convertible with a white racing stripe",
  "Hotelbesitzer": "a creepy cartoon hotel owner with glowing eyes and wild hair, comic horror"
}

---

🎨 Step 3: Generate Images

🧾 Using Defaults (no arguments)

poetry run generate-images

This uses:

• Prompt file: scripts/data/image_prompts.json

• Output directory: assets/illustrations

🧾 With Custom Prompt File or Output Directory

poetry run generate-images \
  --prompt-file assets/prompts/holiday_ch1_prompts.json \
  --output-dir assets/illustrations

🔐 Use a different API key

poetry run generate-images \
  --api-key sk-your-deepai-key

If --api-key is omitted, the script uses the value from .env.

---

✍️ Step 4: Inject Images into Markdown Chapters

poetry run inject-images

🧪 Dry-run

poetry run inject-images --dry-run

Adds image markdown like:

![image](../assets/illustrations/holiday_ch1_01_arrival.png)

---

🧾 Example Summary Output

📄 Chapters processed:    1
🧠 Prompts in total:     5
✅ Images generated:     4
⏭️  Skipped (exists):    1
❌ Errors:               0

---

✅ Tips

• Use consistent filenames like holiday_ch1_01_arrival.png

• Define default prompt styles to unify your visual output

• Separate characters from prompts for modularity

• Use --api-key to override or rotate credentials

• Works great with inject-images and Pandoc/HTML exporters

---