GenImageText

⚠️ IMPORTANT: This is NOT an image generator. It adds perfect text to images created by your AI tools.

Fix imperfect AI-generated text in images by separating image generation and text overlay.

🌐 English | 简体中文 | 繁體中文 | 日本語 | 한국어

What This Tool Does

AI-generated images often contain garbled or imperfect text, especially for Chinese, Japanese, Korean (CJK) and other non-Latin scripts. This tool solves this problem by separating the workflow:

1. This Skill separates your prompt → image-only prompt + text requirements
2. Your AI Tool generates the clean base image (Midjourney, DALL-E, Stable Diffusion, etc.)
3. This Skill analyzes the image for optimal text placement
4. This Skill renders perfect text with professional typography

Supported AI Image Generators

Use any of these tools for Step 2 (image generation):

Key Point: This skill does NOT generate images. It only adds text to images created by the tools above.

For AI Agents (Natural Language Installation)

Copy and paste this prompt to your LLM agent (Claude Code, Kimi Code, Cursor, etc.):

Install the GenImageText skill to my workspace.
Clone from: https://github.com/stephenlzc/GenImageText
Set up all dependencies and verify the installation by running a test with Chinese text extraction.

Installation

Requirements

• Python 3.8+
• Python packages: pip install Pillow numpy

Git Clone

git clone https://github.com/stephenlzc/GenImageText
cd GenImageText

Usage

Step 1: Separate Prompt (This Skill)

from scripts.prompt_separator import separate_prompt

result = separate_prompt("Movie poster with 'Interstellar' title")
# result['image_prompt']: Clean visual description without text
# result['text_requirements']: Structured text data

Step 2: Generate Base Image (Your AI Tool)

⚠️ This step uses YOUR AI image generator, NOT this skill.

Use the image_prompt with your preferred AI image generator:

• Midjourney - Discord-based generation
• DALL-E 3 (ChatGPT Plus, OpenAI API)
• Stable Diffusion - Local or cloud-based
• Google Gemini/Imagen
• Adobe Firefly
• Microsoft Bing Image Creator (Free)
• Any other AI image tool you prefer

Step 3: Analyze Image (This Skill)

from scripts.image_analyzer import analyze_image, get_text_placement_suggestions

analysis = analyze_image("base_image.png", text_requirements)
placements = get_text_placement_suggestions(analysis, text_requirements)

Step 4: Render Text (This Skill)

from scripts.text_renderer import render_text_on_image

output_path = render_text_on_image(
    image_path="base_image.png",
    output_path="final_image.png",
    placements=placements,
    user_choices={
        "font_style": "modern",
        "effects": ["shadow", "outline"]
    }
)

Font Handling

Fonts are loaded with the following priority:

1. User-provided font path: If specified
2. Skill assets: Check assets/fonts/ directory
3. System fonts: Search common system font directories
4. Fallback: Default PIL font

Font Recommendations by Language

简体中文 (Simplified Chinese)

繁體中文 (Traditional Chinese)

한국어 (Korean)

English / Latin

Download Fonts

You can manually download fonts from Google Fonts or Noto Fonts and place them in assets/fonts/:

• Noto CJK Fonts: https://www.google.com/get/noto/
• Roboto: https://fonts.google.com/specimen/Roboto
• Open Sans: https://fonts.google.com/specimen/Open+Sans

All fonts are free for commercial use under SIL Open Font License or Apache License 2.0.

Project Structure

GenImageText/
├── scripts/                # Python scripts
│   ├── prompt_separator.py
│   ├── image_analyzer.py
│   └── text_renderer.py
├── assets/fonts/           # Fonts directory
└── references/             # Reference materials

License

MIT © stephenlzc

🌍 Languages

• 简体中文 - 简体中文文档
• 繁體中文 - 繁體中文文檔
• 日本語 - 日本語ドキュメント
• 한국어 - 한국어 문서