Just tried the open-source gimi-illustration-skill by teacher Xiaomi. Straight to the point: this tool is incredibly useful.

Feed an article into it, and the AI automatically analyzes the structure, extracts key points, and generates hand-drawn illustrations to match. The style is consistent and not templated.

• Supports custom IP characters for your own visual identity
• Adjustable image dimensions — works in both landscape and portrait
• Can add titles or keep it pure image, adaptable to different scenarios

Reports, guides, tutorials — any structured content can be visualized with one click. The hand-drawn style is much lighter than traditional infographics.

Open-sourced on GitHub, ready to use: https://github.com/GiMi-Xiaomi/gimi-illustration-skill…

If you work with content, give it a try — it’ll save you a lot of time on illustrations.

GiMi-Xiaomi/gimi-illustration-skill

Source: https://github.com/GiMi-Xiaomi/gimi-illustration-skill

gimi-illustration

Input Chinese article content → Output whimsical hand-drawn style illustrations. Open-source AI Agent Skill.

Skill code is under MIT License; Gimi character image is subject to IP-NOTICE.md.

Turn the ideas, processes, and metaphors in your article into readable hand-drawn illustrations. Default 16:9 landscape, supports Gimi IP or no character modes.

Preview

Images are served via jsDelivr (https://www.jsdelivr.com/) CDN to avoid failures when fetching raw images from GitHub in certain regions.

No Character

Practice Loop

AI Practice Loop

Demo Main Line

Demo Main Line — Three Steps

Shareable Link

Portfolio Share Link

Gimi IP

Writing Pitfalls

Writing Pitfalls Checklist

AI Modifying Files

AI File Modification Chaos

Sedentary Neck & Shoulders

Sedentary Neck & Shoulder Fatigue

Who It’s For

Quick Start

1. Installation

Codex (recommended, primary testing environment)

git clone https://github.com/GiMi-Xiaomi/gimi-illustration-skill.git
cd gimi-illustration-skill
mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
cp -R . "${CODEX_HOME:-$HOME/.codex}/skills/gimi-illustration"

Cursor (also works)

mkdir -p ~/.cursor/skills
ln -s "$(pwd)" ~/.cursor/skills/gimi-illustration

Or place inside project: .cursor/skills/gimi-illustration/

2. Usage

Send a trigger phrase + article body in conversation:

illustration

[paste your article body]

Trigger words: 配图 / 给文章配图 / 帮我配图 / illustration

The AI will automatically follow the SKILL.md workflow: understand the text → plan the illustration strategy → assemble prompts → call the platform’s image generation → quality assurance → save to outputs/.

3. Common options (can be specified in the text)

Output Structure

outputs/{YYYYMMDD}-{slug}/
  shot-config.md    # illustration strategy (delivered with images)
  01-{slug}.png
  02-{slug}.png     # continues numbering for multiple images

Directory Structure

Image Generation Environment

The Skill handles strategy and prompts; image generation is performed by the current AI platform’s built-in tools:

For Gimi IP mode, the reference images and calibration images must be placed in context before generation. See references/ip/gimi/ip.md.

License

• Skill code and documentation structure: MIT License
• Gimi character image (girl in a hat): See IP-NOTICE.md, not covered under the MIT license

About the Author

Gimi (米未可) — UX Designer / AI Builder

Using Skill + Vibe Coding to improve work efficiency and illustrate content.

• Xiaohongshu: 米未可 (https://www.xiaohongshu.com/user/profile/593a3b5033594a18f9aea21f)

Gimi 米未可 (@Gimi_min): I made a Skill for hand-drawn style illustrations~ Its core function: you provide the article or voice-over text, and it automatically analyzes the content, extracts key points, and generates consistent illustrations. Key features: • Automatically generates images based on content, with consistent style across the set • You can choose whether to show the IP character, image size, and whether to include a title • Suitable for reports, travel guides, tutorial explanations, and many other scenarios github