A comprehensive CLI toolset designed to strip AI detection patterns and metadata from images generated by models like Midjourney and DALL-E.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install humanize-image
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install humanize-image using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is humanize-image?
The humanize-image skill provides a robust pipeline for de-fingerprinting AI-generated content. It addresses the growing need for privacy and authenticity by transforming detectable synthetic patterns into human-camera-like textures. By leveraging powerful Openclaw Skills, developers can automate the removal of EXIF tags, C2PA watermarks, and frequency domain patterns that identify images as non-human.
This skill is essential for creators and researchers who need to ensure their visual assets are treated as traditional photography by automated systems. It uses a multi-stage transformation process including grain injection and color adjustment to break down the mathematical signatures left by diffusion models. Whether you are working with Midjourney, DALL-E 3, or Stable Diffusion, this tool provides the necessary layers to reduce detection probability significantly.
humanize-image Use Cases
- Removing AI-generated metadata and C2PA watermarks for privacy and security.
- Bypassing AI detectors like Hive Moderation, Illuminarty, and AI or Not for research purposes.
- Batch processing large image galleries to normalize visual textures and noise distribution.
- Preparing synthetic visual assets for platforms with strict automated filtering or content labeling.
- Stress testing AI detection models by identifying successful bypass patterns.
How humanize-image Works
- The skill analyzes the input image to identify detection vectors such as EXIF tags and DCT coefficient patterns.
- It initiates a metadata stripping process using ExifTool to remove all traceable origin tags and C2PA watermarks.
- A processing pipeline applies grain addition (Poisson/Gaussian noise) to simulate authentic camera sensor noise.
- Color adjustments and edge disruptions (blur/sharpening) are applied to break down over-smoothed diffusion patterns.
- The image undergoes a resize cycle and JPEG recompression to introduce natural resampling artifacts.
- A final cleanup pass ensures no metadata leakage occurs before the humanized output is generated.
humanize-image Setup
To utilize this skill within your Openclaw Skills environment, ensure you have the core system dependencies installed.
# Install system dependencies on Debian/Ubuntu
sudo apt update
sudo apt install -y imagemagick libimage-exiftool-perl python3 python3-pip
# Install Python libraries
pip3 install Pillow numpy
# Verify your installation
bash scripts/check_deps.sh
humanize-image Data Schema & Taxonomy
The humanize-image skill organizes data through a transformation pipeline that modifies the visual and binary structure of image files.
| Component | Purpose | Data Handling |
|---|---|---|
| Input Buffer | Temporary storage for raw AI images | Supports PNG, JPG, WEBP, and more |
| Metadata Layer | Extraction and removal | Strips EXIF, XMP, JUMBF, and C2PA tags |
| Pixel Array | Noise and grain injection | Modifies pixel distribution via NumPy |
| Output File | Final humanized image | Defaults to JPEG at 75-95 quality |
| Batch Queue | Recursive directory processing | Handles folder structures via the --batch flag |
humanize-image Advanced Features
- Multi-strength processing levels (light, medium, heavy) to balance image quality versus bypass success rates.
- Pure Bash implementation for lightweight environments where Python and Pillow are unavailable.
- Support for custom processing pipelines, allowing users to tweak noise variance and contrast levels manually.
- Automated batch mode for processing entire collections of Openclaw Skills generated assets.
- Integration with verification workflows to test images against multiple online detectors automatically.
SKILL.md
Loading
Related Openclaw Skills
AI Image De-Fingerprinting Skill
A comprehensive utility for removing AI-generated signatures, metadata, and pixel patterns to make synthetic images indistinguishable from camera photography.
Browser Automation Ultra
A high-performance framework to convert expensive browser-tool interactions into reusable, zero-token Playwright scripts with human-like behavior.
Trip Search
An intelligent travel planning agent for finding real-time transport and accommodation options while managing trip expenses.
Trip Save
A specialized AI skill for capturing, auto-tagging, and storing travel destinations from digital content and text commands.
lambda-lang: Native Agent-to-Agent Language
A high-efficiency, machine-native language designed for direct agent-to-agent communication without human linguistic overhead.
Moltbook
A comprehensive integration allowing AI agents to actively participate in the Moltbook social ecosystem through posts, comments, and community engagement.
Featured*
