deai-image
por swaylq
Git & GitHubResumen
Detect and remove AI fingerprints from AI-generated images.
Descripción
Qué hace esta skill
CLI tool that applies a 7-stage pipeline to AI-generated images to reduce their detectability by AI image detectors. Strips EXIF and C2PA metadata, adds synthetic camera noise, adjusts color distribution, and recompresses the file. Works with output from Midjourney, DALL-E, Stable Diffusion, Flux, and other generators.
Combines metadata stripping, frequency pattern disruption, noise injection, and recompression in a single command rather than running separate tools for each step.
Cuándo usarla
- Stripping C2PA metadata from DALL-E exports before sharing
- Batch cleaning a folder of Midjourney images for a client portfolio
- Testing how well different processing strengths bypass Hive or Illuminarty detectors
- Removing generation metadata from AI art before uploading to stock sites
- Converting AI-generated product shots to reduce detection flags on e-commerce platforms
Flujo de trabajo de ejemplo
User asks: Remove AI fingerprints from my Midjourney portrait at medium strength
- 1
Run deai.py on the input file with default medium strength
- 2
Strip EXIF and C2PA metadata via ExifTool
- 3
Add Poisson/Gaussian grain and adjust contrast and saturation
- 4
Perform resize cycle and JPEG recompression to introduce natural artifacts
- 5
Run a final metadata clean pass and write the output file
Processed JPEG with 50-65% bypass rate against common AI detectors and metadata fully removed
Requisitos
- ImageMagick 7.0+
- ExifTool
Instalación
clawhub install swaylq/deai-imageEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.