What gets removed
- Content Credentials (C2PA), the signed record that an image came from ChatGPT, Firefly, a camera, etc.
- EXIF: camera and phone model, dates, software, serial numbers, and GPS location
- AI prompt data (Stable Diffusion, ComfyUI, Midjourney) and “AI-generated” tags
- XMP and Photoshop data, comments, embedded thumbnails, and bytes hidden after the end of the image
The picture itself is never re-encoded: the compressed image data is copied byte for byte, so quality and pixels stay identical.
What it can’t do
Invisible watermarks woven into the pixels themselves (for example Google SynthID) are not touched by this mode. Records an AI service keeps on its own servers can’t be removed from here either. HEIC, WebP and other formats aren’t supported yet.
Safe mode, ported from image-fingerprint-remover (MIT). Works offline.