Technical & Liability Disclaimer
Clear boundaries on what client-side metadata cleaners can and cannot remove from digital media.
1. The Boundary: Header Metadata vs. Pixel-Level Watermarks
MetadataRemover is specifically engineered to parse and excise container-level header and chunk metadata. This includes:
- EXIF camera information (shutter speed, ISO, device make/model, serial numbers).
- GPS geotag coordinates (latitude, longitude, altitude, timestamp).
- IPTC and XMP copyright data, creator credits, and editing histories.
- PNG text chunks (
tEXt,zTXt,iTXt) containing AI prompts, seed numbers, and generation workflows. - C2PA Content Credentials manifests embedded in file headers.
2. What Metadata Cleaners CANNOT Remove: Google SynthID & Pixel Watermarks
Certain modern AI generators embed watermarks directly into the pixel raster itself rather than file headers. The most prominent example is Google SynthID (used by Imagen and Gemini).
Metadata cleaners remove container-level headers (EXIF, IPTC, XMP, C2PA) but do not alter raw pixel data or invisible pixel-embedded watermarks such as Google SynthID. Invisible watermarks reside within the visual bitstream rather than container headers. Removing header metadata does NOT remove SynthID. Even severe lossy recompression or Mode B canvas re-encoding cannot guarantee the elimination of pixel-level watermarks.
We believe in absolute technical honesty: any tool claiming to "strip 100% of all AI watermarks" without acknowledging pixel-level steganography is misleading you. When our engine detects potential SynthID or pixel-level signals, it discloses them as immutable / non-removable.
3. Mode A Lossless Stripping vs. Mode B Raster Re-Encoding
Mode A (Lossless Binary Excision) removes header metadata chunks while leaving the compressed pixel scan streams (DCT blocks in JPEG, IDAT in PNG) 100% byte-for-byte identical. This preserves pure original image quality but leaves perceptual hashes intact.
Mode B (Canvas Re-Encoding & Dithering) decodes the image into raw pixels on an in-browser canvas and re-encodes it. While this resets file container structures and alters perceptual hashes (pHash/dHash), it involves lossy re-encoding and can slightly alter image compression.
4. Educational & Privacy Purpose
This service is provided for personal privacy protection and legitimate digital hygiene. It is not intended for circumventing copyright controls or altering evidentiary records. Users assume full responsibility for complying with applicable laws in their jurisdiction.