Inspect, clean, inspect again
Selected files are read into a browser worker. The parser checks their structure and lists recognized metadata blocks. Cleaning creates a separate byte buffer, which is then parsed again. Your original file is never overwritten.
Before a download is offered, the app compares SHA-256 hashes of the encoded media payloads before and after the edit. If they differ, processing fails rather than offering an unverified output. You can download the report as JSON.
JPEG, PNG and WebP behavior
JPEG: remove application metadata and comments, keep ICC and Adobe color interpretation, preserve compressed scans, and retain a minimal EXIF orientation record when required. Standard baseline and progressive JPEGs are supported. MPF multi-picture/HDR gain-map files are rejected.
PNG: validate chunk checksums, remove text, credential and non-display ancillary chunks, retain image data, transparency, recognized color and animation chunks. Unknown critical chunks are rejected.
WebP: remove non-display metadata chunks, update RIFF length and EXIF/XMP flags, keep image, alpha, animation and color chunks. Orientation is retained when needed. Unknown structures are not silently converted.
Two image modes, one clear trade-off
Privacy clean removes original EXIF, including potential GPS and camera information. It keeps a minimal orientation record if needed. Keep camera EXIF retains the complete EXIF block, including any private or AI-related fields inside it.
Both modes remove XMP, IPTC and text containers; credits stored there are removed too. Neither mode invents camera details or changes the picture to disguise its origin. Save the original when its provenance or attribution matters.
Video scope and verification limits
MP4/MOV: recognized C2PA/XMP UUID, XMP_, user-data, metadata and supported software-information boxes are replaced by same-size empty free boxes. Media offsets, audio and video data stay in place. Unrecognized UUIDs are retained with a warning.
Video cleanup is not exhaustive anonymization: timed metadata, track timestamps and unsupported fields may remain. For videos, the SHA-256 comparison covers mdat payloads, not a full semantic decode of every track. Play the downloaded file in a compatible player before publishing.
The parser validates supported container structure, not the full correctness of every codec bitstream. A structurally valid but already damaged media stream can still fail to display. The app does not validate C2PA signatures, detect AI origin, remove pixel watermarks or test a social platform’s classification.
Limits and supported browsers
A batch accepts up to 50 files, with a 100 MB limit per file and 250 MB total. Processing is sequential to limit peak memory use. Large files and ZIP creation need available device memory.
Use a current browser with Web Workers, typed arrays and Web Crypto over HTTPS. File selection and individual downloads work without an account. Video preview and playback depend on the original codec and your player. HEIC, AVIF, GIF, RAW, PDF, MKV and WebM are not supported.