How AI Content Remover Works

When you select a file, client-side metadata is ignored. The server inspects the binary magic bytes (such as JPEG FF D8 FF, PNG 89 50 4E 47, or MP4 ftyp box) to verify genuine file formats before any processing begins. Malicious scripts, executables, and path traversal attempts are rejected immediately.

Files are saved into isolated private directories indexed strictly by randomized UUIDs (e.g., storage/app/private/media/{uuid}/). Files are never placed inside the public web root and cannot be browsed or indexed by search engine spiders.

For images, ImageMagick 7 (magick identify -verbose) or the PHP Imagick extension extracts EXIF, GPS coordinates, IPTC records, XMP XML packets, and ICC color profile metadata. For videos, FFprobe executes with structured JSON output arguments to inspect audio/video stream parameters, container atoms, and encoder credits.

When you request a cleaned copy:
• Images: Auto-orient transformation is performed to preserve natural visual orientation, followed by a surgical -strip command to remove headers.
• Videos: FFmpeg runs in stream-copy mode (-c copy -map_metadata -1 -map_chapters -1). Visual pixels and audio samples are passed through bit-for-bit with zero re-encoding or compression artifacts.

Before delivering the file, our verification service re-analyzes the cleaned copy and compares it against the original inspection report. We confirm which removable fields were stripped and verify file size differential and dimension parity.

We do not retain user files. A Laravel scheduled worker runs continuously to permanently delete files and directories that exceed the 60-minute retention limit.