Strips C2PA / Content Credentials, EXIF, XMP, IPTC, GPS coordinates and PNG text chunks before you publish. The default mode is lossless: the pixels stay byte for byte, only the container's service blocks are cut out.
Your files never leave your device. Everything runs inside your browser: no file is uploaded to a server.
Drop files hereor click to choose · Ctrl+V pastes from the clipboard
JPEG · PNG · WebP · GIF · TIFF · AVIF · HEIC · SVG · BMP · ICO
Re-encoding removes everything for certain, including exotic blocks, but it compresses the picture again and loses some quality. For JPEG it is usually unnecessary.
“Lossless” means the compressed image data is copied as is and only the container structure changes. The file hash changes, the pixels do not.
| Format | What is removed | Method |
|---|---|---|
| JPEG / JPG | EXIF (including GPS and the camera serial number), XMP, IPTC/Photoshop, APP11 JUMBF (C2PA), MPF, COM comments | lossless |
| PNG / APNG | eXIf, tEXt / zTXt / iTXt (where Stable Diffusion writes prompts), caBX (C2PA), tIME, foreign chunks | lossless |
| WebP | EXIF, XMP, non-standard chunks; the VP8X flags are rewritten correctly | lossless |
| GIF | Comments, XMP, application extensions except the NETSCAPE loop | lossless |
| TIFF | EXIF/GPS/IPTC/XMP/Photoshop tags; the IFD is rebuilt, strips and tiles are copied as is | lossless, beta |
| AVIF / HEIC / HEIF | Exif, XMP and C2PA items are zeroed in place so the offsets stay valid | in place, beta |
| SVG | <metadata>, RDF / Dublin Core, Inkscape and Adobe traces, comments, optionally scripts | XML rewrite |
| BMP / ICO | These containers hold no metadata; the file is checked and returned unchanged | check |
| RAW (CR2, NEF, ARW, DNG…) | Inspection only. Cleaning a RAW would destroy it — convert to JPEG or TIFF in an editor first | read only |
C2PA (Content Credentials) is a cryptographically signed manifest describing where a file came from: what created it, what edited it, whether a generative model was involved. Photoshop, Firefly, DALL·E and Leica and Sony cameras all write it. Platforms such as Instagram, LinkedIn and TikTok read that manifest and attach a “Made with AI” label. In JPEG it lives in the APP11 (JUMBF) segment, in PNG in the caBX chunk, in AVIF/HEIC as a separate item inside the meta box. Proyavka removes all three.
Orientation is often stored not in the pixels but in the EXIF Orientation tag: the sensor writes the frame as is and the tag says “rotate by 90°”. Remove EXIF and the rotation goes with it. That is why “bake rotation” is on by default: such files are re-encoded with the rotation already applied. Everything else stays lossless.
ICC describes the colour space and contains nothing personal. Without it sRGB images barely change, but Display P3 or Adobe RGB visibly wash out or oversaturate. Keep it unless you are fighting for every kilobyte. That said, an ICC profile can technically carry a custom description string, so untick the box if you are being careful.
The dangerous parts are GPS coordinates (the spot where you stood, accurate to a building), the camera and lens serial numbers (they link every photo of yours on the internet together), the owner name, the date and time, the software and OS version, and for images from neural networks the full prompt text inside PNG chunks. Proyavka shows all of it in the “what was inside” block before removal, so you can see what was leaking.
Visible watermarks and invisible pixel-level marks (Google's SynthID, statistical watermarks) live in the pixels themselves — they survive any container cleaning and most re-encoding. The file name and the creation time on disk are metadata too: switch on “anonymise file names”, and dates inside the ZIP archive are zeroed anyway. Remember that the platform will stamp its own marks on upload regardless.
In lossless mode the compressed data is copied byte for byte: the size difference equals exactly the volume of removed metadata, sometimes megabytes of previews and profiles. Quality does not change physically. In “re-encode” mode the picture is compressed again and quality depends on the slider.