AI Metadata Cleaner is a free in-browser metadata and AI provenance cleaner for images, video, PDF, and RAW files. It inspects and strips EXIF, GPS, XMP, IPTC, C2PA Content Credentials, and AI generator tags from Stable Diffusion, Midjourney, GPT-Image-2, and Veo 3. The work runs locally in the browser via WebAssembly, with zero file uploads. Pricing is 100 percent free, with no sign-up required, and processing stays on the device. What the cleaner covers AI Metadata Cleaner is built for AI artists, photographers, and privacy-conscious creators. It detects and strips EXIF camera data, GPS location tags, XMP and IPTC blocks, Stable Diffusion and ComfyUI generation prompt chunks, and C2PA or JUMBF container markers. Files are handled in the browser. They are not uploaded to a server. Supported formats Image files include JPEG, PNG, WebP, AVIF, and HEIC. Camera RAW photos are read with LibRaw WebAssembly and include DNG, CR2, NEF, and ARW. Video files include MP4, MOV, and WebM, processed with FFmpeg.wasm. PDF documents are processed with pdf-lib so author fields, XMP, and document properties can be cleaned. Features you can use locally A read-only provenance inspector shows EXIF, GPS, XMP, and C2PA markers without changing the file. Batch processing accepts up to 20 files and offers one-click ZIP export. An optional HTML5 Canvas re-encode is available when a fresh image container is preferable to an in-place metadata edit. The same pipeline covers a single file or a short batch. Camera RAW workflow Photographers often need camera EXIF and location tags removed before a RAW file is shared, archived, or sent to a client. [Browser-Based Camera RAW (CR2, CR3, NEF, ARW, DNG) EXIF & MakerNotes Cleaner](https://aimetadatacleaner.pro/raw-metadata-cleaner) is the local page for that job. LibRaw WebAssembly parses the RAW container in the browser, so GPS coordinates and camera blocks can be inspected and stripped without sending the photo anywhere. Honest boundaries The cleaner strips container-level metadata and C2PA/JUMBF markers locally. It does not remove invisible frequency-domain pixel watermarks such as Google SynthID. It does not guarantee how third-party social platforms apply AI labels. Those limits are part of the product, not a hidden extra step. Why a local clean matters Hidden metadata can include GPS coordinates, camera details, generator prompts, and content-credential manifests. Stripping those blocks before a file is published reduces what a later viewer or platform can read from the container. Because the cleaner never uploads the file, the original stays on the computer that opened it. That model fits private client work, AI art exports, and ordinary photos that should not advertise where they were taken. The inspector is read-only. Batch mode exports a ZIP of cleaned files. Social platforms may still apply their own AI labels after the container is cleaned, and frequency-domain pixel watermarks such as Google SynthID remain in the image.