Frequently asked questions

It reads the metadata inside an image and removes what you choose. That covers EXIF including GPS, XMP, IPTC, C2PA content credentials and PNG text chunks. It shows you the real field names and values before anything is removed, and verifies the result afterwards.

Yes, with no account, no sign up and no quota. The work runs on your own machine rather than on a server, so there is nothing for us to meter or charge for.

No. It runs in your browser. After your first visit the page is cached, so it also works offline.

JPEG, PNG and WebP, up to 25 MB per file and 30 files per batch. AVIF and HEIC are not supported yet because they use a different container structure that needs its own parser.

Thirty per batch, and as many batches as you like. Each file is listed separately with its own findings so you can see which ones carried markers.

No. Parsing and stripping run in a Web Worker in your browser. There is no upload endpoint that could receive them. Turn off your network and the tool keeps working, which is the simplest way to verify that for yourself.

We never receive them, so there is nothing to store. We could not produce your images if asked, because they never leave your device.

No advertising or analytics cookies. The only thing stored in your browser is your light or dark theme preference, which stays on your device.

Yes, after one online visit. A service worker caches the page and its code, and the engine makes no network calls, so everything keeps working with the network off.

It is verifiable. Disconnect from the internet, reload, and clean a file. If anything were being uploaded, that would fail.

No. Metadata sits in length-prefixed blocks beside the compressed pixel data, so removal is a copy-and-skip operation. The compressed image is never decoded or re-encoded, and the output is byte-for-byte identical in its pixel data.

Because decoding and re-saving is far easier to build. We measured a leading competitor on one pass: 52.3 percent of pixels altered, PSNR 39.78 dB, maximum per-channel difference of 46 out of 255, and a PNG 15.8 percent larger than the input, while the page claimed the pixels were unchanged.

Compare the quantization tables, the DQT segments, between input and output. If they differ, the file was re-encoded regardless of what the marketing says.

Three things: that the pixel payload is byte-for-byte identical to the input, that the output is no larger than the input, and that zero removable metadata fields remain on a fresh re-scan. All three must pass before you see "verified clean".

Because metadata is usually a tiny fraction of a photo. A few hundred bytes to a few kilobytes against several megabytes of pixels. A large reduction would mean something re-compressed your image.

Because PNG text chunks can be genuinely large. A ComfyUI workflow stored in a PNG runs to tens of kilobytes. That is real data being removed, not compression.

Almost always because an AI-assisted edit wrote provenance into the file. Generative Fill, AI background removal, AI sky replacement and Lightroom AI Denoise all do this. The pixels are yours; the marker came from your editor, and the platform is reading the marker.

It removes the file-level markers that most commonly trigger it, C2PA and XMP, which handles the large majority of cases. It cannot help if a platform is using a visual classifier or a pixel watermark, because those are not metadata.

Not in place. Platforms apply the label at upload and do not re-evaluate afterwards. Delete the post, clean the file, and post it again.

No. When a label comes from metadata, there is no toggle that removes it. The only reliable fix is to clean the file before uploading.

Because platforms read every slide separately. Cleaning only the first image is one of the most common reasons a label persists after someone believes they have fixed it.

It removes metadata, because a screenshot is a new file built from pixels on screen. But you lose resolution, gain your display colour profile, and may pick up device metadata. It also does nothing about pixel watermarks, which are designed to survive exactly that.

No. SynthID lives in the pixel values, and this tool deliberately never alters pixels. No metadata tool can remove it. Anything claiming otherwise is either degrading your image and hoping, or overstating what it does.

We can tell you whether the file declares itself as AI-generated, which is a different and more honest question. A generated image stripped of metadata reads clean here, and a real photograph edited with Generative Fill reads as AI-involved.

Not yet. We detect that a manifest is present and remove it, but we do not verify who signed it or whether the signature is valid. Signature validation is planned, and we would rather say so than imply a check we do not perform.

No. MP4 and MOV use a different container and can carry their own C2PA. This tool handles JPEG, PNG and WebP only, and we will not claim a video is clean when we cannot read it.

No. PDF metadata lives in a document information dictionary and an XMP stream inside an entirely different file format. It needs its own parser, which we have not built.

HEIC and AVIF use the ISOBMFF box structure rather than JPEG segments or PNG chunks, so they need a separate implementation. Rather than convert your HEIC to PNG behind your back, which some tools do, we reject the format until we can handle it properly.

Everything removes all removable metadata: EXIF, GPS, XMP, C2PA, IPTC and PNG text. AI markers only removes C2PA, XMP and PNG text, and removes EXIF only when it carries a generator fingerprint, so camera settings and copyright survive.

Yes, and that is important to know. GPS lives inside the EXIF block that AI-only mode preserves. If privacy rather than labelling is your goal, use Everything.

Not as a single switch today, because both live in the same EXIF block. Everything removes both, AI-only keeps both. A GPS-only option is a fair request and is on the list.

Usually yes, which is why it is enabled by default. The ICC profile tells a browser how to render your colours, and removing it can visibly shift them, particularly on wide-gamut displays.

Use AI markers only. Creator, credit and copyright fields in EXIF and IPTC are preserved, while C2PA, XMP generation tags and AI-fingerprinted EXIF are removed.

From files you own, ordinarily yes. Stripping GPS and device identifiers before publishing is routine privacy practice and predates AI labelling by two decades.

No. The EU AI Act places transparency obligations on providers and deployers of generative systems, and many platforms require disclosure in their terms. Those duties attach to how you publish, not to what your file contains.

You should not. Removing copyright management information to conceal ownership carries specific liability in several jurisdictions, including under section 1202 of the US DMCA. That is separate from any infringement claim.

Removing metadata from your own files is ordinary practice, and platforms strip metadata themselves on upload. The risk is not in cleaning a file; it is in misrepresenting AI content where disclosure is required.