AI Metadata Remover

Every hidden tag that says "made with AI"

Modern image files carry far more than pixels. Generators, editors, phones and cameras all stash data in side blocks of the file. This tool parses the container, keeps only what the decoder needs to draw the picture, and drops everything else.

Three steps, zero uploads

1. Drop, paste or browsePick one image or fifty. The file is read into your browser's memory with the standard File API — the page never contacts a server.
2. The container is rewrittenThe tool walks the JPEG segments, PNG chunks or WebP RIFF chunks, copies the compressed pixel data untouched, and skips every metadata block. It also runs a signature scan so the report can name the generator it found.
3. Download the clean fileYou get a new file with -clean added to the name, usually a little smaller, visually identical. Reload the page and everything is gone from memory.

Which AI tools tag their images

A quick map of where each generator leaves its fingerprint and whether a metadata remover can deal with it. Pixel watermarks are listed honestly — no metadata tool removes those.

What platforms read

Instagram, Facebook and Threads decide on the “AI info” label mostly from C2PA and IPTC metadata. LinkedIn shows a Content Credentials icon when it finds a C2PA manifest. TikTok and Pinterest auto-label uploads carrying Content Credentials. YouTube reads C2PA to show a “captured with a camera” badge on eligible videos and separately requires creators to disclose realistic altered media. All of those systems look at the file's metadata first — which is exactly what this tool rewrites.

Built by an editor who ships images every day

I edit YouTube videos for geography, history, economics and true-crime channels. A normal week means dozens of thumbnails, B-roll stills and map textures going through Photoshop, and more and more of them touch a generative tool somewhere — a background extended with generative fill, a stock photo upscaled, a texture cleaned up. Then a client posts a completely real photo of themselves to Instagram and it gets an “AI info” badge because the sky was extended by a few pixels.

The tools I found for this were either paid, capped at three images a day, or — worse — uploaded the file to someone's server first. Thumbnail files for unreleased videos are confidential. So I wrote a small parser that runs in the browser, and put it here for anyone who needs it.

Good reasons to strip metadata

• Privacy: phone photos carry exact GPS coordinates, device serial numbers and timestamps. Strip them before posting a house, a kid or a workplace.
• Portfolio hygiene: your Stable Diffusion prompt, seed and model hash don't belong in a client deliverable or a print-on-demand upload.
• False-positive labels: a real photograph retouched with a generative tool shouldn't be presented as fully synthetic.
• Smaller files: a single Content Credentials manifest can be 30–100 KB. Stripping it speeds up every page the image sits on.
• Sanitised screenshots and assets: remove editor history and comment blocks before sharing files with third parties.

Ariyan Khan

Freelance video editor working remotely with YouTube creators, coaches and founders. I build the tools I need for my own editing workflow and share the useful ones here. More about me → · Also free: Map Maker for highlighting countries.

What a metadata remover cannot do

• It cannot remove pixel watermarks. Google SynthID, Meta's invisible marker and similar techniques are embedded in the image content itself. They survive metadata stripping, cropping and mild compression by design. This tool does not attempt to defeat them.
• It cannot fool pixel-based AI detectors. Classifiers that judge texture, noise and frequency patterns don't read tags at all.
• It cannot control platform policy. Meta, TikTok, LinkedIn and YouTube can add new signals, run their own detection, or ask you directly. Treat the result as “metadata-clean”, not “undetectable”.
• It cannot un-sign a video. This page handles still images only. C2PA in MP4 is a different container and a different tool.

Use it responsibly

Metadata is yours to remove; the obligation to be honest about synthetic media isn't removed with it. YouTube's altered-content disclosure, Meta's policy on realistic AI content, and the EU AI Act's transparency rules apply regardless of what the file's tags say. Clean your files for privacy, quality and accuracy — and disclose AI-generated content where the platform or the law asks you to. I wrote more about where AI actually fits into editing work in Will AI replace video editors?