Image Metadata Remover

The Image Metadata Remover removes readable metadata from image files, including EXIF, XMP, C2PA-readable markers, camera details, and software tags. It processes images locally in your browser, cleaning metadata-level information while keeping the visible image pixels intact. This tool does not remove invisible watermarks such as SynthID. Your files are not uploaded to any server, making it useful before publishing, sharing, or archiving images.

How to Use

1. Upload the PNG, JPG, or WebP image you want to clean
2. The tool parses and removes EXIF, XMP, C2PA-readable metadata locally in your browser
3. Download the cleaned image for sharing, publishing, or storage

Why Choose This Tool

• Local processing — your images stay on your device, useful for private or unpublished assets
• Readable metadata removal — removes EXIF, XMP, C2PA-readable markers, camera data, and software tags
• Simple workflow — no install required, upload an image and export a cleaner version quickly

It reads the file before it changes anything

The first pass scans the raw bytes and reports what is actually in your file, so you are not cleaning blind. It looks for three things: the JPEG APP1 marker (byte pair FF E1) that carries EXIF, the literal Adobe XMP namespace string http://ns.adobe.com/xap/, and the strings c2pa or contentcredentials that appear in a C2PA provenance manifest. It also records the file size and the pixel dimensions.

The same scan is then run against the finished file. Removal is checked with the same detector that found the data in the first place, rather than being assumed.

Cleaning is a canvas re-encode, not a field-by-field strip

The image is decoded, drawn once into an HTML canvas at its exact original dimensions, and then re-encoded into a brand-new file. A canvas holds nothing but an RGBA pixel array — it has no slot for a GPS coordinate, a camera serial or a signed manifest. Everything outside the pixel grid is gone by construction.

This matters because the usual alternative is a blacklist. A field-by-field stripper has to know every container it might meet: APP1 for EXIF, APP13 for IPTC, APP2 for the ICC profile, APP11 JUMBF for C2PA, and in PNG the tEXt, iTXt, zTXt, eXIf and vendor-specific chunks. Miss one — or meet a segment type invented after the image metadata remover was written — and it stays in the file. A re-encode is subtractive: it can only carry forward what a pixel array can hold, so a segment nobody has named yet is dropped along with the ones that are.

Orientation, transparency and the output file

A JPEG shot in portrait usually stores upright pixels plus an EXIF Orientation tag telling the viewer to rotate. Removing that tag naively is how photos end up sideways. Current Chrome, Safari and Firefox apply the orientation while decoding, so the canvas receives the already-rotated pixels and the output stands upright with no tag needed.

PNG transparency is preserved — the canvas keeps the alpha channel and PNG is re-encoded as PNG. The output keeps your original format and dimensions, and downloads with _cleaned appended to the filename so you never overwrite the master by accident.

Selling something online without publishing your front door

A photo taken in your living room carries GPS coordinates accurate to a few metres. Paste them into any map and you have the seller's address, sitting next to the words "cash on collection".

Facebook Marketplace strips EXIF from the public listing image, but eBay, Craigslist, Gumtree, classifieds forums and any listing you email or message directly do not reliably do so. Burglaries traced back to home addresses recovered from listing photos are a documented pattern, not a hypothetical. Clean the file before it leaves your machine and the platform's behaviour stops being your problem.

A journalist or researcher handling a source photo

Three fields do the damage here: GPS, the capture timestamp accurate to the second, and the camera body serial number. Together they place a source at a specific place at a specific minute.

The serial number is the one people forget. It links an anonymous photo to any other photo from the same body — including one already published under a byline. Strip GPS from one file and not another and the serial still ties them together. Note that the transport matters as much as the file: Signal, WhatsApp and Telegram all strip metadata when you send through the photo picker, and all three pass the original file through untouched when you send it as a document or "send as file", which is exactly what people choose when they want full quality.

Posting a photo of your child

The risk is not the single photo, it is the set. A handful of geotagged images taken over a few weeks reconstructs a routine: the house, the school gate, the Saturday playground, and the hours when nobody is home. Software reads that out of a folder in seconds.

Public feeds on Instagram, Facebook, X and TikTok strip location server-side. Group chats that forward original files, cloud-storage share links, school portals, family blogs and email attachments do not. Clean the file once and it is safe on every one of those paths.

A photographer sending work to a client or a portfolio site

A finished export typically carries the body serial, the exact lens, the shutter and ISO, an edit history in XMP, and a software tag naming your retouching stack and its version. Some of that you may be happy to publish; the serial and the client-identifying IPTC fields usually not.

One workflow detail: clean at the very end, from the master, and prefer PNG output when fidelity is critical, because PNG is re-encoded with no quality parameter while JPEG and WebP go through a 0.95 re-compression. Never make the cleaned copy your archive file — the metadata is unrecoverable once it is gone.

Windows, macOS, iPhone or Android — and how the built-in options compare

There is no Windows build and no Mac build because there is nothing to build. This runs inside the browser tab you already have open, so the same page behaves identically on Windows 10 and 11, macOS, Linux, ChromeOS, iPhone and Android. Nothing to download, no installer, no admin rights — which is the deciding factor on a managed work laptop where policy blocks software installation and a desktop EXIF tool is simply not an option.

Windows does ship a free option worth knowing: right-click a photo, then Properties, Details, "Remove Properties and Personal Information". It clears the fields Explorer and ordinary EXIF readers display. What it is not is a rebuild — security researcher Didier Stevens showed in 2022 that Explorer deletes the EXIF directory entries that point at each value while leaving the value strings themselves in the JPEG APP1 segment, so strings or a hex editor still recovers them. He reproduced it on fully patched Windows 7, Windows 10 21H2 and Windows 11. The dialog also only lists properties the Windows property system recognises, so XMP blocks and camera maker notes are never offered.

macOS handles location well and only location: open the file in Preview, Tools then Show Inspector (Cmd+I), pick the "i" More Info tab, and if a GPS sub-tab is there, "Remove Location Info" rewrites the file in place. There is no remove-all button, so camera model, serial, timestamps and maker notes stay. exiftool -all= photo.jpg is the thorough answer and genuinely the best tool for the job, but it is a command line and an install via Homebrew. On iOS 13 and later the share sheet has Options at the top with a Location toggle; switching it off strips GPS from that one share and nothing else, and it resets next time. Google Photos on Android is narrower still — "Remove geo location in items shared by link" is off by default and applies only to link shares.

So: if you just want GPS off one photo before you text it, use the toggle your phone already has. Come here when you want every non-pixel segment gone rather than a subset, when you want to see what the file contained before and after, or when installing anything is off the table.

JPEG and WebP are re-compressed at quality 0.95

This is real, if small, generation loss. A canvas re-encode cannot preserve the original JPEG coefficients, so the output is a fresh lossy encode at quality 0.95 — one generation of loss, invisible on a screen, measurable in a pixel-difference test. PNG has no quality parameter and comes back with identical pixel values.

The practical rule: clean once, from your master file. Running the same JPEG through repeatedly stacks a fresh generation of loss each time. If you need a strictly untouched image, keep the original and treat the cleaned copy as a publishing artefact.

The embedded colour profile does not survive

The ICC profile is metadata and goes with the rest. That is a genuine trade-off of the canvas approach, not a bug.

It matters for wide-gamut files. A 2D canvas works in sRGB by default, so a Display P3 original — which is what an iPhone shoots by default — or an Adobe RGB export from Lightroom is converted into sRGB on the way through. Saturated reds and greens that live outside sRGB have nowhere to go and clip, which reads as slightly duller colour. If the file was already sRGB, and most web images are, nothing visible changes. For print or soft-proofing work, keep the profiled master and clean only the copy you are publishing.

It does not remove SynthID or any invisible watermark

SynthID lives in the pixel values themselves, and this metadata remover copies pixels faithfully. The watermark survives on purpose.

Google designed it to survive exactly the kind of handling a canvas re-encode does — and considerably worse. Independent testing has found it still detectable after hundreds of successive lossy re-compressions and after a screenshot, because the signal transfers with the pixels into the new file. Removing metadata removes the label, not the signal: an AI image with its C2PA manifest gone is not an undetectable image, and this page will not pretend otherwise.

Metadata removal is not anonymity

A clean file still identifies you in four other ways, and no image tool can fix them.

The picture itself: a house number, a street sign, a reflection in a window, the view from a balcony. The filename: IMG_4821.jpg says little, flat-viewing-oakwood-road.jpg says a lot — rename before you send. The channel: the account, the email headers, the IP the upload came from. And the platform: services that strip metadata for viewers generally read and keep it server-side first, so stripping on their end protects you from other users, not from them. Treat this image metadata remover as one control among several, not a cloak.

My cleaned JPEG is bigger than the original

Expected, and not a fault. Quality 0.95 is higher than most cameras and export presets use — a typical camera JPEG or Lightroom export sits nearer 0.85. Re-encoding at a higher setting writes more data than the removed metadata saved, so the file can grow even though information was taken out. If file size matters, run a separate compression pass after cleaning.

Colours look slightly duller after cleaning

The original was wide-gamut and its ICC profile did not survive the re-encode. iPhone photos are Display P3 by default; Lightroom and Capture One often export Adobe RGB. The canvas works in sRGB, so out-of-gamut saturated colour clips. Fix it upstream: export an sRGB copy from your editor first, then clean that, and the shift disappears because there was no wider gamut to lose.

It reports no EXIF, XMP or C2PA found

Usually correct. Screenshots, images already round-tripped through a social platform, and most AI-generated PNGs genuinely carry none of the three. The detector looks for the JPEG APP1 marker, the Adobe XMP namespace string, and the literal c2pa / contentcredentials strings, so it names those three specifically rather than every field in the file. Cleaning is still worth doing: the canvas re-encode rebuilds the file regardless, dropping segments the scanner never looked for.

Instagram already strips EXIF — do I need this?

For a public post on Instagram, Facebook, X, LinkedIn or TikTok, no. For everything else, yes. Email attachments, Google Drive and Dropbox share links, forum uploads, marketplace listings, personal sites and Discord all pass the file through with metadata intact. WhatsApp and Telegram are the specific trap: both strip through the photo picker and both preserve the untouched original when you use "send as file" or document mode, which is what people pick when they want full quality. Also worth knowing — platforms that strip for viewers still read the data server-side on upload.

I already used Windows' "Remove Properties and Personal Information" — do I still need this?

Usually yes. That feature clears the fields Explorer and ordinary EXIF readers display, which is why the Details tab looks empty afterwards. What it does not do is take the data out of the file. Didier Stevens documented in 2022 that Explorer deletes the EXIF directory entries pointing at each value while leaving the value strings sitting in the JPEG APP1 segment — strings or a hex editor still finds them, on fully patched Windows 7, Windows 10 21H2 and Windows 11 alike. A quick tell: if the file size barely moved, the bytes did not go anywhere.

Drop the Explorer-cleaned file here and the first-pass scan will still report EXIF present, because it looks for the APP1 marker and Explorer does not remove that segment. The canvas re-encode then builds a new file out of pixels alone, so there is no segment left for orphaned strings to sit in.

My animated WebP came back as a single frame, or a huge photo came back blank

Two different hard limits of the canvas path. A canvas holds one frame, so an animated WebP or APNG is decoded to frame one and re-encoded as a still — for animation you need a tool that rewrites container chunks instead. Separately, browsers cap total canvas area far below the file-size cap, and Safari on iOS is the strictest: past roughly 16.7 million pixels — which a 24 MP phone photo at 6000x4000 exceeds — it hands back an empty bitmap instead of raising an error. Run large files on a desktop browser, or resize first.

It works on my laptop but does nothing on my iPhone

Same code, different ceiling. The mobile size cap is 100 MB rather than 500 MB, but the limit you are far likelier to hit is the canvas area cap in the entry just above. Whether it bites depends on your camera setting: a 12 MP shot at 4032x3024 is comfortably under it, a 24 MP shot at 6000x4000 is over. Chrome and Firefox on iOS run on the same WebKit engine, so switching browsers on the same iPhone does not raise the ceiling.

In order of effort: AirDrop or email the file to a desktop browser and clean it there, export a smaller copy first, or drop the camera to a lower resolution. Chrome on Android allows considerably more area, but it is still bounded by device memory, so a large panorama can fail there too.

How do I confirm the metadata is really gone?

Check with a tool that is not this one. On macOS, open the cleaned file in Preview and press Cmd+I for the Inspector — the EXIF and GPS tabs should be absent. On Windows, right-click the file, then Properties, then Details: camera, GPS and author rows should be blank. For a complete dump on any platform, exiftool -a -G1 yourfile.jpg lists every remaining tag. Verifying with an independent reader is better practice than trusting any cleaner's own report, including ours.

Your files never leave the browser tab

There is no upload step, and you do not have to take that on faith. Open DevTools, switch to the Network panel, drop in a file and clean it: no request carries your image. The image metadata remover is a plain ES module served from this site, so you can also read metadata.js in the Sources panel and see that the only things it touches are a FileReader, an Image and a canvas.

Nothing is stored, because there is nowhere to store it — closing the tab discards everything. No account, no queue, no retention period to trust. Our full privacy policy covers the analytics and advertising the site itself uses.

Use it on files you own or are authorised to handle

Clean your own photos, your client's deliverables, or files you have permission to process. Do not use it to sanitise images obtained without authorisation, to interfere with an investigation, or to strip provenance from someone else's work before republishing it.

Removing metadata from your own photos is a normal privacy practice everywhere. Removing copyright management information from another person's image in order to redistribute it is a separate matter with real legal exposure in several jurisdictions, including under Section 1202 of the US DMCA. See our acceptable use policy for the full terms.

A clean file is not an unlicensed file

Deleting the copyright, creator and licence fields deletes the record, not the obligation. A stock photo licensed for editorial use only is still editorial-only after its IPTC block is gone; a CC BY image still requires attribution when the credit string no longer travels with the file.

The same applies to AI provenance. Dropping a C2PA manifest removes a claim about where an image came from — it does not change where it came from, and it does not remove an invisible watermark from the pixels. If you are publishing under rules that require disclosing AI-generated imagery, this metadata remover does not exempt you from them.