What a photo can say about you that the picture doesn't
The reader is a small parser written for this page. It walks the file's own structure and reports what it recognizes: camera make and model, software, dates, serial numbers, an owner or artist name, GPS latitude, longitude and altitude, embedded thumbnails, color profiles, XMP fields such as the creator tool, IPTC captions and credits, and the text fields PNG files can carry. Each row is tagged Location, Time, Device, Identity, Provenance or Technical, so you can see which ones matter before you share. The usual surprise is the GPS row on a photo of something indoors, taken at home.
It reads JPEG, PNG and WebP. HEIC and AVIF files are identified, but their metadata isn't listed.
How to strip metadata from a photo
- Pick a photo with the chooser above, or drop it on the box.
- Read the list. If the GPS row shows coordinates, anyone who receives this file can see where it was taken.
- Choose the output format. Same as the original is the default. Set a quality for JPEG or WebP; 92 is a sensible start.
- Press Create clean copy. The page decodes the image, draws it on a blank canvas, and encodes a new file from that canvas.
- Read the second list, which is the same reader run over the new file. Then download it.
What each row tag means and what to do about it
| Tag | Typical contents | What to do before sharing |
|---|---|---|
| Location | GPS latitude, longitude and altitude | Remove for anything posted publicly or sent to a stranger. Home, school and work coordinates are the risk. |
| Time | Capture date and time, sometimes a time zone offset | Remove when the date tells a reader when you were away from home or when a child was somewhere. |
| Device | Camera make and model, lens, serial number, software | Remove serial numbers in particular, since they can tie several photos to one device. |
| Identity | Owner, artist or copyright strings, captions, credits | Keep it when you want credit on a portfolio image. Remove it when the photo should not name you. |
| Provenance | Content Credentials, IPTC source type, generator text | Read it before you decide. Removal is covered in the AI watermark guide, with the limits. |
| Technical | Dimensions, color profile, thumbnails, file structure | Mostly harmless. An embedded thumbnail can be worth removing, since it is a second copy of the picture. |
Where the metadata sits in each format
The three formats the reader opens store the same kinds of information in different places. A JPEG keeps EXIF and XMP in marked segments near the start of the file, ahead of the compressed picture data. A PNG keeps text in named chunks, and it can also carry an EXIF chunk. A WebP file is a container that holds the picture in one chunk and can hold EXIF and XMP in others. The parser follows those structures and reports each field it recognizes by name, which is why the output reads as a list of labelled facts instead of one dump of text. A field that a camera or app wrote in a private, undocumented layout may appear with a generic label or not at all, and that's one reason the second list on the clean copy is worth reading.
Who uses a viewer like this, and for what
- Anyone selling or giving away items online. Photos of a room or a car park can carry the coordinates of the place where the object sits. Check the Location row before you attach the picture to a listing.
- Parents and teachers sharing pictures of children. A school event photo with a time and a GPS fix tells a stranger where a child was and when.
- Photographers who want credit. The Identity rows show whether your name and copyright text made it into the file, so you can keep them on portfolio images and drop them elsewhere.
- People who received a file and want to know about it. The viewer reads the labels and fields without needing the original app that made the image.
- Anyone preparing a document or report. Screenshots and exports can carry software strings and dates you didn't mean to publish.
When the page shows less than you expected
A short list is common and not always good news. Messaging apps, social platforms and many email tools re-encode pictures on upload, and in doing so they usually drop the EXIF. A photo you downloaded from a social site often arrives with almost nothing, while the same photo sent as a file or an attachment keeps everything. If the list is empty, the file isn't necessarily clean in every sense, because the picture can still show a street sign, a reflection or a face. The viewer reads the file's data. It doesn't look at what the photo shows.
If you pick a HEIC photo from an iPhone and the page lists little, that's the format limit described above and not proof that the photo has no location. Export a JPEG copy first, then read that here. The iPhone guides linked in this page walk through the settings that control what a new photo records in the first place.
A worked example: a photo for a rental listing
You photograph the living room of a flat you're letting. The viewer shows a Location row with latitude and longitude accurate to a few meters, a Time row with Saturday morning, and a Device row with the phone model and software version. The renters need the room, so you choose JPEG at quality 92, press Create clean copy, and read the second list. It shows the image width and height and the basic file structure and nothing in the Location, Time or Device groups. You attach the copy to the listing and keep the original in your own storage, where the place and time are still on file in case a dispute needs them.
Before you share a photo: a short routine
Open the photo here first and read the tagged rows. Location and Device rows are the usual leaks, Identity rows can carry an owner name, and Time rows give away when you were somewhere. Decide what the recipient needs. A buyer needs the picture, not the coordinates of your kitchen. Then create the clean copy, read the second list to confirm the rows are gone, and send the copy rather than the original. Keep the original for yourself, since it holds the capture time and place you may want later.
Why the clean copy carries no EXIF
A canvas holds pixels and nothing else. When the browser encodes it, the new file contains the picture plus the basic file structure the browser itself writes. None of the original's EXIF, XMP, IPTC, text fields or manifest is copied, because the tool never reads them into the canvas. The second list exists so you don't have to take that on trust.
Reading Content Credentials and AI labels in a file
The viewer can tell you whether the file contains a Content Credentials (C2PA) marker, and it prints a few readable strings from it, such as an IPTC digital source type that labels media as AI-generated. That's a reading of what is written in the file. It doesn't check the signature, so a marker found here doesn't mean the credential is valid, and the absence of one doesn't mean an image is real. For background, see what Content Credentials record and how to check whether an image is AI generated.
What changes in the picture itself
- JPEG and WebP are lossy. Every re-encode adds some compression. Higher quality means a larger file and less change.
- PNG is lossless but the file can grow.
- Transparency is flattened onto white when the output is JPEG.
- Animation is lost. Only the first frame is kept.
- Orientation is applied to the pixels, so the photo stays upright without a tag.
- A second embedded image, such as an HDR gain map, is dropped. You keep the standard picture.
- Color profiles are not copied. Wide-gamut photos may shift slightly.
If your browser can't encode WebP, it produces PNG instead and the page says so.
Where your file goes while you use this page
The page reads the file with the browser's file API, decodes it with the browser's image decoder, and encodes it on a canvas. The script contains no network calls, and closing the tab discards everything. To find out what a particular phone puts in its photos, EXIF metadata on iPhone photos and removing location data from iPhone photos go through the settings, and finding where a photo was taken covers the reading side.
Questions people ask
Does this remove AI watermarks?
It removes file metadata, including a Content Credentials manifest, from the copy it makes. It doesn't remove watermarks that live in the pixels, such as SynthID, or visible logos. See the AI watermark guide.
Can it prove an image was made by AI?
No. A label in the file is a claim written by whatever software handled the image, and its absence proves nothing, because labels are easy to lose. Treat it as one clue among several.
Does it work with HEIC photos from an iPhone?
It identifies HEIC but can't list its metadata. If your browser can display the file, the clean-copy step may still work and produces a JPEG or PNG. See removing metadata on an iPhone.
Will the cleaned photo look worse?
Slightly, for JPEG and WebP, because each encode compresses again. At quality 92 or above the difference is usually hard to see. Choose PNG if you want no further loss.
Is my photo uploaded?
No. The reading and re-encoding happen in the tab, and you can confirm that in your browser's network panel while you use it. Closing the tab discards everything the page held.
Can I clean several photos at once?
Not on this page. It handles one file at a time, so run each photo through in turn and check the second list for every copy before you share it.
What is photo metadata?
It's data stored inside an image file next to the picture: camera make and model, date and time, GPS position, software, thumbnails, and sometimes an owner name or a provenance label.
What are the steps to strip the hidden data from a picture?
Choose the photo here, press Create clean copy, check the second list and download the copy. It's rebuilt from pixels alone, so the original fields aren't carried across.
Do the pixels change when the hidden data is removed?
The picture is re-encoded, which adds a little compression for JPEG and WebP. At quality 92 or higher the change is usually hard to see, and PNG adds none.
Will the viewer show where a picture was taken?
Yes, when the file has GPS fields. The row shows latitude, longitude and altitude, tagged as Location. Finding where a photo was taken covers reading them.
Can a file tell me whether an image came from an AI tool?
Read the Provenance rows for a Content Credentials marker or an IPTC digital source type. A label counts as one clue, and a missing label proves nothing. See checking whether an image is AI generated.
Which AI tools can the page recognise from a file?
It prints readable strings that are written in the file, such as the software or creator tool and a digital source type. If the file carries no such string, the page has nothing to name.
Can I delete the AI labels stored in a file?
Create the clean copy. Labels stored as file metadata aren't copied into it. Marks that live in the pixels, such as visible logos or invisible watermarks, stay.
Can the Content Credentials record be taken out of a file?
The clean copy drops the manifest because the tool never reads it into the canvas. The second list confirms it. More on the format is in what Content Credentials record.
Does cleaning a file erase its Content Credentials?
Yes from the copy it makes, since that copy holds pixels only. Your original file keeps its credentials untouched.
Does cleaning touch invisible watermarks such as SynthID?
No. Those live in the pixels, and re-encoding the same pixels doesn't remove them. The page doesn't claim to.
Which image formats are supported?
JPEG, PNG and WebP are read and listed. HEIC and AVIF are identified, but their metadata isn't listed.
Sources