Image Metadata Viewer

1. 1
   Drop your image: Drop a JPEG, PNG, WebP, GIF, BMP, TIFF, HEIC, AVIF, SVG or RAW file. Only the few hundred kilobytes holding the metadata are read, so a 24 MB photo opens as quickly as a small one.
2. 2
   Read the table: Every tag appears, grouped by what it describes: camera, exposure, lens and focus, date and time, location, author and rights, image, colour and the manufacturer's maker note. Search by tag name or by value.
3. 3
   Check the privacy scan: The panel above the table names what identifies you — coordinates, the camera's serial number, the owner's name, the capture time and the device — with a count for each.
4. 4
   Choose a clean: Remove location, remove personal fields, or remove all metadata. The list of exactly which tags will go is shown before you export, not after.
5. 5
   Export: The cleaned copy is assembled from the original's own segments, so the compressed image data is copied byte for byte and the picture is unchanged. Or download the whole table as JSON, CSV or plain text.

Every tag a photo carries — EXIF, IPTC, XMP, the colour profile and the manufacturer's own private note — read from the file's own tables rather than summarised.

A photo carries far more than the camera and the date. Alongside EXIF sit IPTC captions and credits that press and stock workflows run on, an XMP packet holding whatever the last editor did, an ICC profile deciding what the colours mean, and a private block the manufacturer writes in its own layout. Most online viewers show the first of those and none of the rest.

• EXIF, in full: every directory — IFD0, the Exif sub-directory, GPS, Interoperability and the thumbnail directory — rather than a fixed list of popular tags.
• IPTC-IIM: caption, headline, byline, credit, source, keywords and the copyright notice. This is the block the best-known online viewer never showed.
• XMP: the raw packet and its properties, including ratings, develop settings, drone telemetry and the edit history left behind by desktop editors.
• ICC: profile description, class, colour space, rendering intent and white point — the answer to why an image looks different on another screen.
• Maker notes: Canon, Nikon, Sony, Fujifilm, Olympus and Panasonic, each addressed the way that manufacturer actually writes it.

Tags are grouped on screen by what they describe, but each one belongs to a block inside the file. The block is shown beside every row, because the same name can mean different things in different places.

Metadata is addressed, not searched for. Every container here carries a table saying where its metadata is — a JPEG a chain of markers, a PNG a chunk list, HEIC and AVIF an item-location box, TIFF and every RAW derived from it a directory offset. The reader follows those tables and fetches only the ranges they name.

• Locate: walk the container's own structure to find each metadata block. A 24 MB photo costs a 256 KB read, because the tags start in the first few bytes.
• Decode: read every directory entry with its real type and count, so a rational stays a fraction and an array stays an array.
• Place the maker note: work out which manufacturer wrote it and what its internal offsets are relative to — the outer file header for Canon, Sony and Panasonic, the note's own start for Fujifilm, Olympus and Nikon.
• Group: file each tag by its identity rather than by the label on screen, so the grouping works the same in every language.
• Scan for risk: flag the tags that name a place, a person, a device or a moment.

• Runs in the background: the reading happens in the background, so the page never freezes on a large file.
• Preview without decoding: the thumbnail the camera already embedded is used as the preview, which is why a RAW file shows a picture the browser could not otherwise open.
• Unknown tags are shown, not dropped: a tag with no name in the specification appears under its id. A nameless tag with a readable value is still information.
• Search across names and values: find every row mentioning a lens, a date or a serial number at once.
• One file or a hundred: a batch reads into one table per image and exports as a single archive with a combined report.

Nothing is decoded, so a file the browser cannot display can still be read completely. That is what makes RAW work here.

Every online metadata cleaner benchmarked for this tool works the same way: decode the image, draw it, encode it again. That works, and it silently re-compresses the photo. Here a clean is a copy of the container's segment list — the metadata segments are dropped or overwritten with zeros, and the compressed image data is copied across byte for byte. The exported file decodes to pixels identical to the original's.

• Remove location: the GPS directory and the place-name fields in IPTC and XMP. Camera settings, lens and capture time stay.
• Remove personal: the above plus serial numbers, owner and author fields, capture times, editing history and the maker note.
• Remove all metadata: every metadata block the container marks removable.
• The colour profile always stays: removing it would change how the picture looks, which makes it an edit to the image rather than a privacy measure.
• Removed bytes are zeroed, not orphaned: the coordinates are overwritten where they sit, so they cannot be recovered from the file afterwards.

Anything that starts with a question about a file rather than about a picture: where it came from, what took it, whether it is the same file you were sent, and what it would tell a stranger.

• Before posting a photo: check the privacy scan, remove location, and post a file whose pixels are unchanged.
• Settling how a shot was taken: read the exposure, the lens, the focus mode and the drive mode, including the settings only the maker note records.
• Checking a delivery: confirm the caption, credit and copyright a client asked for are actually written into the IPTC block.
• Working out a colour shift: read the ICC profile description and rendering intent.
• Verifying a file is the file: compute its SHA-256 and compare it against the one you were given.
• Reading a RAW you cannot open: a CR3 or an ARW gives up its full directory chain here even though nothing on the page can display it.

Numbers from the tool's own measurements on a desktop machine, not estimates.

• A 24 MB photo is read in 0.23 ms, touching 256 KB of the file — the tags live in the first few bytes, so the rest is never read.
• Nine metadata blocks are parsed: EXIF, GPS, Interoperability, the thumbnail directory, IPTC-IIM, XMP, ICC, Photoshop resources and the maker note.
• Six manufacturers' maker notes are decoded, each with the addressing rule that manufacturer actually uses.
• Nineteen input formats are read, including seven RAW formats and three ISO base-media containers.
• A cleaned export differs from its source by 0 of 255 on every channel, measured by decoding both.
• Everything happens on your device. No image is uploaded, and nothing is sent anywhere unless you click the map link yourself.

A metadata tool is easy to be confidently wrong with, because a wrong answer looks exactly like a right one. These are the things worth knowing.

• Capture times carry no timezone. They are the wall-clock time where the shutter fired, so they are shown and written back unchanged — a viewer that converts them through UTC moves every holiday photo by the traveller's offset.
• Emptying an editable field deletes that tag rather than leaving the old value in place.
• Editing rewrites the metadata block, so the maker note has to be carried across and re-addressed. The result is parsed again and compared against the original; if it does not match exactly the note is removed and you are told.
• HEIC, AVIF and RAW files are read but not rewritten. Their metadata is addressed by an item table whose offsets all shift when a block changes size, and a rewrite that gets that wrong corrupts the file quietly.
• A file with no metadata is a real answer, not a failure. Screenshots never carry EXIF, and social platforms strip it before delivery.

• For PDF, Word, video or audio metadata, use the dedicated tool for that format — the blocks are completely different.
• For changing what a picture looks like, use the editor or the converter. Nothing here touches a pixel.
• For proving where a photo was taken, treat the coordinates as a claim rather than as evidence: every field in a file can be written by hand.

A 24 MB photo is read in 0.23 ms by fetching the 256 KB the file's own tables point at, rather than loading the whole image.

• Image analyzer
• Batch rename
• Image converter
• Image compressor