Image metadata viewer

Quick start

1. Add your images. Drag one image, several, or a whole folder anywhere onto the page, or click the drop box to choose files. Each image gets a card, and the first card opens by itself.
2. Read the summary first. The box called What this file reveals lists the important findings in plain words, sorted from most to least sensitive: location, serial numbers, names, dates, then smaller items.
3. Open the details. Below the summary, the metadata is split into groups such as Camera and lens, Location (GPS) and Software and authorship. Click a card's header to fold it open or shut.
4. Copy or export. Copy one file's metadata as JSON, download it, or use the buttons above the list to export every file as JSON or CSV.

If you only want the metadata gone, the Image Metadata Remover cleans many files at once. Each card here also has a Download without metadata button for JPEG, PNG, WebP and GIF files.

What the viewer shows

Photos and generated images carry several separate layers of hidden information. The viewer reads all of them and keeps them in labeled groups, so you can see where each fact came from.

Rows marked with a red dot are the kind that can identify a person or a device, such as GPS coordinates, serial numbers, an artist name or a contact address. Our guide on what EXIF data is and what it reveals explains each layer in more depth.

The "What this file reveals" box

Raw metadata is long and technical, so each card starts with a short summary written for people. Every finding has a label:

• Sensitive: exact GPS coordinates, camera or lens serial numbers, and fields that name a person.
• Notable: the date and time with its time zone offset, an embedded preview image, free-text comments, extra data hidden after the end of the image, and text hidden in the pixels.
• Minor: the device model, the editing software, an edit history, unique IDs and manufacturer notes.
• Info: generation settings from an image generator, and Content Credentials.
• Clean: shown when none of the above was found.

A file can be clean and still be sensitive in other ways, because a picture can show a street sign, a house number or a face. The viewer reads the file's metadata only. It does not look at what is in the picture.

GPS location: what you see and how far to trust it

When a photo has GPS data, the Location group lists latitude and longitude as decimal degrees and as degrees, minutes and seconds, then altitude, the direction the camera faced, speed and the map datum if the camera stored them. Two links open the spot in OpenStreetMap or Google Maps. Nothing is sent to a map service until you click one of them.

The coordinates are whatever the camera or phone wrote at the moment of capture. Accuracy depends on the device and on how it found its position. A phone with a clear sky view is usually much closer than one that guessed from nearby Wi-Fi networks or cell towers indoors. The values can also be edited or added by software afterwards, so treat them as a clue rather than proof. If EXIF has no GPS block but the XMP block carries coordinates, the viewer uses those and says so.

Prompts and settings from image generators

Many image generators save their settings inside the file they produce. The viewer recognizes these formats and splits out the prompt, the negative prompt and the settings, each with its own copy button:

• Stable Diffusion web UI, Forge and SD.Next write a parameters text chunk in PNG files, and the same text in the EXIF UserComment of JPEG and WebP files. The viewer separates the prompt, the Negative prompt: line and the final line of settings such as steps, sampler, CFG scale, seed, size and model.
• ComfyUI stores its node graph as prompt and workflow text chunks. The viewer follows the sampler's positive and negative inputs back to the text nodes and lists the seed, steps, checkpoint and similar values.
• NovelAI writes a Comment chunk with the prompt and settings, and may also hide the same data in the pixels.
• InvokeAI, Fooocus-style JSON parameters and Midjourney description text with a job ID are recognized too. Tools that only name themselves in the Software field or the XMP creator tool are reported by name.

The viewer also checks PNG and WebP pixels for hidden pixel data, a trick where text is stored in the lowest bit of the alpha or color channels. It is invisible to the eye and most software never shows it. Our guide to reading the prompt from a Stable Diffusion image covers where each tool keeps its settings and why they sometimes disappear. If you plan to reuse recovered prompts in a dataset, see captioning a LoRA dataset.

Content Credentials (C2PA) in plain terms

Content Credentials are a signed record attached to an image that can say which tool made it and what edits were applied. When the viewer finds one it shows the claim generator, the list of actions (including the declared digital source type, such as an image created by a trained model), the names in the signing certificate and the certificate dates. If the file holds earlier manifests from previous edits, it says how many.

The viewer does not verify the signature. It reports what is written in the file, which anyone could have written. To check that a manifest is intact and trusted, use an official verifier, and read what Content Credentials and watermarks can and cannot tell you.

Other hidden extras it looks for

• Embedded thumbnails. Many cameras store a small preview in the EXIF block. The viewer shows it, because after a crop or an edit it can still show the original framing.
• Data after the end of the image. Some phones append a short video clip or extra images after the JPEG data, and other tools append arbitrary bytes. The viewer measures the extra data and guesses what it is.
• Maker notes. Manufacturers keep private settings in a MakerNote block that the viewer does not decode. It reports the block's size and flags it because it can hold serial numbers.
• File structure. The last group lists each metadata block and how many bytes it uses, which helps when you wonder why a small picture has a big file size.

Supported formats

Browsers other than Safari usually cannot display HEIC pictures, so those cards show the format name instead of a thumbnail. The metadata is still read from the file.

What it can't tell you

• Invisible watermarks. Watermarks woven into the pixels, such as SynthID or Digimarc, need a dedicated detector. A metadata viewer cannot see them.
• Visible watermarks. Logos and text drawn on the picture are part of the image. The viewer does not read the picture's content.
• Whether a picture is generated. Finding generation data is strong evidence. Finding none proves nothing, because a screenshot, a re-save or a platform upload removes it.
• Whether metadata is true. Dates, GPS coordinates, camera names and even Content Credentials text can be edited or copied between files.
• Everything in MakerNotes. They are reported by size, not decoded.

Exports and the next step

The JSON export keeps the group structure of the cards, with the summary sentences included. The CSV export has one row per field, with the file name, group, field and value, which suits spreadsheets and quick audits of a whole folder. If you find location data you do not want to share, our guide on removing location data from photos shows the options on phones and computers, and the metadata remover cleans files in a batch.