Inspect image metadata, colour and quality without uploading
Drop a photo or graphic to read its dimensions, format structure, EXIF and GPS metadata, colour profile, palette, histogram, exposure, clipping, focus estimate, edge map and perceptual fingerprints. Decoding and analysis happen only in this browser tab.
📥
Drop an image to inspect it
or click to browse · paste from clipboard · drop a whole folder
JPEG, PNG, GIF, WebP and BMP have detailed format parsing. Other formats are analysed when your browser can decode them.
One image has two very different layers
The visible pixels are only one layer. The file around them can record the camera and lens, capture time,
editing software, copyright owner, an embedded preview and sometimes the exact GPS coordinates where the
shutter was pressed. This inspector reports pixels and metadata separately so you can see both.
Technical structure — dimensions, bit depth, compression, colour model, pixel density, progressive encoding, chroma subsampling, transparency, animation and frame count.
Metadata — common EXIF camera fields, dates, exposure settings, lens and serial identifiers, GPS, ICC profile headers, plus the presence of XMP and IPTC blocks.
Pixel analysis — RGB and luminance distributions, dominant palette, average colour, clipping, contrast, focus and noise estimates, and JPEG block-pattern signals.
Visual fingerprints — an edge map plus aHash and dHash values for finding images that look similar even when their file bytes differ.
How to read the quality estimates
Sharpness, noise, exposure and compression artefacts do not have universal pass/fail numbers. A deliberately
soft portrait, a dark night scene and a grainy film scan can all be excellent images. The scores here are
measurements and heuristics intended for comparison, not aesthetic verdicts.
Sharpness uses variance in a Laplacian filter. Defined edges raise it; blur lowers it. Image content and resolution also affect the value.
Noise measures local residuals in relatively smooth areas. Texture, grain and aggressive sharpening can look like noise.
Clipping counts nearly black or nearly white pixels. Small percentages are normal; large values may indicate lost shadow or highlight detail.
Block pattern compares changes at eight-pixel boundaries with neighbouring changes. It can flag JPEG artefacts, but geometric artwork can produce the same pattern.
Private by architecture, not by policy
The selected file is passed directly from your file picker to JavaScript in this tab. Metadata parsing,
browser image decoding, canvas sampling, charts and report generation all happen on this device. No endpoint
receives the image, its filename, its metadata or the analysis result.
For memory safety, pixel analysis downsamples very large images to at most about 1.4 million pixels and
metadata scanning reads at most the first 64 MiB. The displayed dimensions remain the original dimensions.
On unusually large files, the interface states when trailing-data detection or late metadata could not be checked.
Finding no supported metadata is not proof that a file is clean.
Proprietary maker notes and unknown application blocks may still exist. Likewise, metadata describes what a
file claims about itself and can be edited or forged.
🔒 Why this is safe for confidential files
Nothing is uploaded The file is read directly from your disk by JavaScript in this tab. No network request carries its contents.
No server to trust There is no backend. Even if we wanted your files, there is nowhere for them to go.
Works with the network off Load the page once, disconnect, and every tool still works. That is the strongest proof that nothing is being sent.
Nothing is retained Close the tab and the file is gone from memory. No history, no accounts, no storage.
Image inspection questions
No. The file is read by JavaScript in your browser and parsed on your device. Pixel decoding and the quality analysis also happen locally. You can load the page, disconnect from the internet and continue using the inspector.
Open the Image tab after selecting it. If readable GPS latitude and longitude tags exist, a prominent location warning and the coordinates appear in the EXIF card. Some apps store location outside standard EXIF, so absence of that warning is not an absolute guarantee.
It measures edge energy, so both focus and subject matter affect it. A perfectly focused blank wall scores lower than detailed foliage. Compare like with like — repeated shots of the same scene or outputs from the same workflow — rather than treating it as a universal quality grade.
A perceptual hash describes the visual pattern of an image rather than its exact bytes. Resizing or recompressing an image can leave a similar aHash or dHash, while its SHA-256 changes completely. Perceptual hashes help find likely visual duplicates; they are not cryptographic and should not verify file integrity.
It detects animation and counts frames in GIF, animated WebP and APNG files. Pixel analysis describes the browser-decoded representative frame, not every frame in the animation.