Inspect
Duplicate image finder
Find exact copies and look-alike photos in a set of images or a whole folder, compare them side by side and decide what to keep.
Identical bytes are proven by SHA-256. Visual similarity is an estimate: always check a group before you tidy up.
Strict finds near-identical copies only; Loose also catches heavier edits but may group photos that only look alike.
Nothing is deleted. Use the report or the list to tidy your folders yourself; this page cannot change files on your device.
Exact duplicates
Identical bytes (same SHA-256): these files are perfect copies.
Visually similar
Look alike to the perceptual hash: resized, recompressed, converted or lightly edited versions. Check each group before relying on it.
Unique images
Files compared by bytes only
How matching works
Exact copies and similar pictures
Exact duplicates are files whose bytes are the same, so their SHA-256 hashes are equal. They are certain matches, whatever their names or folders.
Visually similar images are compared with a perceptual hash (pHash): each picture is reduced to a 32 × 32 grey grid, its broad shapes are summarised in 64 bits, and pictures whose fingerprints differ in only a few bits are grouped. That survives resizing, recompression, format changes and small brightness edits.
Limits
What it can and cannot tell
Similarity is an estimate. Burst photos, scans of similar pages or plain backgrounds can look alike; cropped, mirrored or heavily edited versions may be missed. Nearly blank pictures are only compared by bytes, because their fingerprints are meaningless.
In each group the largest picture (most pixels, then largest file) is suggested to keep. You can choose another, and nothing is ever deleted for you.
FAQ
Duplicate photo questions.
How does it find duplicates?
Two ways. Exact duplicates have identical bytes: the tool computes the SHA-256 hash of every file, and equal hashes mean perfect copies. Visually similar images are found with a perceptual hash, a 64-bit fingerprint of what the picture looks like, so resized, recompressed or converted versions still match.
What do Strict, Balanced and Loose mean?
They set how different two perceptual fingerprints may be and still count as similar: up to 4, 10 or 16 of 64 bits. Strict finds near-identical copies. Balanced also finds resized and recompressed versions. Loose catches heavier edits but may group photos that only look alike, such as a burst of very similar shots.
Can it find rotated copies?
Yes. With “Also match rotated copies” on, each picture is also compared turned by 90°, 180° and 270°. Mirrored or cropped versions are not matched reliably.
Does it delete my duplicates?
No. A web page cannot change the files on your device, and nothing is deleted. Choose which picture to keep in each group, then use the CSV or JSON report, or copy the list of extra copies, to tidy your folders yourself.
How many images can I check?
Up to 2,000 images at once, each up to 100 MB. The work runs in several background workers so the page stays responsive, and you can cancel at any time. Very large collections are faster on a computer than on a phone.
Are my photos uploaded?
No. Images are read and compared in your browser on your device. Thumbnails are made locally and discarded when you leave or start over. Nothing about the images, including names and hashes, is sent to analytics.