Will it delete the duplicates from my computer?

No — and it can’t: a web page has no permission to delete files on your disk. The tool only reads your photos and finds the duplicates. Download the “kept photos” ZIP to replace the original folder, or export the CSV list and delete the rows marked keep = no in your file manager. Photos are read locally in your browser and never uploaded.

What is the difference between “exact” and “similar”?

Exact means the files are byte-for-byte identical (same SHA-256), usually the same file copied several times. Similar means the pictures look the same or nearly so — a photo that was resized, recompressed as JPG, converted to WebP, slightly colour-corrected or given a small watermark. Those files differ in content, so perceptual hashes are needed to spot them.

What similarity threshold should I use?

The threshold is how many of the 64 fingerprint bits may differ. 0 finds only visually identical pictures; around 8 reliably catches resized, recompressed and converted copies with very few false matches; 12 and above also groups burst shots and similar compositions of the same scene, which is handy for culling bursts but worth checking by eye. “Combined” averages all three hash distances and is steadier than any single hash.

Can it handle thousands of photos?

Yes. Each photo is decoded in a Worker, immediately shrunk to 320 px for hashing and released, so memory use stays low, and up to four Workers run in parallel depending on your CPU. A typical computer analyses dozens of photos per second, depending on size and format (HEIC needs an extra decoder and is slower). The comparison step computes pairwise Hamming distances and takes well under a second even for thousands of images.

👯

👯Duplicate Photo Finder - Find Similar & Identical Images

Phone backups, camera exports and saved chat images are full of duplicates: the same picture stored several times, or resized, recompressed or converted to another format. Drop in hundreds or thousands of photos, or a whole folder, and the analysis runs in parallel Web Workers. First every file gets a SHA-256 hash, so byte-identical files are grouped straight away. Then each image is shrunk to a tiny greyscale version and fingerprinted three ways with 64-bit perceptual hashes — aHash (average), dHash (neighbouring gradients) and pHash (low frequencies of a 32×32 DCT) — and similarity is measured as Hamming distance, so the threshold slider goes from “visually identical” to “different versions of the same shot”. Each group is shown side by side with thumbnails, resolution, file size, date taken (from EXIF, or the file’s modified date) and format. Pick which photo to keep automatically — highest resolution, newest, oldest, smallest or largest file — or tick them by hand, then export the list as CSV or download the kept set as a ZIP. Note that a web page cannot delete files on your computer: the tool finds duplicates and you do the deleting. Photos are only read inside your browser and never uploaded; each one is downscaled and released right after decoding, so even large sets stay light on memory, and you can cancel at any time.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Duplicate Photo Finder

1Choose photos, choose a whole folder, or drag photos/folders in
2Pick the comparison method and similarity threshold, then start the scan and wait for the progress bar
3Review each group, auto-select or tick the photos to keep, then export the CSV list or download the kept photos as a ZIP

FAQ

Will it delete the duplicates from my computer?: No — and it can’t: a web page has no permission to delete files on your disk. The tool only reads your photos and finds the duplicates. Download the “kept photos” ZIP to replace the original folder, or export the CSV list and delete the rows marked keep = no in your file manager. Photos are read locally in your browser and never uploaded.
What is the difference between “exact” and “similar”?: Exact means the files are byte-for-byte identical (same SHA-256), usually the same file copied several times. Similar means the pictures look the same or nearly so — a photo that was resized, recompressed as JPG, converted to WebP, slightly colour-corrected or given a small watermark. Those files differ in content, so perceptual hashes are needed to spot them.
What similarity threshold should I use?: The threshold is how many of the 64 fingerprint bits may differ. 0 finds only visually identical pictures; around 8 reliably catches resized, recompressed and converted copies with very few false matches; 12 and above also groups burst shots and similar compositions of the same scene, which is handy for culling bursts but worth checking by eye. “Combined” averages all three hash distances and is steadier than any single hash.
Can it handle thousands of photos?: Yes. Each photo is decoded in a Worker, immediately shrunk to 320 px for hashing and released, so memory use stays low, and up to four Workers run in parallel depending on your CPU. A typical computer analyses dozens of photos per second, depending on size and format (HEIC needs an extra decoder and is slower). The comparison step computes pairwise Hamming distances and takes well under a second even for thousands of images.