Similar Images.
Near-duplicate photos and clips, grouped by how they look.
A perceptual hash (dHash) compares how images actually look, so it clusters resized copies, edits, re-exports and re-encoded clips that exact-hash duplicate detection would never match. It groups them by perceptual distance, keeps the highest-resolution original, and stages only the redundant copies.
- perceptual dHash
- resized and edited
- re-encoded clips
- keeps the best
What you see while it scans
Every Atlas tool has its own loading animation, drawn from the work that tool actually does. This one shows tiles being compared and linked as the perceptual pass finds lookalikes, the same dHash comparison that runs while Similar Images works, rendered live on this page.
Near-duplicate photos and clips
-
1
Find near, not just exact
Similar Images uses a perceptual hash to cluster photos and clips that look alike, catching resized and re-encoded versions that exact-hash duplicate finding misses.
-
2
One best per cluster
Each cluster keeps a best image scored at the top, with the near-copies scored by how close they are, so you keep the best and stage the rest.
-
3
Tune the threshold
A similarity threshold lets you decide how close counts as a match, tightening or loosening the clusters live.
-
4
Assets stay out of it
Game and app assets, mods and textures are excluded from matching, so only your actual photos and clips show up here.
Your photo library is full of near-identical shots, resized copies and re-encoded clips that exact duplicate detection walks right past.
Look-alike photos and clips grouped into clusters, the best of each kept, the near-copies staged, with your game assets left untouched.
