Private file comparison
Find lookalikes inside the files you own
Rank your own files against a master, or group them. Nothing is uploaded, and this does not search the web, the archive or a trademark register.
1. Master image
2. Reference images to compare (up to 60)
Static SVG, PNG, JPEG and WebP only. Maximum 60 references, 5 MB per image, 40 MB including the master and 16 megapixels per raster. ZIP: 80 MB compressed, 500 entries, 40 MB of supported images after expansion. A new selection replaces the current references. Unsupported file types are skipped.
Advanced review settings
Lower distances narrow the review. These controls do not set a confidence percentage. Grouping compares references with one another and does not require a master.
0 references selected. Source images, filenames and results stay in this page’s memory.
Saved fingerprints and backups
Saving is optional and keeps filenames, dimensions and compact fingerprints on this device. Anyone with access to this browser profile may read them. Original images are not stored. V1 data is read only; it is never rewritten or deleted here. Collections are limited to 200 entries without automatic eviction.
0 saved fingerprints available.
Next step
For a formal check, search a trademark register
A visual ranking of your own files is not a trademark search or legal clearance. Before you adopt or register a mark, search the official registers for the markets you trade in, and ask a qualified trademark adviser.
- WIPO Global Brand Database International and many national records, including a search by image.
- USPTO Trademark Search United States federal trademark records.
- EUIPO / TMDN TMview European Union and participating national offices.
- DPDT Bangladesh Department of Patents, Designs and Trademarks: filing and published trademark journals.
Read the signals, then inspect the originals
Identical SHA-256 values identify identical file bytes. For other files, the tool first removes a plain white, black or single-colour background and trims empty margins. When both files separate cleanly, it compares the outline of the artwork on a coarse grid, grayscale difference hashes on white and black backgrounds, an alpha-weighted colour histogram and proportions. The plate of an app icon or solid badge is removed too, so the mark drawn on it is compared, and the separate left emblem of a lockup (a symbol beside its wordmark) is also read on its own; the row says when that happened. The outline is measured, not the structure inside a mark. Photos, screenshots, pairs of app-icon plates that match more closely as whole files, and saved fingerprints are compared as whole files with the hashes, histogram and aspect ratio only. Each row says which reading was used and gets a plain-language tier (near-duplicate, similar or low overlap) read from the distance. These fixed heuristics are useful for sorting a review set, but they are not a trained recognition model, a probability or a calibrated confidence percentage.
Two different logos can share a coarse fingerprint. Compression, rotation, cropping, fine detail and different lockups can change the ordering. A matching preview does not establish ownership, permission, originality or legal clearance. Folder and ZIP inputs use the same file checks. Reports are evidence for a human review, not automatic cleanup instructions.
What stays on your device?
Original images and previews stay in this page’s memory. Clear the files or close the page to discard them. If you choose to save, filenames, dimensions and compact fingerprints remain in this browser profile until you delete the saved collection. Anyone with access to that profile may read them. Exported JSON, CSV and printed reports also remain wherever you save them.
Existing v1 collections are read without modification and can be backed up or compared separately. They contain no original image bytes or exact-file hashes. The v2 review method is different, so its distance must not be compared with v1. Opening a different domain or browser profile does not transfer either collection automatically.
How do grouping and sensitivity work?
Grouping checks every pair in the selected reference set, bounded to 60 current files and 200 saved records per fingerprint version. Links below the selected distance form connected review groups. A chain can connect two dissimilar endpoints, so a group does not establish that all its files are duplicates. JSON exports preserve the individual links for inspection. Nothing is automatically merged, deleted or saved.
The display filter and group distance are under “Advanced review settings”. The display filter narrows visible comparison results; exports still contain all compared rows. Legacy shape weighting appears only when a v1 collection exists and changes only the v1 recipe. Changing a recipe or group threshold clears affected results so a previous grouping cannot be mistaken for a new calculation. These settings are review controls, not calibrated match probabilities.
The page itself still loads scripts and fonts over the network. Artwork, filenames and fingerprints are not sent through network requests or analytics by this tool.
Methods and limitations → · Search the public archive for similar logos (a different, server-assisted workflow) →