Find duplicate & similar photos without upload

No. The photos are opened and compared in your browser and do not leave your device. The AI model and the runtime are served by mytoolio itself – no other provider is involved.

The slider sets the similarity from which photos form a group. From about 90% it is almost always the same picture – resized, saved again, slightly cropped or changed in brightness and colour. Between 75 and 90% you mostly find several shots of the same scene, such as bursts. Below that the tool finds similar subjects, for example two different bridges at night. Files that are equal byte for byte are labelled “identical file”.

The one with the most pixels; at equal resolution the sharper one (measured in the middle of the picture), then the larger file. Only copies that are at least 90% like a kept photo are pre-marked for sorting out – merely similar photos stay on “keep” at first and you decide yourself. Every mark can be changed with a click; the magnifier shows the photo large, and the arrow keys page through the group.

Only if you explicitly want it to. By default it creates a list of the marked files. If you open the folder with “Choose folder” in Chrome or Edge on a computer, the tool can move the marked photos into the subfolder “_sorted-out” after your permission – from there you can bring them back at any time. There is also “delete for good”: the tool asks first, with the number and size, deletes only files that are unchanged since the check, and refuses as long as a group would keep no photo. Deleted photos do not go to the recycle bin – if in doubt, move them instead.

Up to 10,000 photos on a computer and 2,000 on a phone. With graphics-card support (WebGPU in Chrome and Edge) the tool handles about ten photos per second, so 1,000 photos in roughly two minutes; most of that time goes into opening the files. Without WebGPU the processor manages about three photos per second.

Everything your browser can display: JPG, PNG, WebP, AVIF, GIF and BMP, in Safari also HEIC. RAW files and videos are skipped. Mirrored copies are usually recognised, copies rotated by 90° are not. After the first download the model stays in the browser cache, then the tool also works without an internet connection.

Yes. “🔍 Find similar” in the large view, or “Find photos like this …” with any photo, shows all checked photos that look like it, sorted by similarity – for example all pictures of the same place or subject. The search photo itself is not added to the list. With “Date taken from … to …” only photos from that span count, for the groups as well as for the search. The date comes from the camera data (EXIF) of the photo; without it, the file’s modification date is used.