Product

Five Files That Look the Same. Which One Is the Original?

Find every resized, cropped or flipped copy of a photo in your library, even under a new name, and copy the original's rights information onto them in one edit.

Growth Team

5 min read

Find duplicates in FileSpin: an original tote bag photo beside its mirrored, resized and cropped copies

A partner asks for the spring campaign image at four o’clock on a Friday. Their launch email goes out on Monday. You search the library for “spring hero” and get five results that look the same. You pick the first one and send it.

On Monday the email goes out with the photo flipped. It was the version a designer mirrored for a banner, so the logo on the product reads backwards. Nobody did anything careless. The library had five copies of one photo and no way of saying which was which.

FileSpin’s similar-image detection is built for this moment. Pick any image and FileSpin shows you the other copies of that same photo in your library, even when the file has been resized, cropped or flipped and saved under a new name. Each copy is labelled with how it changed, and you can copy the original’s information, such as the photographer and the licence, onto all of them in one edit.

How one photo becomes five files

Nobody sets out to fill a library with copies. It happens because a good photo is useful.

The agency delivers the master. The web team saves a smaller version for the homepage. The social team crops it to fit the feed. A designer flips it so the model faces the headline. A regional team exports their own version and uploads it as banner-de-final2.jpg. Every one of those files was made for a good reason, and each one goes back into the library as a separate asset that has no idea where it came from.

Three things that go wrong

The wrong version goes out. Search results don’t know which file is the master. People pick the one at the top, which might be the low-resolution web copy, the tight social crop or the flipped banner. The mistake usually shows up after it’s published.

People lose time checking. “Is this the latest one?” turns into a message to the designer, then to the brand manager, then a wait. Multiply that by every request in a busy week.

Information gets left behind. The original carries the photographer’s name, the licence and the date the licence ends. The copies usually carry nothing. When the licence runs out and someone archives the original, the copies stay in the library, still being picked for new work.

Why most duplicate checks miss these copies

Most media libraries have some kind of duplicate check, but it usually looks for the same file, not the same photo.

Many DAMs only flag a file that is identical byte for byte, or pixel for pixel. That catches someone uploading the exact same file twice. It doesn’t catch a copy that was resized, re-saved or cropped, because to a computer that is a different file.

File names don’t help either. hero.jpg, IMG_4471.jpg and banner-de-final2.jpg can all be the same picture.

Visual search (“find similar images”) answers a different question. It finds more pictures of the same kind of thing, such as other red trainers or other beach scenes. That’s useful for browsing, but here it’s the wrong answer. You don’t want more photos like this one. You want this photo, wherever it ended up.

How FileSpin spots the same photo

FileSpin looks at the picture itself, not the file or its name. It recognises a photo that has been resized, saved in another format, colour-corrected, mirrored, lightly cropped or stamped with a logo or text. It also knows the difference between a copy of a photo and a different photo of the same product. A second shot taken a moment later from another angle is not a copy, so it isn’t listed.

You don’t have to read percentages and guess what happened. FileSpin labels every result:

  • 97% similar: how close the copy is to the image you started from. The closest are listed first.

  • Mirrored: the copy is flipped.

  • Cropped or edited: part of the picture was cut away, or something was added.

If FileSpin already knows how two images are related, for example because one was made from the other inside FileSpin, it shows that too.

Similar images in the FileSpin dashboard: a mirrored copy at 100% similar, and a cropped and a rotated copy each labelled Cropped or edited

The Similar images view in FileSpin. Each copy shows how close it is and how it changed.

What it won’t find

An honest tool is easier to trust, so here is what it leaves out:

  • Heavily changed copies. A crop that cuts away half the picture, a photo dropped into a collage, or a background-removed cut-out is found only some of the time.

  • Other kinds of file. Videos and documents are matched only when they are exact copies.

  • The whole library at once. You check one image at a time, starting from the photo you care about. There is no single “every duplicate in the library” report.

For developers, the full detection rates, the API and what the AI Agent receives are in our guide Duplicates, Derivatives and Lineage.

From “found” to “fixed”

Finding the copies is only useful if you can do something about them. In FileSpin it takes a minute:

  1. Open the original and choose Find duplicates.

  2. Switch from Exact copies to Similar images.

  3. Select the copies and edit their details in one go, giving them the original’s photographer, licence and end date. If you change your mind, the dashboard keeps an undo for seven days.

  4. Add the copies nobody needs to a collection, so a person can review them before anything is removed.

You can also ask the FileSpin AI Agent: “Does the spring hero image have copies in the library?” It checks, tells you what it found and how each copy differs, and offers to copy the original’s details across or collect the copies for review. It does nothing until you say yes, and it never deletes files.

Edits you make inside FileSpin don’t add to the pile in the first place. Resizes and crops are served from the original rather than saved as new files, and an AI edit such as a background removal stays linked to the photo it came from.

Try it on ten images

You don’t need to clean up the whole library to find out whether you have this problem. Start with the images that matter most:

  1. Pick ten images in heavy use: the homepage hero, your best-selling product shots and this season’s campaign image.

  2. Check each one for similar images.

  3. Note every copy that’s missing its licence or usage details.

  4. Copy the details across, and collect the spare copies for review.

If ten images turn up even one stray copy, the rest of the library probably holds more.

If you’re choosing a media library, add one question to your list: “If I upload a cropped, flipped copy of one of my photos, will the system tell me where it came from?” The answer tells you a lot about how the system handles versions and rights.

Frequently asked questions

Can FileSpin find duplicate images with different file names?

Yes. It finds exact copies whatever their names, and similar copies too: resized, re-saved, colour-corrected, mirrored or lightly cropped versions of the same photo.

Will it match a different photo of the same product?

No. It looks for copies of the same photograph. A separate shot of the same product isn’t a copy. Very similar studio shots can occasionally appear, so check them before you change anything.

Can the AI Agent delete duplicates?

No. The agent can find copies, collect them for review and copy information onto them once you confirm. Deleting is always done by a person.

Your next media workflow

See where FileSpin fits.

Discuss the media you handle, the systems you use and the output your team needs. Then agree a useful first evaluation.

Your next media workflow

See where FileSpin fits.

Discuss the media you handle, the systems you use and the output your team needs. Then agree a useful first evaluation.

Your next media workflow

See where FileSpin fits.

Discuss the media you handle, the systems you use and the output your team needs. Then agree a useful first evaluation.