Skip to content
All tools

Compare two images

Are they the same picture? Drop two images to get a similarity score, a check for exact copies, and a map of exactly where they differ.

6%

0% marks every pixel that changed at all, including compression noise you can’t see. 6% ignores that noise and keeps real changes.

Outline each changed area Boxes the places where the content changed, not just the pixels, so a small edit can’t hide in compression noise.

Bytes of your files sent anywhere: 0 B

How to use it

  1. 01

    Add your files

    Drop the two images you want to compare. The first one you add is the reference.

  2. 02

    Check the settings

    Choose how small a difference to ignore. The default skips invisible compression noise.

  3. 03

    Save the result

    Get the verdict, drag the divider between the two, and save the difference map.

Why it doesn’t upload

The images people compare are often private: a contract before and after, ID photos, unreleased designs. The comparison is arithmetic on pixels your browser has already decoded, so there is nothing a server could add.

How to check for yourself

Worth knowing

  • If the two images are different sizes, the second is scaled to the first’s frame before comparing. A crop or a change of shape therefore shows up as a difference everywhere.
  • Images larger than 2,048 pixels across are compared at that size. An exact pixel-for-pixel check still runs at full size for images up to 16 megapixels.
  • The comparison is of what the images look like, so two files can be “pixel for pixel identical” and still differ in metadata or file size.

Three answers, in order

First, are they the same file? Both files are fingerprinted with SHA-256. If the fingerprints match, one is an exact copy of the other and there is nothing more to say.

Second, is every pixel the same? Files can differ while the picture doesn’t: a PNG saved from a PNG, or a copy with its metadata stripped. At full resolution every pixel of one is checked against the other.

Third, how different are they, and where? The difference map shows the reference image faded to grey, with every pixel that changed by more than the tolerance painted red. Separately, each 16-pixel patch is checked for whether its structure changed. Compression and resizing bend pixel values but keep shapes, so they don’t count. New text or a removed object breaks the shape, and those patches are grouped into the outlined “changed areas”.

The percentage is the overall structural similarity (SSIM). It is the verdict only when nothing stands out locally. An edited invoice can be 99% similar overall and still have one changed number, which is why changed areas come first.

Reading the difference map

Scattered red specks all over a photo usually mean compression: the same picture saved at a different quality. Raise the tolerance a few percent and they disappear. A solid red patch or a box around one area means something in the picture actually changed there.

Red outlines along every edge mean the second image is slightly shifted, scaled or sharpened. A picture that is entirely red is a different picture, or the same scene from a different moment.

Questions

How do I check if two images are exactly the same?

Drop both here. The result first checks whether the files are byte-for-byte identical, then whether every pixel is the same, which catches copies saved with different metadata or in a lossless format. If neither, it measures how similar they look.

What does the similarity score mean?

It is SSIM (structural similarity), a standard measure that compares brightness, contrast and texture in small patches the way eyes do. 100% is identical; above about 99% differences can’t be seen; below about 95% you can see them without looking hard.

Why are two copies of the same photo “different”?

Because saving a JPG again, uploading it to a chat app or resizing it changes pixel values slightly, even when you can’t see it. That is what the tolerance setting is for: at 6% that noise is ignored and only real changes are marked.

Can it find what was edited in a photo?

Yes, if you have the original. Changed areas are found by structure rather than raw pixel values, so an altered number, a removed object or a retouched patch is boxed even when the rest of the picture was re-compressed.

Can I compare a JPG with a PNG, or a HEIC with a JPG?

Yes. Both images are decoded to pixels first, so the formats don’t need to match. Comparing an iPhone HEIC with the JPG made from it is a quick way to see what the conversion cost.

Other tools