Compare two images
Two pictures next to each other, a map of what changed, or a number.
Your file
Drop a file here, or click to choose
It stays on this device. Nothing is sent anywhere.
The first file is the reference, this one is what it gets measured against.
Settings
What each mode is for
- Side by side is for judging quality settings, where seeing both at once beats remembering the first.
- Difference blacks out everything identical and lights up everything changed. This is how you spot a watermark, a retouch, or a compression artefact you could not otherwise see.
- Score gives a single similarity figure, useful for checking whether an export actually changed anything.
Same size helps
Images are scaled to a shared canvas, so different sizes still compare, but a difference map is only precise when both start at the same dimensions. Resize one to match the other first if you need exact results.
Reading the numbers
The metrics mode returns two figures, and they measure different things.
PSNR is peak signal-to-noise ratio, in decibels, and it is essentially a measure of raw pixel error. Above about 40 dB the two images are visually identical for practical purposes. Between 30 and 40 dB a difference exists and most people will not spot it without flipping between them. Below 30 dB it is visible. The scale is logarithmic, so a few decibels matter more than they look like they should, and infinity means the files are pixel-for-pixel identical.
SSIM is structural similarity, scored from 0 to 1, and it tries to model what a person would actually notice rather than counting errors. It weights structure and contrast over absolute pixel values. Above 0.98 is generally indistinguishable; below 0.95 something is visibly going on. SSIM is the better guide of the two when you are judging compression, because it does not punish an encoder for changes the eye ignores.
The amplify control
Raw differences between two versions of the same image are almost entirely black, because the changes are a handful of levels out of 256. The amplify setting multiplies the brightness of the difference so those small numbers become visible. At 4 you can see compression artefacts clustering around edges. At 10 and above you will see noise the encoder introduced in areas that look perfectly smooth in both originals. Neither is a fault in the tool; you are looking at real differences the eye normally has no chance of resolving.
What this is actually good for
- Settling a quality argument with yourself. Export at quality 85 and quality 95, compare, and find out whether the larger file bought anything.
- Checking that a conversion did nothing. A lossless operation should score near-infinite PSNR. If it does not, something re-encoded when you thought it was copying.
- Finding a watermark or a retouch by putting the two versions in difference mode, where anything changed lights up and everything untouched stays black.
- Verifying an upload survived. Platforms re-encode what you give them. Compare what you sent with what came back and you can see exactly what they did to it.
For the same job on video, compare videos works frame by frame.
Common questions
The difference is completely black.
The two images are identical as far as the pixels are concerned.
Can it tell me if a photo was edited?
It can show you where two versions differ. It cannot analyze a single image on its own for signs of editing.