Load a before and after image, drag the divider to reveal each version, or switch to a pixel comparison that paints every changed area in red. The threshold slider lets you ignore compression noise and keep only edits that a person can actually see.
It works for design revisions, photo retouching, screenshot regression checks and any pair of files that should look the same but somehow do not. All processing runs locally in your browser. Your data is not uploaded to any server. 100% Local
Every pixel comparison has to decide what counts as different. At threshold zero, a JPEG saved twice or a file converted between colour profiles will light up almost everywhere, because recompression nudges values by one or two units. Around eight to fifteen, that noise disappears and real edits remain: a moved button, a new caption, a colour change in a logo. If a client insists nothing changed, a low threshold plus the changed-pixel count is a useful conversation starter; if you only need to know whether the layout shifted, a higher value is faster to read.
Use the same dimensions and the same crop for both files whenever possible. When aspect ratios differ, both images are fitted into one canvas with letterboxing, so only the overlapping region is genuinely compared and the empty bands can look like differences at low thresholds. Export both versions from the same tool with identical quality settings, and avoid comparing a PNG against a JPEG of the same artwork, because the compression artefacts alone will dominate the result. For screenshots, capture at the same zoom level and the same window size.
A designer sends version four and asks whether anything moved. Instead of opening both files and squinting, load them here and switch to differences only. Padding changes, shifted icons and recoloured buttons appear immediately as red shapes, and the pixel count gives a rough sense of how large the change is. This is also useful when reviewing your own work after a long break, because the eye tends to normalise small drifts that a pixel map exposes instantly.
Retouchers and clients often want to know exactly what was removed: a blemish, a stray hair, a logo on a shirt. A high threshold hides grain and shows only the healed areas. Keep the threshold moderate, though, because aggressive retouching with texture rebuilding can leave soft differences that a strict threshold would flag as noise. Comparing the final export against the approved version is also a practical way to confirm that nothing else was altered during the last save.
Front-end teams capture screenshots in tests and compare builds. A diff view answers the first question quickly: did anything change at all? If the answer is yes, differences-only mode shows whether the change is confined to one component or spread across the page. Font loading, image lazy-loading and animation timing can all produce false positives, so capture screenshots after the page settles, and raise the threshold slightly to ignore anti-aliasing on text edges.
After compressing a photo or converting it to another format, compare the result with the original to judge quality. At a low threshold the changed-pixel count will be large even for a good encode, because every block of a lossy format differs slightly. What matters is where the differences cluster: flat sky should stay quiet while edges and fine texture may shift. If the map is bright in areas that should be smooth, the quality setting was too aggressive and the file should be re-exported.
The threshold decides how large a per-pixel color change must be before it counts as different. A low value catches tiny shifts from recompression or anti-aliasing; a high value filters that noise and leaves only visible edits such as moved objects or changed text.
JPEG compression, color profile conversion, resizing and even a one-pixel offset change individual pixel values. Raise the threshold until the noise disappears, or export both files again from the same source with identical settings before comparing.
It hides both photos and paints only the changed pixels in red on a dark background. This is the fastest way to see where two layouts, screenshots or edited photos diverge, especially when the change is small and easy to miss in a slider view.
Not strictly, but the comparison is most accurate when they match. Both images are fitted into a shared canvas and letterboxed, so different aspect ratios leave empty bands and the overlapping area is what actually gets compared.
No. Both files are decoded locally in the browser, drawn onto canvases, and compared pixel by pixel in JavaScript. Nothing leaves your device, which matters when the images are unreleased designs or private documents.