Image Diff
Pixel-by-pixel comparison of two images with adjustable tolerance. The canonical visual-regression QA tool — runs in your browser because the screenshots are private.
Original (A)
Drop image A
JPG, PNG, WebP — never uploaded
Changed (B)
Drop image B
JPG, PNG, WebP — never uploaded
How to compare two images online
Drop the two images you want to compare. ToolChop decodes each into a canvas locally, scales them to a common size if needed, and walks every pixel computing the Euclidean RGB distance. Pixels that differ by more than your chosen tolerance are highlighted in red on a dimmed version of image A — so you see what changed, in context. The header reports the exact differing-pixel count and percentage.
Why a local diff matters
Image diff is the standard tool for visual-regression QA — and the screenshots you diff almost always contain something private: internal UI, customer data on screen, pre-release features, signed-in account details. Uploading them to a third-party service for the simple operation of comparing pixels is exactly the data leak you are trying to avoid. ToolChop runs the diff entirely in your browser. You can confirm in DevTools → Network that no request fires.
Picking the right tolerance
- 0 — strictest. Any color shift counts. Catches sub-pixel rendering changes.
- 5–15 — typical for QA. Ignores anti-aliasing and font-hinting noise; flags real visual changes.
- 15–30 — for comparisons across JPEG re-encoding. Ignores compression artifacts.
- 60+ — only flags clearly different colors (red vs blue, not red vs slightly-darker-red).
What you can do
- Pixel-by-pixel diff with adjustable color-distance tolerance
- Differing-pixel count, percentage, and red overlay on dimmed background
- Toggle between overlay, side-by-side, A-only, and B-only views
- Download the diff overlay as a PNG for bug reports
- Handles mismatched dimensions by scaling to the larger of the two
Frequently asked questions
How do I compare two images online for free?
Drop image A on the left and image B on the right. ToolChop diffs them pixel-by-pixel in your browser, highlights every differing pixel in red on a dimmed background, and reports the exact count and percentage of pixels that differ. Adjust the Tolerance slider to ignore minor compression artifacts. No account, no upload, no daily limit.
Does ToolChop upload my images?
No. Both images are decoded into canvases in your browser and diffed locally. The files never leave your device — important when the images are screenshots of internal dashboards, UI mockups, customer-facing surfaces, or any visual content you should not casually share.
Why is the privacy story for an image diff important?
Image diff is the canonical QA tool for visual regression: comparing a 'before' screenshot of your app against an 'after' screenshot to spot unintended changes. Those screenshots contain internal UI, customer data, signed-in account details, and pre-release feature surfaces. Uploading them to a third-party diff service is a needless leak. ToolChop runs the diff entirely in your browser.
What does the Tolerance slider do?
It is the maximum Euclidean RGB distance two pixels can have and still count as 'unchanged'. 0 means any color shift counts as a difference. 15 catches what the human eye notices but ignores typical JPEG compression noise. 30 ignores moderate compression and anti-aliasing artifacts. 60+ only flags clearly different colors (e.g. red vs blue).
What if the two images have different dimensions?
ToolChop scales both to the larger of the two dimensions using smooth resampling, then diffs them at that size. A warning banner makes this clear. If you need exact-dimension diffing, resize them to match first using our Image Resizer.
Can I use this for visual regression testing?
Yes — that is the main use. Drop a baseline screenshot and a current screenshot, set Tolerance to 5–15 to ignore antialiasing noise, and the percentage and pixel count tell you exactly what changed. Download the overlay PNG to attach to a bug report.
How is the 'differing pixels' percentage calculated?
(Pixels where color distance > tolerance) ÷ (total pixels in the comparison canvas) × 100. A 0.05% number means 1 in 2,000 pixels differ — usually just font hinting or sub-pixel rendering. A 5% number is a real change that humans will notice.
Does the diff handle anti-aliasing well?
Set Tolerance to 10–20 to ignore typical anti-aliasing differences (1-pixel font edge shifts, slight color blending). At Tolerance 0 anti-aliasing will dominate the diff for any text-heavy screenshot, which is rarely what you want.
What does the red overlay actually show?
Red pixels are pixels whose color distance between A and B exceeded the tolerance. The unchanged pixels are kept as a dimmed grayscale version of A, so the red diff stands out clearly without losing context of where in the image the change is.
What about images with transparency?
Transparent regions are composited onto white before diffing, so a partial alpha change shows up as a color change. If you need to diff with alpha awareness specifically, that is a more specialized tool — most QA use cases want the visual-equivalent comparison ToolChop provides.
Is there a file size limit?
Only your browser's memory. ToolChop comfortably diffs 4K-resolution images. The diff itself is one pass over the pixel buffer — fast even at high resolutions.
Why use ToolChop instead of an online diff service that uploads my images?
The screenshots people diff are almost always internal: customer data on the page, signed-in account details, pre-release features, internal dashboards. Uploading them to a third-party service for the simple operation of comparing pixels is exactly the data path you should avoid. ToolChop diffs entirely in your browser — DevTools → Network confirms no request fires.