Compare Two Images and Highlight Changes
Compare two screenshots or renders pixel by pixel, count what changed and export a mask that marks every difference. Runs in your tab.
What this page does
Compares two images pixel by pixel, counts how many pixels differ by more than your tolerance, and draws a mask over them on a copy of the first image. Different sizes are padded to the larger canvas rather than refused.
How to use it
Choose a file, set the options and press Render. Example: two 800 × 600 screenshots that differ in one button becomes a mask PNG plus a text count of the changed pixels and their bounding box.
How it works
Both images are drawn at the top-left of a canvas sized to the larger of the two, with uncovered pixels treated as transparent. A pixel is a difference when any colour channel differs by more than the tolerance, and the mask paints those pixels in the highlight colour.
What it refuses
Fewer or more than two files, a zero-size image, a file that cannot be decoded, or a picture over the per-file cap is refused by name.
Where it runs
The picture is decoded and redrawn on a canvas in this tab: there is no upload, no network request and no model, so the pixels never leave your device. It opens PNG, JPEG, WebP, GIF, BMP and AVIF up to 16,000,000 pixels (16 MP) and 25 MB per file; anything larger is refused with its measured size rather than drawn blank.
Compares two images pixel by pixel, counts how many pixels differ by more than your tolerance, and draws a mask over them on a copy of the first image. Different sizes are padded to the larger canvas rather than refused.
Both images are drawn at the top-left of a canvas sized to the larger of the two, with uncovered pixels treated as transparent. A pixel is a difference when any colour channel differs by more than the tolerance, and the mask paints those pixels in the highlight colour. The picture is decoded and redrawn on a canvas in this tab: there is no upload, no network request and no model, so the pixels never leave your device.
Worked example: two 800 × 600 screenshots that differ in one button becomes a mask PNG plus a text count of the changed pixels and their bounding box.
How to use the compare two images and highlight changes
Choose 2 to 2 image files, set the options, and press Render. The result appears below the form.
Example: two 800 × 600 screenshots that differ in one button becomes a mask PNG plus a text count of the changed pixels and their bounding box.
The picture is decoded and redrawn on a canvas in this tab: there is no upload, no network request and no model, so the pixels never leave your device.
- Tolerance per channel (0-255) — 0 to 255
- Highlight colour (hex) — a hex colour
- Show under the mask — one of first image, white, black
- Output format — one of png, jpeg, webp
- Quality (JPEG and WebP) — 1 to 100
What the canvas draws
Compares two images pixel by pixel, counts how many pixels differ by more than your tolerance, and draws a mask over them on a copy of the first image. Different sizes are padded to the larger canvas rather than refused.
Both images are drawn at the top-left of a canvas sized to the larger of the two, with uncovered pixels treated as transparent. A pixel is a difference when any colour channel differs by more than the tolerance, and the mask paints those pixels in the highlight colour.
- Every option is declared by this page, so an unknown choice or an out-of-range number is refused by name before any pixel is drawn.
- The preview and the download are the same canvas: what you see is the file you get.
What it refuses, and what it leaves out
Fewer or more than two files, a zero-size image, a file that cannot be decoded, or a picture over the per-file cap is refused by name.
It opens PNG, JPEG, WebP, GIF, BMP and AVIF up to 16,000,000 pixels (16 MP) and 25 MB per file; anything larger is refused with its measured size rather than drawn blank.
- Not the right page for: comparing different-sized images as if they were aligned
- Not the right page for: a semantic diff of the DOM or the text
What to Expect
Best for
- catching a visual regression after a CSS change
- checking that an export matches a design
- spotting what changed between two renders
Not ideal for
- comparing different-sized images as if they were aligned
- a semantic diff of the DOM or the text
What this tool keeps
- Every differing pixel's position
- A count and a bounding box for the change
- The first image, readable under the mask
What may need cleanup
- Ignoring anti-aliasing by raising the tolerance
- Cropping both images to the same region for a meaningful count
Common errors
- A tolerance of 0 compares exact bytes
- Images of different sizes are padded, so the padding itself counts as changed
- JPEG re-encoding can move pixels by a few levels; raise the tolerance
Example use cases
- A before/after of a UI fix
- Verifying an image export
Sample input
two 800 × 600 screenshots that differ in one button
Sample output
a mask PNG plus a text count of the changed pixels and their bounding box
Who this is for
- Front-end developers reviewing a visual change
- QA engineers checking a render
Frequently asked questions
Does my image get uploaded anywhere?
The picture is decoded and redrawn on a canvas in this tab: there is no upload, no network request and no model, so the pixels never leave your device.
What is the maximum image size?
It opens PNG, JPEG, WebP, GIF, BMP and AVIF up to 16,000,000 pixels (16 MP) and 25 MB per file; anything larger is refused with its measured size rather than drawn blank. The limit exists because a browser tab has to hold the decoded bitmap in memory.
What happens if the file cannot be decoded?
The page refuses it by name — unsupported type, decode failure, zero pixels, or above the pixel cap — and shows the measured size rather than drawing a blank canvas. The file you chose stays selected so you can correct one option and try again.
Why do two identical-looking screenshots show differences?
Compression and anti-aliasing move pixel values by a few levels. Raise the tolerance to 16 or 24 and the noise disappears while real changes remain.
What happens if the images are different sizes?
Both are placed at the top-left of a canvas as large as the bigger one, and the uncovered area is treated as transparent, so it counts as changed. That is deliberate: two renders of different sizes are not the same image.
Build custom tools for your store
Join the SDK program to create useful ecommerce utilities for your store.