Node.js library for comparing PNG-images, taking into account human color
perception. It is created specially for the needs of visual regression testing
gemini utility, but can be used
for other purposes.
var looksSame = ;;
Parameters can be paths to files or buffer with compressed
By default, it will detect only noticeable differences. If you wish to detect any difference,
You can also adjust the ΔE value that will be treated as error in non-strict mode:
tolerance in non-strict mode is 2.3 which is enough for the most cases.
tolerance to 0 will produce the same result as
strict: true, but strict mode
Attempt to set
tolerance in strict mode will produce an error.
Some devices can have different proportion between physical and logical screen resolutions also
pixel ratio. Default value for this proportion is 1.
This param also affects the comparison result, so it can be set manually with
Comparing images with ignoring caret
Text caret in text input elements it is a pain for visual regression tasks, because it is always blinks. These diffs will be ignored by default. You can use
ignoreCaret option with
false value to disable ignoring such diffs. In that way text caret will be marked as diffs.
ignoreCaret can be set independently of one another.
Comparing images with ignoring antialiasing
Some images has difference while comparing because of antialiasing. These diffs will be ignored by default. You can use
ignoreAntialiasing option with
false value to disable ignoring such diffs. In that way antialiased pixels will be marked as diffs. Read more about anti-aliasing algorithm.
Sometimes the antialiasing algorithm can work incorrectly due to some features of the browser rendering engine. Use the option
antialiasingTolerance to make the algorithm less strict. With this option you can specify the minimum difference in brightness (zero by default) between the darkest/lightest pixel (which is adjacent to the antialiasing pixel) and theirs adjacent pixels.
We recommend that you don't increase this value above 10. If you need to increase more than 10 then this is definitely not antialiasing.
Getting diff bounds
Looksame returns information about diff bounds. It returns only first pixel if you passed
stopOnFirstFail option with
true value. The whole diff area would be returned if
stopOnFirstFail option is not passed or it's passed with
Getting diff clusters
Looksame returns diff bounds divided into clusters if option
shouldCluster passed with
true value. Moreover you can pass clusters size using
Building diff image
Building diff image as a Buffer
If you don't want the diff image to be written on disk, then simply don't
diff: path to the
createDiff method. The callback will then
Buffer containing the diff as the 2nd argument.
If you just need to compare two colors you can use