Resolve multiple image conflicts at the same time

Usage no npm install needed!

<script type="module">
  import multiImageMergetool from '';


multi-image-mergetool Build status

Resolve multiple image conflicts at the same time

This was built to resolve conflicts in full page test screenshots where auxilary content may have changed (e.g. navigation). We prefer full page screenshots over ignored content/component based screenshots as it is less brittle and catches unexpected issues.


  • Find/resolve common image changes via overlay selection
  • Support for any image configuration (e.g. same directory, same name/different directory, Gemini, specific names)



Screenshot from tests

Getting Started

Install the module via:

# Install multi-image-mergetool globally
npm install -g multi-image-mergetool

# Run multi-image-mergetool against a Gemini test suite
multi-image-mergetool --loader gemini
# Comparing images...
# ✘ gemini/screens/root/default-small/Chrome.png
# ✘ gemini/screens/root/default-medium/Chrome.png
# ✓ gemini/screens/root/default-large/Chrome.png
# ✓ gemini/screens/root/z-default-large2/Chrome.png
# ✓ gemini/screens/root/z-default-large3/Chrome.png
# Images matched: 3 of 5
# Server is listening on http://localhost:2020/

# Browser window will automatically be opened

# Alternatively compare one-off images by their paths
multi-image-mergetool \
    --current-images path/to/current1.png path/to/current2.png \
    --ref-images path/to/ref1.png path/to/ref2.png

# Optionally define custom diff paths
multi-image-mergetool \
    --current-images path/to/current1.png path/to/current2.png \
    --ref-images path/to/ref1.png path/to/ref2.png \
    --diff-images path/to/diff1.png path/to/diff2.png


Support this project and others by twolfson via donations



Our CLI supports the following options:

$ multi-image-mergetool --help
multi-image-mergetool [options] --current-images <current-images...> --ref-images <ref-images...>

  --current-images   Current images for comparison (required if no --loader)  [array]
  --ref-images       Locations to load/save reference images (required if no --loader)  [array]
  --diff-images      Locations to save diff images  [array]
  --loader           Loading mechanism to find images  [string] [choices: "gemini"]
  --assert           Only perform assertions (no server). If all images match, exits with 0. Otherwise, exits with 1  [boolean]
  --port, -p         Port for server to listen on  [number] [default: 2020]
  --hostname, -h     Hostname for server to listen on  [string] [default: "localhost"]
  --verbose          Enable verbose logging  [boolean]
  --no-browser-open  Prevent browser window from opening automatically  [boolean]
  --version          Show version number  [boolean]
  --help             Show help  [boolean]

  Load from paths:
    multi-image-mergetool --current-images current1.png current2.png --ref-images ref1.png ref2.png
    multi-image-mergetool --current-images current1.png current2.png --ref-images ref1.png ref2.png --diff-images diff1.png diff2.png
  Load from `gemini` and `gemini-report` folders:
    multi-image-mergetool --loader gemini


  • The --current-images, --ref-images, and --diff-images must be the same length and order
    • It's okay if the ref images and diff images don't exist yet, we will show prompts in the UI to handle new images

Architecture choices

We chose to use a server/browser implementation over a desktop application (e.g. Electron) for more flexibility with little development cost. It allows us to support virtualized environments (e.g. Vagrant, Docker) without asking our users to bend over backwards.



This repository was initially inspired by gemini-gui so we want to maintain support for a similar setup. To get Gemini set up, run the following:

# Generate `gemini` and `gemini-report`
gemini-test --reporter html

# Use `multi-image-mergetool` with `gemini` folders
multi-image-mergetool --loader gemini


This isn't efficient due to comparing images twice and waiting for all screenshots to be taken before starting comparisons. Ideally we wrap multi-image-mergetool with a better integration or add data persistence to this library (unlikely). If you write a wrapper, please submit a PR updating this documentation.



To get a local copy of multi-image-mergetool set up, run the following steps:

# Clone the repository
git clone
cd multi-image-mergetool

# Install our dependencies
npm install

# Compile our latest assets
npm run build

# Start our local multi-image-mergetool
npm run start-gemini
# or bin/multi-image-mergetool

Building files

We use gulp and browserify to compile assets. These can be run once via:

npm run build

or continuously via:

npm run develop

npm run develop will additionally start a LiveReload server


We have test suites set up on Mocha and Karma. They can be run via:

npm test

To exclusively run the browser or server tests, use:

npm run test-browser # Browser tests
npm run test-server # Server tests

# Additional helpers
npm run test-karma-develop # Watches/runs Karma tests on PhantomJS and asserts screenshots

Additionally, during testing we generate screenshots. We can assert and approve they are as expected via:

npm run test-browser-mergetool-screenshots


Our release process has a bunch of steps so we use foundry to make them reproducable. To perform a new release, run the following:

  • Update with new release version
  • Run foundry release <version>
    • This will take care of compiling browser-dist, updating .npmignore to match .gitignore, tagging repository, updating package.json, pushing commit/tag to GitHub, and publishing to npm


In lieu of a formal styleguide, take care to maintain the existing coding style. Add unit tests for any new or changed functionality. Lint via npm run lint and test via npm test.

Prior art

multi-image-mergetool was inspired by:


As of Oct 22 2016, Todd Wolfson has released this repository and its contents to the public domain.

It has been released under the UNLICENSE.