Visual Regression Tracker

Visit Website

Introduction

Visual Regression Tracker (VRT) is an ‘open source, self hosted solution for visual testing and managing results of visual testing’, according to its README. The service receives images, performs a comparison with the previously accepted baseline, and gives immediate results so unexpected changes are caught.

Its main selling point is independence. VRT is automation-framework independent, so teams integrate it with the tools they already use, and platform independent, so it works for web, mobile, desktop or anything else that can produce a screenshot. SDKs and agents exist for JavaScript, Java, Python and .NET, and any other language can use the REST API. There are integrations for Playwright, Cypress, Selenium-based stacks and other frameworks through these SDKs.

The README lists baseline history to see how each baseline image changed over time, ignore regions to improve stability by excluding unimportant or uncontrollable parts of an image, Docker-based setup with everything inside images, self-hosting to keep data inside your intranet, and a standalone JAR for comparing PDFs and folders of images.

Comparison is configurable per project. VRT supports Pixelmatch for fast pixel-by-pixel comparison (the default), Looks-Same for perceptual comparison that tolerates minor rendering differences, Odiff with layout detection and anti-aliasing support, and a VLM option that combines Pixelmatch with Vision Language Model analysis of whether differences are noticeable to humans.

Results are organised as test variations, each a history of baselines identified by name, branch, operating system, browser, viewport and device, so the same screen can have different baselines across environments. Reviewers approve or reject differences in the web UI. The project released 5.9.0 in October 2026 and had commits in November 2026.

In the quality loop VRT sits at the end-to-end and release stages for teams that want visual checks without sending screenshots to a SaaS provider.

The source code is at Visual-Regression-Tracker/Visual-Regression-Tracker on GitHub under the Apache-2.0 licence, with about 726 stars; the most recent commit was on 8 November 2026.

Choosing VRT is mainly a decision about ownership. Commercial services handle storage, scaling and the review interface for a monthly fee, while VRT gives you those pieces as containers that you deploy, secure and back up. For organisations that cannot send screenshots of internal applications to a third party, or that want visual checks for desktop or embedded interfaces that SaaS tools do not support well, that trade is often worth it. Teams should plan storage for baseline history, decide who approves changes, and agree how branch baselines merge back into the main branch before rolling VRT out widely. A short pilot on one application, with a few dozen screenshots, is usually enough to judge the effort involved.

Features

1. Framework independent: Use with any automation tool.
2. Platform independent: Web, mobile, desktop and more.
3. Baseline history: Track how baselines changed.
4. Ignore regions: Exclude unstable areas.
5. Multiple engines: Pixelmatch, Looks-Same, Odiff and VLM.
6. SDKs: JavaScript, Java, Python and .NET, plus REST API.
7. Branch-aware variations: Baselines per branch, browser, viewport and device.
8. Self-hosted: Data stays in your network.
9. Docker setup: Everything runs from images.
10. PDF comparison: Standalone JAR for PDFs and folders.

Pros

1. Free and self-hosted: No per-screenshot costs.
2. Flexible comparisons: Choose an engine per project.
3. Free to use: Open source under the Apache-2.0 licence, so the core tool costs nothing to use.
4. Actively maintained: The repository had commits as recent as 8 November 2026.

Cons

1. You run it: Hosting, upgrades and storage are your job.
2. Smaller community: About 730 GitHub stars.
3. UI less polished: Compared with commercial tools.

Usage and Links (Example)

# Start the service with Docker Compose (see the README for the current files)
git clone https://github.com/Visual-Regression-Tracker/Visual-Regression-Tracker.git
cd Visual-Regression-Tracker
docker compose up -d

// Playwright example with the JS SDK
import { VisualRegressionTracker } from "@visual-regression-tracker/sdk-js";
const vrt = new VisualRegressionTracker({ apiUrl: "http://localhost:4200", project: "web", apiKey: process.env.VRT_KEY, branchName: "main" });
await vrt.start();
await vrt.track({ name: "home", imageBase64: (await page.screenshot()).toString("base64") });
await vrt.stop();

Practical tips:

1. Use ignore regions for dates, ads and animated content.
2. Pick Looks-Same or Odiff if pixel comparisons are too noisy across machines.
3. Run screenshots in the same Docker browser image locally and in CI.

Resources:

• Visual Regression Tracker website

Pricing

Visual Regression Tracker is free and open source under the Apache-2.0 licence.

Recommended for?

1. Teams needing self-hosted visual testing.
2. Polyglot organisations using several automation frameworks.
3. Not a fit if you want a fully managed service.

Conclusion

Visual Regression Tracker is a capable, actively maintained self-hosted alternative to commercial visual testing platforms. Its framework independence and choice of comparison engines are real strengths.

Budget time for hosting and for stabilising screenshots, as with any visual testing tool.

On CheckOps, Visual Regression Tracker is filed under the End-to-end & functional, Prototype & UX testing and Release & CI gates stages. Related tools listed here include Argos, Percy, Applitools, BackstopJS, reg-suit and pixelmatch. Before choosing, compare them on licence, supported languages and platforms, hosting model, and how well each fits your existing CI pipeline. Facts on this page were checked against the official website, documentation and repository on 11 November 2026; statements the vendor makes about its own product are marked as vendor claims.