Open source visual regression testing tools
The open-source visual regression testing tools still worth using in 2026 — Playwright, BackstopJS, reg-suit, Loki, Argos, jest-image-snapshot and pixelmatch — with licences, maintenance status, and what each one does not do.
Open source removes the licence cost from visual regression testing. It does not remove the work — it moves it onto your team, which is the trade that matters when you are deciding. This is the current state of the maintained options, plus honest notes on the two that are no longer safe defaults.
Quick comparison
| Tool | Licence | Captures? | Compares? | Review UI | Status |
|---|---|---|---|---|---|
| Playwright | Apache 2.0 | Yes | Yes | HTML report | Actively developed |
| BackstopJS | MIT | Yes | Yes | HTML report | Seeking a maintainer |
| reg-suit | MIT | No | Yes | Static report | Actively maintained |
| Loki | MIT | Yes (Storybook) | Yes | CLI approval | Actively maintained |
| Argos | Open source | Via your tests | Yes | Hosted UI | Actively developed |
| jest-image-snapshot | Apache 2.0 | No | Yes | None | Maintained |
| pixelmatch | ISC | No | Yes (library) | None | Maintained |
| Lost Pixel | MIT | — | — | — | Archived Apr 2026 |
The tools
Best for: Almost everyone starting from zero in 2026.
- License
- Apache 2.0
- Pricing
- Free
If you have no existing investment, start here and stop reading. It is the only option on this list where capture, comparison, retry logic, and reporting all ship together and are all actively developed. Setup details in our Playwright guide.
- Capture and comparison in one tool, with no assembly required
toHaveScreenshot()retries until the rendered image is stable before diffing- Backed by Microsoft with a fast release cadence
- Built-in HTML reporter shows expected, actual, and diff side by side
- A spec per page or component, authored and maintained by you
- Platform-specific baselines (
name-chromium-linux.png) mean local and CI images diverge - Committed PNGs grow the repository
Best for: Existing installations, and small page-level suites where you accept the maintenance risk.
- License
- MIT (7.2k stars)
- Pricing
- Free
For a decade this was the answer to this question. The code still works and 7.2k stars of ecosystem is real. But the maintainer signal in the README is not something to discover after you have built a 300-scenario suite on it — factor it into the decision now, especially for a new project.
- The most mature open-source tool for whole-page regression testing
- Capture via Puppeteer, or Playwright for Firefox and WebKit
- Docker rendering keeps local and CI results consistent
- Scenario config is declarative JSON, which is easy to generate programmatically
- The README states the project needs a new maintainer or owner
- One hand-written scenario per page
- No collaborative review or approval
Best for: Teams that already produce screenshots and want Git-aware baselines without a vendor.
- License
- MIT (1.3k stars)
- Pricing
- Free, plus your own S3 or GCS storage costs
The most commonly misread tool in this category. reg-suit is the comparison and reporting half of a visual testing system, deliberately so. Pair it with Playwright or Puppeteer for capture and you get Git-aware baselines that never enter your repository — which is a genuinely better architecture than committing PNGs.
- Actively maintained
- Capture-agnostic: works with images from any source
- Baselines keyed to Git history rather than to a directory you must commit
- Publishes reports to S3 or GCS and notifies GitHub, GitLab, and Slack
- Does not capture screenshots — this is a comparer only
- You assemble capture, storage, and CI wiring yourself
- Review is a static HTML report, not a collaborative approval flow
Best for: Storybook component libraries that want zero vendor spend.
- License
- MIT (1.9k stars)
- Pricing
- Free
Effectively the open-source counterpart to Chromatic's component coverage, without the hosted review workflow. A good fit for a library team comfortable approving baselines from a terminal. See Storybook visual regression testing for how it compares to the current Vitest-based approaches.
- Purpose-built for Storybook with minimal configuration
- Chrome via Docker, AWS Lambda, or locally
- Also targets iOS simulators and Android emulators
- Docker rendering makes results reproducible across operating systems
- Storybook only — it cannot test a real route
- Approval happens at the CLI, with no review UI
- Coverage is limited to the stories someone wrote
Best for: Teams that want an open-source core but a real review UI.
- License
- Open source, with a hosted commercial tier
- Pricing
- Hobby free 5,000 screenshots/mo · Pro $100/mo for 35,000
Included here because the core is open source, but be clear-eyed: the thing that makes Argos better than BackstopJS — the review workflow — is the hosted product. As a free tool it is a 5,000-screenshot tier; as a $100/mo tool it is the value pick of the hosted category.
- The only entry here with a collaborative approval UI
- Capture runs locally in your own browser, so rendering matches CI
- Direct Playwright and Storybook integrations
- Self-hosting remains possible
- The useful tier is the hosted paid one, so this is only partly a free option
- You still write the tests that produce the screenshots
- Smaller ecosystem than the incumbents
Best for: Jest suites that already produce images and want a matcher.
- License
- Apache 2.0 (3.9k stars)
- Pricing
- Free
A matcher rather than a tool. Useful glue when Puppeteer or Playwright is already producing buffers inside Jest tests, and largely superseded by Playwright's own assertions for new work.
- Drops into an existing Jest suite as
toMatchImageSnapshot() - Configurable failure thresholds and diff output
- Maintained by American Express
- Comparison only — you supply the screenshot
- Inherits Jest's snapshot-approval habits, where
-ubecomes reflexive - No review UI and no cross-platform baseline handling
Best for: Building your own comparison step.
- License
- ISC (6.9k stars)
- Pricing
- Free
Worth knowing about because it is the primitive underneath several tools on this list, including Playwright's own comparison. Reach for it when you are building something bespoke, not when you want visual testing today.
- Tiny, fast, zero-dependency image diffing
- Anti-aliasing detection built in
- The library Playwright itself uses under the hood
- A library, not a tool — no capture, storage, baselines, or reporting
- Everything around the diff is yours to build
What open source does not give you
Every tool above solves capture, comparison, or both. None of them solves the three things that actually determine whether a visual suite is still running in a year:
- Coverage that follows the product. Every option here requires a human to add a scenario, spec, or story for each new page. Coverage decays at exactly the rate team discipline does.
- Determinism. Fonts, animation, clocks, and live data are your problem in all seven. See fixing flaky visual tests for the work involved.
- Change targeting. None of them map a diff to the routes it can affect, so you either recapture everything or maintain that mapping by hand. Visual testing in CI/CD covers the alternatives.
That is not an argument against open source — for a stable design system with fifty stories, Playwright or Loki is genuinely the right answer and paying for a platform would be waste. It is an argument for counting the maintenance hours honestly before concluding that free is cheaper.
Frequently asked questions
What is the best open source visual regression testing tool?
Playwright, for most teams starting in 2026 — it is the only option where capture, comparison, retry logic, and reporting ship together and are all actively developed. reg-suit is the better choice if you already produce screenshots and want Git-keyed baselines, and Loki if your coverage lives entirely in Storybook.
Is BackstopJS still maintained?
The project is MIT-licensed with around 7.2k stars and the code works, but its README currently asks for a new maintainer or owner. That is worth weighing before building a large new suite on it, though it remains reasonable for existing installations and small page-level coverage.
What happened to Lost Pixel?
The Lost Pixel repository was archived on 22 April 2026 when the team joined Figma, and the product is being sunset. The repository is read-only. Any comparison article still recommending it has not been updated since then.
Can you do visual regression testing for free?
Yes. Playwright, BackstopJS, reg-suit, Loki, jest-image-snapshot and pixelmatch are all free and open source, and Argos, Chromatic, Percy and SmartUI offer free hosted tiers between 2,000 and 5,000 screenshots per month. The cost you cannot avoid is the engineering time to author coverage and keep it deterministic.
Sources
- Playwright visual comparisons documentationchecked 2026-09-13
- BackstopJS repositorychecked 2026-09-13
- reg-suit repositorychecked 2026-09-13
- Loki repositorychecked 2026-09-13
- Lost Pixel repository (archived)checked 2026-09-13
- Argos pricingchecked 2026-09-13
- jest-image-snapshot repositorychecked 2026-09-13
- pixelmatch repositorychecked 2026-09-13