Playwright vs Percy for visual testing

This is not really a like-for-like comparison, and framing it as one is how teams end up disappointed. Playwright is a test runner with built-in image comparison. Percy is a hosted review platform that you point a test runner at — including Playwright. The honest question is not which, but whether you need the hosted half at all.

At a glance

What Playwright gives you for free

More than people expect. toHaveScreenshot() retries until the rendered image is stable before comparing, which eliminates a whole class of timing flake that naive screenshot-then-diff implementations suffer from. It uses pixelmatch underneath, handles animation suppression, masking and per-pixel tolerance, and ships an HTML reporter showing expected, actual and diff side by side.

await expect(page).toHaveScreenshot("pricing.png", {
  threshold: 0.2,
  maxDiffPixelRatio: 0.01,
  animations: "disabled",
  mask: [page.getByTestId("last-updated")],
});

For a team of engineers covering thirty stable routes, that is genuinely sufficient, and paying for a platform would be waste. Setup detail in Playwright visual regression testing.

What you are actually buying with Percy

Three things, and only the first is hard to replicate.

1. A review workflow for non-engineers

This is the real product. When a designer needs to confirm that a spacing change is intended, "pull the branch, run the suite, open the HTML report" is not a workflow they will use. A hosted UI with approvals, comments and an audit trail is genuinely difficult to build and the main reason to pay.

2. Baselines outside your repository

Playwright commits PNGs. A hundred routes across two viewports and two platforms is four hundred binaries, re-committed on every redesign, permanently in your history. Git LFS mitigates it if you adopt it before the first baseline lands; afterwards the blobs stay in every clone forever.

3. Cheap responsive coverage

Percy's DOM upload means one capture can be re-rendered at several widths server-side. With Playwright, each viewport is a separate test execution. If you cover three viewports, that is three times the CI minutes — a real cost that does not appear on Playwright's $0 price tag.

The honest cost comparison

Playwright is free of licence cost, not free.

Note the last row. Percy does not reduce the authoring burden at all — you still write the test, you just add percySnapshot() to it. If your problem is that nobody has time to maintain the suite, Percy does not solve it and neither does Playwright.

You can also use both

Worth stating because the versus framing obscures it: Percy has an official Playwright integration. Teams commonly run Playwright's built-in assertions for fast engineer-facing checks in CI, and Percy for the routes where design review matters. The two are not mutually exclusive, and the decision is per-route rather than global.

When self-hosting stops paying

1. 01A second person needs to approve a change. Writing files to a directory stops being a workflow. This is the clearest threshold.
2. 02Baselines exceed a few hundred images. Repository weight becomes everyone's problem, including CI clone times.
3. 03You are comparing branches, not just commits. Branch-aware baseline selection is a real engineering problem — see screenshot comparison tools.
4. 04Your `ROUTES` list is out of date. That is a coverage problem no platform on this page solves, though a whole category of tools now generates its own coverage.

If it is the first three, a hosted platform is worth it — though Percy is not the cheapest one with published pricing; see Percy alternatives. If it is the fourth, neither tool helps.