See every pixel that changed

One screen captured once — one page, in one browser, at one size. A run over ten pages is ten snapshots. Free gives you 1,000 as a lifetime allowance rather than a trial that expires, and the paid plan is a flat monthly price for the whole organisation with no per-user charge.

The free plan gives you 1,000 snapshots as a lifetime allowance with no card, and Premium is one flat monthly price for 5,000 snapshots every month across the whole organisation — see the pricing page for the current price. Neither plan is charged per user or per parallel session.

Desktop browser sessions in Live — Chrome, Edge, Firefox and Safari on real Windows and macOS. Capture from real phones works from a Live session too; scheduled runs are desktop browsers only for now, and capture from automated Selenium and Playwright runs is not part of this release. We would rather say that here than let you find out after you have bought it.

Yes, but not with a scheduled run. Runs do not sign in — they open a public address only, so they cover pages a visitor can reach without a password. For anything behind a login you have two ways past it: open a Live session and sign in yourself, then capture from there, or let an Automate test do it as part of its run.

Two controls, for two kinds of noise. A per-project threshold handles the diffuse rendering variation that appears everywhere — anti-aliasing, sub-pixel shifts. Ignore regions mask a specific rectangle that legitimately changes, like a clock or a carousel, in both images before they are compared and from the next capture onwards. Using the wrong one for the wrong problem is what makes visual tools get switched off.

Yes, every time, until you approve them. That is the point of a baseline: it does not know which changes were deliberate, so a redesign produces a queue of diffs on the day you ship it. Approve them in one pass and the queue goes quiet again — the noise is front-loaded onto the day you changed something, which is the day you can explain it.

Capture does: you can take a visual snapshot from a Live session on a real phone or tablet. Scheduled runs are desktop browsers only today — the mobile leg of scheduling is a known gap rather than a setting you are missing.

Whoever can tell whether a change was intended — usually a designer or product owner rather than whoever ran the tests. That is a cost question everywhere else in this category, because reviewing means holding a paid seat. Here user accounts are not the billing unit, so the person best placed to judge a diff can open it without anyone buying them a licence first.

Capture stops until the allowance resets or you upgrade; nothing is deleted and no baseline is lost. The free allowance is 1,000 snapshots as a lifetime total rather than a monthly one, which is enough to run a real trial on a real set of screens rather than a demo.

Those are evidence of what happened; nobody compares them to anything. Visual testing adds the missing half — an approved version of the screen, a comparison against it, and a person who decides whether the difference is intended. The CLI turns the folder you already produce into exactly that.