Stable screenshots and accessibility details

Use this page when animation frames or a blinking caret make visual comparisons vary, or when an accessibility finding needs a closer look at the affected node. First set up a running gallery using the Musea guide, then choose capture settings and inspect the report before changing a baseline.

Musea freezes CSS animations and hides the text caret during screenshots by default. Finite animations finish and infinite animations return to their initial state. Existing baselines can change when they previously captured a live frame.

Configure capture behavior in the Musea plugin:

musea({
  vrt: {
    capture: {
      animations: "disabled", // "allow" keeps animations running
      reducedMotion: "reduce", // default: "no-preference"
      caret: "hide", // "initial" keeps the text caret
    },
  },
});

Run musea-vrt --a11y --json to save accessibility reports. Each violation retains its nodes count and adds targets containing the selector arrays, HTML snippet, failure summary, and any/all/none checks. Check data includes measured contrast colors and ratios when axe provides them.

incompleteResults carries the same details for findings needing manual review; incomplete remains a count. Expand a node in the HTML report to inspect its markup, failure reason, and measured check data.

For the first run, keep the gallery server running and use the visual regression workflow to compare the baseline, current image, and diff. Check whether a change comes from your component or the capture settings before approving a new baseline. Setting reducedMotion to "reduce" asks the page to use its reduced-motion behavior; the default remains "no-preference", independently of animation freezing.

The CLI writes accessibility reports to the output directory, .vize by default. With --json, inspect a11y-report.json; without it, inspect a11y-report.html. Review incompleteResults manually rather than treating those findings as passes. An audit error means a variant was not audited, so inspect that error before drawing a conclusion from the findings count.