Comparison
Coverli and Chromatic, side by side.
Chromatic defined visual testing. We built it for the states that matter to accessibility. Here is what each does, stated plainly enough that a Chromatic engineer could read it without wincing.
| Capability | Coverli | Chromatic |
|---|---|---|
| Resting-state screenshots per commit | Yes | Yes |
| Skip unchanged stories | Yes — build-graph change detection | Yes — TurboSnap |
| Interaction states (open selects, dialogs, menus, checked controls) | Automatic — no test code | Requires writing play functions per component |
| Focus-indicator screenshots (WCAG 2.4.7 evidence) | Yes — every interactive element, opt-in per project | No |
| Flake handling | Recapture-and-verify guard; unstable stories flagged as unstable, never as changes | Sensitivity thresholds and per-story tuning |
| False-positive posture | Deterministic capture: unchanged code produces byte-identical images | Anti-aliasing settings, diff thresholds |
| Region comments on changes | Yes — click-to-pin markers or drawn regions; comment, or comment-and-reject | Yes |
| Rejection feedback in the pull request | Annotated screenshot embedded in the comment — flagged regions drawn and numbered, no login needed | Link out to their app |
| Screen reader output testing in the same run | Yes — it is the core of the platform | No |
| Accessibility audit with cited WCAG evidence | Yes | No |
| Immutable approval trail (who approved what, when) | Yes — governance-grade | Partial |
| Browser matrix (multiple engines) | One controlled engine today; regressions live in your code, not in browser weather | Multiple browsers |
| Themes and viewport modes | roadmapBaseline model is built for it | Yes |
| Per-branch baselines with auto-accept on merge | One approved baseline lineage — simpler, audit-friendly | Yes |
| Pricing model | Included with accessibility testing | Separate per-snapshot product |
Rows marked roadmap are not shipped. We would rather tell you that than let you find out during a pilot.
“Chromatic photographs your components at rest in five browsers. We photograph them open, checked, and focused — in the same run that tests what they announce to a screen reader.”