Visual
The lower-level screenshot-comparison surface behind visual QA workflows such as `$visual-verdict` and iterative UI fidelity checks.
Visual is OMX's screenshot-comparison and image-analysis surface. It powers workflows that compare a generated UI against a reference and turn the differences into actionable feedback.
What it is
Use this surface when a workflow needs to answer questions like:
- does the generated UI match the reference closely enough?
- where do layout, spacing, or hierarchy diverge?
- is this screenshot good enough to stop iterating?
Most users reach this capability through a higher-level workflow such as $visual-verdict, rather than invoking the capability model directly.
What it evaluates
Visual comparison in OMX typically looks at:
- overall layout structure
- spacing and alignment
- typography choices and weight
- color differences
- component sizing and visual hierarchy
- regressions between iterations
The goal is not just to produce a judgment, but to drive the next edit.
Typical workflow shape
- Capture or provide a reference image.
- Capture the current generated screenshot.
- Compare the two.
- Turn the differences into concrete edits.
- Re-run until the verdict is acceptable.
That pattern is why visual QA shows up naturally inside $ralph, $ultraqa, and UI-cloning loops.
Tool surface vs workflow skill
- Visual is the lower-level comparison capability.
$visual-verdictis the user-facing workflow that packages that capability into a repeatable QA step.
If you are navigating the architecture, this page is the capability view. If you are trying to use OMX interactively, the skill page is usually the better starting point.
Related
$visual-verdict— main user-facing screenshot comparison workflow$web-clone— uses visual comparison as an iteration gate- HUD — useful for monitoring long-running UI verification loops