Visual tests often flood QA teams with minor diffs (font antialiasing, dynamic ads, etc.). Instead of manual pixel checks, let AI-driven tools compare UI snapshots and flag only real layout bugs. For example, Tricentis Testim integrates with Applitools Eyes: it takes a baseline screenshot (first run) and then automatically compares new screenshots on each build using AI.
Step-by-step method (Testim+Applitools example):
-
Capture a Baseline: In Testim, enable the Visual Validation (Applitools) step. Run the test once to store a “golden” screenshot in Applitools Eyes.
-
Run & Compare: On each test run, Applitools Eyes grabs the current UI screenshot and uses Visual AI to compare it against the baseline. It highlights discrepancies that matter (layout shifts, missing elements, color changes) and ignores irrelevant noise.
-
Tweak Sensitivity: Configure the match level or ignore zones. For example, set Applitools’ Match Level to “Content” or “Layout” (instead of strict pixel match) so dynamic text/ads don’t break the test. You can also draw masks around known-changing areas to skip them.
-
Review AI Results: Applitools groups similar visual bugs and presents them in its dashboard. Review only the flagged differences. This AI-driven approach often finds problems (overlaps, missing images, style regressions) faster than writing manual assertions.
Tool-agnostic alternative: If you’re not using Testim, a similar trick is to write a simple script with OpenCV or scikit-image. For example, grab baseline and new screenshots and compute their Structural Similarity Index (SSIM) to generate a diff mask. Then threshold the diff and draw contours around changes (skipping tiny areas, e.g. area<100px) to focus on meaningful shifts. Finally, fail the test only if significant differences remain.
Using AI-based image comparison (whether via Testim/Applitools or a CV script) cuts through noise and lets the test point you to real visual bugs. This makes visual regression testing much faster and more reliable for QA.