Text that renders a hair differently
Anti-aliasingDid the title change, or only how iOS draws it?
Lumen QA reads every screen of every release the way a person would, and flags only the changes people will see.
A pixel diff is full of changes nobody would ever see. Lumen reads each one, gives it a reason, and keeps only what a person would notice.
Did the title change, or only how iOS draws it?
The carousel shuffled again. Is anything broken?
Only the clock moved on this screen, right?
Is the hero photo softer, or is it just me?
Can every shopper still read the price?
Can people on a 390 pt phone still reach Pay?
Lumen captures every screen after your end-to-end tests, reads every change, and hands you the few a person would notice. You decide what becomes the new baseline.
01642 shots of 214 screens on three devices, taken right after your end-to-end tests.
02The vision model reads all 1,204 changes and sets aside 1,202, each with a reason you can open.
03Two tickets are filed with both screenshots. Dana approves 4.12 as the new baseline, because the model never does.
A pixel diff found 1,204 changes in release 4.12. Lumen set aside 1,071 anti-aliasing changes, 88 timestamps, 34 carousel changes and 9 resampling changes, each with a reason. 2 changes remain for a person, and Dana approves the new baseline.
Lumen does the reading. Someone on your team decides what becomes the baseline, and every decision is kept.
Approvals
Lumen never accepts a baseline on its own. A named reviewer approves each one, and the audit log keeps who and when.
Tickets
Each flag becomes a ticket with the before and the after, so the fix starts from the evidence.
The honest limit
Lumen sees that a hero photo moved 12 pt or lost contrast, not that it feels off brand. Live prices need a mask.
“It caught a pay button pushed under the fold on small phones, the kind of thing no unit test would ever have seen.”
14 days free, 5,000 screenshots, no card needed