JM Velez

Product & UX · Northeastern University, 2026

Adaptive Liquid Glass

Role
Research, concept, prototype
Team
J. M. Velez, Harsh Kumar and Sujan Pokhrel, as “Team Apple”
Tools
Figma, WCAG contrast checkers
Output
26-screen interactive prototype

The problem

UNDER CONSTRUCTION

"Team Apple" approached the project from three directions: public and reviewer reaction to the release; the revisions Apple itself shipped in subsequent point updates; and our own measurements of the UI shipped against WCAG 2.2. The complaints clustered, and they were not about taste. Text over bright or busy backgrounds fell below the 4.5:1 contrast ratio that WCAG 2.2 AA requires for body text, and the failure was worst exactly where phones get used most — outdoors.

Apple's own hotfixes were instructive. They increased opacity in specific components, which read as an admission that the problem is real, and as a concession that the fix so far has been piecemeal rather than systemic.

Who it fails

Low contrast is an inconvenience for most people and a barrier for some. We scoped to four groups: users with low vision, users in high-glare environments, older users, and users who simply wanted more control over how their device looks.

The decision that shaped everything

The obvious fix is to turn the translucency down. Make the panels opaque, contrast goes up, problem solved.

We rejected it. A uniform fix throws away the aesthetic entirely, which defeats the purpose of a refresh, and it fails the same way the original does — by assuming one setting works everywhere. Liquid Glass is not broken in a dim room over a plain wallpaper. It is broken in direct sunlight over a photograph.

So the system should not have one appearance. It should have a contrast target, and adjust its appearance to hit it.

How it works

Two prototype screens: one sampling wallpaper brightness, contrast, colour complexity and luminance variance; one classifying ambient light into indoor, outdoor, direct sunlight and low light
Detection: The system samples wallpaper pixels beneath each UI element and reads the ambient light sensor, classifying the surroundings into four modes. In this state it has found an 80,000-lux reading and a psychedelic gradient wallpaper, close to a worst case.
Prototype screen showing a decision engine taking brightness, complexity and environment as inputs and outputting blur, opacity, saturation and tint values
Decision: Brightness, colour complexity and environment were the three inputs that moved readability most in testing, so those are what the engine weighs. It outputs four glass parameters, on-device, in a single frame at 60fps.

Tuning four parameters together rather than one is what preserves the look. Raising opacity alone produces flat grey panels. Raising opacity while lifting saturation and shifting tint keeps the glass reading as glass.

Lock screen with Adaptive Liquid Glass active over a colourful wallpaper, showing a 4.5:1 WCAG AA pass indicator
Result over a wallpaper that defeats the shipping UI: legible notifications, diffusion intact, 4.5:1 met.
Accessibility settings screen showing larger text, bold text, increased contrast and reduce motion, with glass contrast raised to 5.2:1
Accessibility preferences can be used to not just restyle the UI, but move the engine's target to accommodate different users' needs. With low-vision settings on, contrast is held at 5.2:1, well past AAA.

What we kept

Fonts and icons are untouched from Liquid Glass 1.0. The diffusion effect that defines the material stays. The engine darkens backgrounds and cuts reflectiveness only as far as compliance requires, and no further. A user who never opens Settings gets an interface that looks like the one Apple shipped and reads in daylight; a user who wants to push contrast to 7:1 can.

Prototype overview screen summarising the four components of Adaptive Legibility
The feature as a user would first meet it, in the onboarding flow.

Reflection

UNDER CONSTRUCTION

Next: Don't Forget to Smile!