Only about 2° of your visual field is sharp. The rest is peripheral — blurred, desaturated, crowded. Scrutinizer renders that reality so you design for the glance, not the screenshot.
Designing in Figma? Try the plugin instead →
Scrutinizer models how spatial resolution, colour sensitivity, and crowding change with distance from where you're looking — the same falloff your retina and cortex impose.
Resolution drops with eccentricity along a log-polar mapping. Scrutinizer renders this falloff with hardware MIP-chain pooling — each level halves spatial resolution, approximating how receptive fields grow with distance from fixation.
Colour perception narrows in the periphery — S-cones are sparse outside the fovea and the parvocellular pathway loses resolution with eccentricity. Scrutinizer models this as selective chrominance pooling in Oklab, preserving luminance while colour bleeds and shifts.
In the periphery, objects aren't invisible — they're unidentifiable. Flankers merge with targets. Scrutinizer renders this as position jitter scaled to integration-field size, following the Texture Tiling Model .
Check whether your primary CTA and key content hold up under peripheral degradation — before you ship. If people have to foveate every element to find what they need, the layout isn't working, and Scrutinizer shows you before they tell you.
The Restricted Focus Viewer paradigm, modernised: see where attention lands and where peripheral clutter competes with the task — with a window that's biologically accurate, not a plain Gaussian blur.
Point the open-source browser at your real app and watch it through human perception — live congestion scores, saliency, and peripheral falloff on the page as it ships. Scrutinizer is open interaction infrastructure: a rendering pipeline you can run on any site or build on, not a sealed app.
Eccentricity-dependent pooling, Oklab chromatic degradation, and Texture-Tiling crowding — all parameterised. Screenshot exports embed the full pipeline config so a result is reproducible, not just a pretty picture.
Every rendering choice is argued from the literature and checked against psychophysics — in long form, in the open. A few entry points to the science:
Red–green opponency collapses in the periphery; blue–yellow holds on. It's been settled vision science for 25 years — and still hasn't reached the design curriculum. Road signs already figured it out.
Read the brief →A 12-band decomposition and a biological desaturation ramp, validated against Bowers 2025, Hansen 2009, and Mullen & Kingdom 2002.
Running Rosenholtz's Feature Congestion on live pages — from the app, the command line, or via MCP for AI-assisted review. Validated at ρ = 0.93.
DoG band decomposition and eccentricity-driven level selection — the trick that turns a graphics shortcut into a model of the retina.
A standalone browser with the full foveated rendering pipeline. Browse any site with real-time peripheral simulation, saliency maps, congestion analysis, and mobile device emulation.
Apply foveated rendering to your Figma frames without leaving the editor. Copy static renders to canvas, or browse with live simulation.
A 30-second calibration that maps your fovea in pixels at your current viewing distance — the same number the demo above estimates.