Focal AI

About this plugin

Designers guess where attention goes. Focal predicts it.Focal uses AI to predict where people will look at your design — generating an attention heatmap in about a second, so you can fix your visual hierarchy before it ships.Right now, you find out whether users actually saw your CTA after launch — in analytics, or never. Real eye-tracking studies answer the question, but they cost thousands and take weeks. So most layouts ship having never been tested for visual hierarchy at all.Focal closes that gap. Select any frame, run a scan, and an AI model predicts where a viewer's eyes are most likely to land. You get an attention heatmap in about a second — adjust your layout, re-scan, and watch the prediction change. It's a feedback loop fast enough to use while you're still designing, not after.Evidence, not opinions. Instead of debating hierarchy in design reviews, point to a number. Focal turns "I think the headline competes with the image" into a measurable attention split you can show your team.What you get from every scanAttention heatmap — a color overlay showing predicted high- and low-attention zones, with adjustable opacity. View it four ways: Heat (standard), Path (predicted look order), Fog (hides everything users would likely miss), and AOI (per-element breakdown).Clarity Score (0–100) — a single read on how clear your hierarchy is, from Cluttered to Strong.Attention metrics — Concentration, Dead Space, Focal Point count, and a top/middle/bottom distribution so you can see if your fold is pulling its weight.Areas of Interest — pick any layer (or let Focal auto-detect buttons, images, and text) and get the % of predicted attention it captures.Actionable suggestions — automatic flags for things like a buried CTA, competing focal points, dead space, or content stranded at low-attention edges.Shareable output — export a heatmap PNG or a full PDF report, or drop the heatmap and analysis back into your file as frames.How it works Focal is powered by an AI model trained on eye-tracking and visual-importance data.Your designs stay private Frames are sent for AI inference, processed in memory, and discarded immediately. Nothing is stored, and nothing is used to train the model.