STEP 01
Trained on real gaze
Trained on image-fixation pairs from real eye-tracking studies. The model learns the patterns the eye obeys — contrast, faces, text, edges, center bias.
Neural attention model · sub-second inference · live
See where
attention lands.
Trained on millions of human gaze fixations, Focal predicts where people will look on your designs — before you ship them. Built into Figma.
Interactive playground
SAMPLES
186ms
INFERENCE TIME
0.87
r vs HUMAN GAZE
millions
FIXATIONS TRAINED ON
0
PIXELS STORED
01
UNDER THE HOOD
A model that learned to look like us.
Focal is a neural attention model trained on millions of human gaze fixations — images annotated with where real people actually looked. It reproduces the same attention an eye-tracker would record, in the time it takes to switch tabs.
STEP 02
Built for design speed
A single forward pass produces a full saliency field. Fast enough to scan, adjust, and scan again — without losing your design flow.
STEP 03
Processed and forgotten
Your design is analysed in memory and discarded. Nothing is stored, logged, or used to train future models. Confidential work stays confidential.
PREDICTION vs GROUND TRUTH
r = 0.87 · same hotspots, same order
FOCAL PREDICTION
HUMAN EYE-TRACKING
02
WHAT EVERY SCAN ANSWERS
Three questions, every time.
What draws the eye
The hotspots that win the first fixations — ranked, scored, located to the pixel.
What gets missed
The cold zones nobody reaches — the CTA below the fold, the price, the disclaimer you were sure they'd read.
How hierarchy holds
Which element wins, by how much, and in what order. Test the fold and the flow at the same time.
Focal lives inside Figma.
Scan any frame without leaving the canvas. The heatmap overlays in place — fix hierarchy while you design, not after you ship.
Install the pluginScan in context
Select any frame, group or component. Iterate on hierarchy without ever leaving the file you're working in.
Test the fold
Switch between mobile, tablet, desktop and custom viewports. Scroll to test fold-by-fold attention before the developer hands it off.
Place the report
Drop the result and analysis into your Figma file as a side-by-side frame. Share evidence, not opinions, in your next review.
WEB APP COMING SOON · UPLOAD ANY SCREENSHOT
EVIDENCE, NOT OPINIONS