Visual Saliency Engine · v2.0

WHERE DOES THE EYE LAND FIRST?

AI-estimated saliency for graphic design. Upload any layout and get a heatmap, scan path, an 8-metric score, and a written report in about a minute.

Drop a design to analyze
Mode: Heat_Map_04
Capabilities
01/ Attention Heatmap

See where the eye lands

A thermal overlay highlights the regions most likely to pull focus in the first few seconds of viewing.

02/ Scan Path

Follow the viewing order

Numbered fixation points trace the sequence a viewer is likely to move through — from first glance to last.

03/ Design Score

Eight metrics, one report

Hierarchy, contrast, typography, readability, balance, spacing, CTA clarity, and accessibility — each scored out of 100.

04/ AI Suggestions

Know what to change

A short written report explains why attention lands where it does and gives concrete edits to try next.

Protocol

Upload Asset

Drop in any static design — JPG, PNG, or WebP up to 20MB.

Vision Model

A vision-language model reads the layout and estimates likely focus regions.

Readable Report

Get a heatmap, scan path, an 8-metric score, and written suggestions.

Use Cases

[UI/UX]
Friction Audits
[AD]
Ad Recall
[BRAND]
Poster Studies
[AGENCY]
Client Reports
[EDU]
Design Critique
[SOCIAL]
Feed Testing
A note on accuracy

HeatSight predicts attention using a vision-language model — it isn’t hardware eye-tracking and it won’t replace real user testing. Use it to pressure-test a layout, sanity-check hierarchy, and move faster between iterations.

FAQ

How accurate is the prediction?+

These are AI estimates, not lab eye-tracking. The model reasons about hierarchy, contrast, and composition to predict where attention is likely to land — useful for iteration, not a substitute for real user testing.

What files can I upload?+

JPG, PNG, and WebP up to 20MB. Best results come from full-resolution exports of the actual design rather than screenshots of screenshots.

Do I need an account to try it?+

Yes — a free account is required to run an analysis. It takes a few seconds and everything is free right now.

Is my upload private?+

Your image is sent to the vision model only to generate the analysis. Results are kept in your browser session and aren't shared.

System ready

Audit your design.

Start analysis