Glance

Best free attention heatmap tools for your next review

For a free attention heatmap, start with Glance if you want a local design workflow. UNISAL and DeepGaze IIE provide research code; pysaliency helps evaluate predictions; the manual squint test checks composition without software. These options solve different jobs. Paid platforms add hosted workflows and support.

Try Glance in browserDownload

Choose by the job you need done

“Best” depends on whether you need a design review, research implementation, metric evaluation or team service. This is not a measured performance ranking. Free code does not automatically grant rights to model weights or datasets.

Free and research options
OptionUseful forLimit to check
GlanceFree, open-source local attention maps and design comparisons; no Glance account.In development. MIT app code; model assets separately licensed. Verify downloads and supported models.
DeepGaze IIE codeResearch implementation of learned saliency prediction.Research only here: no reuse license established for this recommendation. Public code is not permission for commercial reuse or redistribution; request terms.
UNISALResearchers needing image and video saliency code.Apache-2.0 code; archived repository. Check weights, dependencies and dataset terms. Requires a research/developer setup.
pysaliencyDevelopers evaluating models and saliency metrics.A Python evaluation framework, not a polished screenshot-to-heatmap product. Check license and dataset terms.
Squint test by handAnyone checking large light/dark groups immediately.No learned gaze model, measured gaze or accuracy score.

What each free route gives you

Glance is the design-workflow route: keep files local, inspect a prediction and follow the model citation. A browser check makes the first image easy to try; desktop inference can run offline after setup. Check the download page before relying on a particular release. Browser runtime and model caching have separate availability constraints.

DeepGaze IIE and UNISAL are research routes. Expect to prepare inputs, manage dependencies and interpret model outputs. The DeepGaze IIE paper and UNISAL paper explain their methods. They do not establish that every conversion or export reproduces the published benchmark.

Use pysaliency when your question is how to evaluate a model. Use squinting when your question is whether the composition reads as a few clear shapes. Neither belongs in a feature checklist as though it were a hosted design review suite.

Paid options for team workflows

Choose a paid service when its collaboration, reporting or support removes work your team would otherwise do manually. Bring the same image and a clear reporting requirement to each trial or demo.

12 Nov 2026 recheck: Feng-GUI has a public pricing page at the linked URL. The earlier “not found” observation is historical. These are dated price observations, not current quotes. Live plan names and monthly/annual billing can differ. No shared test establishes which product predicts your audience better.

Read the map before acting

Predicted attention is not observed gaze, comprehension or conversion. Keep model, crop and region boundaries fixed when comparing variants. Read the accuracy guide, then follow the landing-page workflow to choose one question for a real user test.

Choose your next check

Try in browserDownload