Skip to content
heatpoints
FRLog inCreate an accountAdd to Chrome

Compare image variants with AI attention heatmaps

Bring your alternatives into one review. Championship calculates a heatmap for each image, compares the same metrics and helps you decide what to test next.

Drop two or more variants

PNG, JPG or WebP, up to 10 MB each. Limited free access, at most one credit per variant.

Make the comparison meaningful

Use variants with the same purpose, dimensions and surrounding context. A square ad and a desktop landing page are different tasks; their score difference is not a useful creative verdict.

  1. Choose your images and check the batch allowance shown by the tool.
  2. Keep the offer and format consistent. Name files after the variable you changed.
  3. Run the comparison and inspect each heatmap alongside its original.
  4. Review the metric differences and export the results as PNG or CSV.

The ranking is a model summary

Championship combines normalised attention intensity, grid hierarchy and intensity variation into a score. It also shows the underlying metrics, including coverage. The top-ranked image is the highest-scoring image under that formula, not a proven advertising winner.

If two variants are close, inspect where the emphasis differs. A change that highlights decoration instead of the offer can increase a metric without helping the design. Do not interpret a score gap as an expected change in click-through rate or sales.

References: Exact metric definitions

A practical comparison brief

For an ad, keep the product and copy constant and compare two crops. For a thumbnail, compare the amount of surrounding detail. For a landing page, adjust the relationship between the hero illustration and the primary action.

Write down the intended effect before running the images. Then inspect that region in the overlays. This produces a specific hypothesis to test, rather than a post-hoc explanation for whichever image ranks first.

References: Review visual hierarchy

Access, credits and saved work

The interface reads the current batch size and credits from your allowance. Initial free access is limited; larger batches or further runs can require an account. A distinct, uncached image can consume a credit. The tool reports the run's usage.

Comparison results can be exported as PNG or CSV. They are not automatically saved as a comparison history in the media library. Keep your export before leaving if you need a lasting record.

References: Current access conditions

Prediction is not an A/B experiment

A live A/B experiment exposes real audiences to variants and measures a defined outcome. Championship analyses pixels without audience exposure. Use it to narrow the variants you want to test, then measure the outcome that matters to you.