Lighting & shadow check

When the light comes from two suns, Chekr notices

Lighting is where composited and generated images quietly fall apart: a subject lit from the left in a scene lit from the right, a product with no contact shadow, a highlight with no source. Viewers rarely name the problem — the image just feels off. Chekr names it.

What lighting errors look like

  • Subjects and backgrounds lit from different directions
  • Missing or detached contact shadows under objects and people
  • Highlights with no plausible light source
  • Shadow softness inconsistent with the scene’s light
  • Color temperature that changes between regions without reason

How Chekr checks it

The lighting pass estimates the scene’s light — direction, softness, temperature — and then audits every subject and object against it. Regions that disagree get pinned with an explanation of what’s inconsistent.

Localized problems (a missing contact shadow) can be fixed with a region regenerate or guided edit; scene-wide inconsistencies are flagged as regenerate-at-source, so you don’t waste edits on an image that needs a new prompt.

Frequently asked

Why is lighting so often wrong in AI images?

Models assemble scenes from parts learned under different lighting. Each part looks right alone; together they disagree about where the light is. Human vision is extremely sensitive to that disagreement, even preconsciously.

Does this matter for flat/graphic creatives?

Less — the check weights findings by how photographic the creative is, and your rule set decides whether lighting findings warn or block.