Render scoring is grading a generated visual against the brand's record instead of eyeballing it. A finished ad is checked against the real product and the brand's rules, so a wrong product, a distorted label, or copy sitting over a face gets caught before anyone ships it.
Why it matters
Generated images fail in specific ways. The product comes back subtly wrong, text lands on a face, an extra object floats in the scene. A quick human glance misses these under volume.
Eyeballing also does not stay consistent. Two reviewers disagree, and the same image can pass on Monday and fail on Friday.
Scoring makes the check repeatable. The same image is measured the same way each time, against the brand's own reference rather than a general sense of good.
How Jinn treats it
Vermeer runs two layers. First a deterministic pixel check reads legibility from the image itself: it measures the contrast of the copy band by the WCAG formula, so faint copy on a busy background is caught by math, not opinion. That layer is authoritative, and a band that fails it can never quietly pass.
Then a vision pass scores four independent axes: whether it is the same product as the reference, whether the product is rendered faithfully, whether copy sits in negative space or over a face, and whether the scene is free of artifacts. Findings surface for review rather than blocking.
Every Vermeer composite gets a pixel-contrast legibility check plus a vision judge, with findings surfaced for review. See it
Related terms
Terms without links are on the roster and get their pages as each definition ships.
Grade the render, do not just glance at it.
A deterministic pixel check plus a vision judge, measured against the brand's real product. See how Vermeer scores every composite.