How accuracy scoring works

The idea

When the Accuracy Check is on, every shot is graded before it reaches you. This is the plain-English version of how that grade is decided.

Vermeer compares each finished render against your real product reference photos and scores how well it holds up across three things that matter for a usable marketing shot.

What gets scored

• Product faithfulness — does the rendered product match your real one? The right label text, the right shape, the right colors. This is the heaviest factor: a wrong label sinks a shot no matter how pretty it is.
• Scene believability — does the world look real? No distorted hands, no melted faces, no impossible poses, no obvious AI tells.
• Commerce-readiness — does it read as a shot you could actually run? A clean, on-brand composition rather than something distracting or off.

Pass, retry, or hold

Each shot gets a score, and that score decides its fate:

• Above the bar — the shot passes and reaches you.
• Below the bar — Vermeer automatically re-renders it, up to a sensible number of retries, before it ever shows up in your set.

So the set you see is the set that already cleared the threshold. The retries happen quietly in the background.

Calibrated per brand

The bar isn’t one-size-fits-all. Vermeer calibrates the passing threshold to each brand, because what counts as “faithful enough” depends on the product — a text-heavy label is held to a tighter standard than a simple one. The result is a quality bar tuned to your product, not a generic average.

Honest about cost

Scoring a shot means sending your reference and the render to a trusted third-party AI to grade it, which adds a small per-image cost — which is exactly why the Accuracy Check is a deliberate per-brand opt-in rather than always-on. You decide where the extra quality is worth it.