# Concept note — A Camera Cannot See the Oil

## Creative approach

The still turns an invisible estimation error into a physical object. A fried meal sits on a worn diner table under ordinary hard overhead light, but the plate has been cut away like a museum specimen. Beneath the food, an amber reservoir reveals the absorbed and pooled fat that a top-down photograph cannot register. A real measuring vessel beside the plate gives that hidden amount a second, immediately readable volume.

The single figure, **≈30 g**, is stencilled into the battered tabletop and dominates the lower third. The only small label points directly to the exposed reservoir. This keeps the argument legible at phone size: the camera sees the meal's shape and colour, but not what cooking has put inside or underneath it.

## Visual system

- Near-9:16 portrait at 941 × 1672 pixels.
- Photographic chipped diner formica, real scratches, hard light, and close food texture.
- Restrained turquoise, oxide red, amber, cream, and black material palette.
- One headline, one locked number, and one functional label.
- No phone interface, scan overlay, calorie-counter graphic, brand, person, body, scale, dieting message, or invented sample size.

## Generation disclosure

The hero image was created with **OpenAI GPT Image (built-in image generation tool)** from an original prompt written for this bounty. The final output was reviewed at full resolution for text accuracy, physical plausibility, evidence restraint, and prohibited imagery.