Visual Prompting Protocol

# Visual Prompting Protocol

> **Objective**: Define the physics and parameters of an image before you generate, in terms that hold across any image model.

This protocol is model-neutral on purpose. It talks about aspect ratio, stylization strength, and exclusions as ideas, not as one product's flags. Whatever tool you use, map these concepts to its controls.

## 1. The universal formula

Every prompt answers four questions:

1.  **Subject**: who or what is it? (Adjectives + noun, doing something specific.)
2.  **Environment**: where is it? (Background, context, time of day.)
3.  **Light**: how is it lit? (Direction, quality, and source: soft key, hard rim, golden hour, single bare bulb.)
4.  **Camera**: how is it shot? (Focal length, aperture, medium or film stock, angle.)

Answer all four and the image has a spine. Skip one and the model fills it with its own average, which is where the generic look comes from.

## 2. Structure the prompt (fields, not prose)

Keep the prompt as separate fields so each can be edited without disturbing the rest. The point is control and reproducibility, not any particular file format. A plain structure like this travels between tools: