The Real-Camera Prompt Cookbook (Part 14): Shot Recipes That Don’t Look Like AI

The Physics of Better AI Images—now in “recipe” form

If you’ve already internalised the failure modes and the realism layers, the next bottleneck isn’t theory.

It’s execution.

This dispatch is a practical cookbook: camera-like prompt templates, shot-by-shot. You’ll be able to take a scene idea and turn it into an image that behaves like a real camera—because the prompt encodes the same constraints your lens and sensor would.

---

The cookbook rule: prompts should “declare the shot”

A strong real-camera prompt usually contains four declarations:

1. Camera position & framing (where the lens sits, what’s in front of it)
2. Optics (focal length / field of view cues)
3. Exposure & sensor character (what part of the dynamic range the image sits in)
4. Optical imperfections & micro-failure (the tiny things that stop it from looking rendered)

---

A reusable master scaffold (copy/paste)

Use this as the backbone, then swap the variables. The point isn’t to be verbose; it’s to be specific in the physics that matter.