Prompting for video · Lesson 1 of 4
The anatomy of a shot
Subject, setting, light, action — in that order. · 4 min read
Video models were trained on footage described the way a camera sees it. The prompts that work best read like a description of one image in motion, not like a story.
Cover four things, roughly in this order:
- Subject: who or what the shot is about, described physically.
- Setting: where it is, and when.
- Light: the time of day, the source, the mood.
- Action: the one thing that happens during the clip.
Put the subject first. Models weight the start of a prompt most, and a prompt that opens with camera movement tends to produce a well-executed move around the wrong thing.
Weak: "A cool video of a woman in a city at night, cinematic."
Strong: "A woman in a yellow raincoat waits at a tram stop, rain-slicked street, neon signs reflecting in puddles, she lowers her umbrella and looks up."