How it works
Video models respond to the same language cinematographers use. Instead of one long ramble, a good prompt answers six questions in order: how big is the shot, who is in it, what are they doing, where are they, how does the camera move, and what do the light and style feel like. This generator asks those questions one at a time and stitches the answers into a single clean paragraph, adding audio direction and technical settings at the end.
A prompt structure that works
| Part | What it controls | Example |
|---|---|---|
| Shot size | How much of the scene the viewer sees | close-up, wide shot, aerial shot |
| Subject + action | The story of the clip in one breath | a diver holds still as a whale glides past |
| Setting | Place, time and atmosphere anchors | deep blue open ocean, dawn, night market |
| Camera movement | Energy and point of view | slow push-in feels intimate; a drone flyover feels epic |
| Lighting + style | The look — realism, anime, commercial gloss | golden hour light, hand-drawn anime look |
| Audio + negatives | Sound direction, and what to keep out | ambient sound; avoid morphing and watermarks |
Tips from real generations
- One clear action beats three competing ones. If the subject does everything, the model does nothing well.
- Name the light. “Golden hour” or “neon night” changes the grade more than any style adjective.
- Keep dialogue short — a single short line per clip is what current models handle reliably.
- Iterate one variable at a time: same prompt, new camera move. That is how you learn what each field does.