Video tutorial · 1:18

The AI video prompt formula: subject, motion, camera

The AI video prompt formula is simple: describe a shot, not a wish. A video model fills every detail you leave out with a guess, so name three things, a subject you can picture, a motion so something happens, and a camera so the model knows where you stand.

Try text to video →Read the transcript

Key takeaways

Key takeaways from the AI video prompt formula

Each part of the formula removes one guess the model would otherwise make. Fewer guesses means fewer retries, and the clip you get stays closer to the one you pictured. Use the four points below as a checklist before you hit generate.

01

Subject

Who or what is in the shot, described so you could picture it.

02

Motion

What moves, and which way.

03

Camera

Where the viewer stands: low tracking shot, slow push-in, close-up.

04

Photos too

For image to video, say what moves and where the camera goes.

Example

AI video prompt formula: before and after

Both prompts describe the same idea. The first one is a wish, so the model has to guess who moves, which way and where the camera is. The second one names the subject, the motion and the camera, which leaves almost nothing to chance.

Before · a wish

a dog on a beach at sunset

After · a shot

A golden retriever sprints along wet sand as the waves break behind it. Low tracking shot, warm sunset light.

Transcript

Video transcript

The full narration, split into the seven chapters of the video. Select a timestamp to play the video from that moment, or read straight through if you prefer text.

Why AI videos look weird

You typed a great idea into an AI video tool. So why did the video come out looking so weird?

Writing a wish

Most of us write a wish. A dog on a beach at sunset. Then we hit generate, and we hope.

Why the model guesses

But the model has to guess everything you left out. Who moves? Which way? Where is the camera? Every guess is a coin flip, and the flips add up fast.

The 3 parts of a shot

A video model thinks in shots, not wishes. A shot needs three things. A subject you can picture clearly. A motion, so something actually happens. And a camera, so the model knows where you stand.

Same idea, done right

So try the same idea again. A golden retriever sprints along wet sand as the waves break behind it. Low tracking shot, warm sunset light. Now there is almost nothing left for the model to guess.

Photos work too

Fewer guesses mean fewer retries. You get closer to the clip you pictured, faster. It works for photos too. Tell a still image what moves and where the camera goes.

Recap

Next time, before you hit generate, check three things. Subject. Motion. Camera. Describe the shot, not the wish. Make your first video free at onvid.ai.

FAQ

AI video prompt formula FAQ

The AI video prompt formula is simple: describe a shot, not a wish. A video model fills every detail you leave out with a guess, so name three things, a subject you can picture, a motion so something happens, and a camera so the model knows where you stand.

What is the AI video prompt formula?

The AI video prompt formula describes a shot in three parts: the subject, the motion and the camera. The subject says who or what is on screen, the motion says what moves and which way, and the camera says where the viewer stands, like a low tracking shot or a slow push-in.

Why does a short prompt give weird AI videos?

A short prompt leaves gaps, and the model fills each gap with a guess about who moves, which way and where the camera is. Every guess can go wrong, so a vague wish often produces odd motion or framing. Naming the subject, motion and camera removes most of the guessing.

How long should a prompt be with the AI video prompt formula?

With the AI video prompt formula, one or two sentences is enough for most clips. The example on this page is 20 words: a golden retriever sprints along wet sand as the waves break behind it, low tracking shot, warm sunset light.

Does the AI video prompt formula work for image to video?

Image to video needs one change. The photo already shows the subject, so the prompt only has to describe the motion and the camera. Tell the still image what moves, like clouds drifting, and where the camera goes, like a slow push-in.

What camera terms work best in AI video prompts?

Simple film terms work best: close-up, wide shot, low angle, tracking shot, slow push-in and handheld. Pick one camera move per shot. Stacking several moves in one prompt gives the model conflicting directions, and the result drifts or shakes.

Should AI video prompts include lighting and style?

Yes, after the three core parts. Lighting such as warm sunset light or soft window light, and a style such as cinematic or anime, refine the look. Add them as a short tail on the prompt, after the subject, the motion and the camera.

Why does the AI video prompt formula reduce retries?

Every missing detail is a separate guess, and the more guesses a prompt leaves, the less likely the clip matches what you pictured. Naming the subject, motion and camera removes the biggest guesses, so more first attempts come out usable.

Can one prompt describe several shots?

Keep one shot per prompt. A clip is one continuous shot, so asking for a cut, a second location or a scene change in the same prompt confuses the model. For a sequence, write one prompt per shot and join the clips afterwards.

What is the most common mistake in AI video prompts?

Describing a feeling instead of something visible. Words like amazing or epic give the model nothing to draw. Replace them with things a camera can see: a golden retriever, wet sand, waves breaking and a low tracking shot.

Does the AI video prompt formula work with every AI video model?

The formula works with any text to video model, because every model has to decide the subject, the motion and the camera. Models differ in how literally they follow camera terms, so rephrase the camera or switch models when one ignores it.

Explore more

How to Write AI Video Prompts That Actually Work

Text to Video AI Generator

Image to Video AI Generator

10 Best AI Video Generators for YouTube in 2026

InVideo vs OnVid: Which Is Better for AI Video?

7 Best InVideo Alternatives in 2026

Write your next shot and see the difference.

A 78-second animated tutorial on the AI video prompt formula: describe a shot with a subject, a motion and a camera instead of a wish, with a before-and-after.

Free to start after you sign up