AI Video Prompts: How to Write Prompts for Sora, Veo and Kling

Text-to-video models such as Sora, Google Veo, Kling, Runway, Hailuo and Wan can produce convincing clips from a paragraph of text — but only if the paragraph reads like a shot description rather than a story. This guide explains how to describe camera, motion, style and sound so the model generates the clip you pictured.

Think in shots, not stories

Most models generate one continuous clip of roughly 5–10 seconds. A prompt that describes a whole plot ("she wakes up, drives to work, then gets promoted") forces the model to cram or skip events. Describe one shot: one camera setup, one main action.

If you need a sequence, generate each shot separately with the same character and style description, then edit them together.

The six building blocks

1. Shot type and framing

extreme wide shot, wide shot, medium shot, close-up, extreme close-up, over-the-shoulder, bird's-eye view, low angle

2. Camera movement

Pick one clear movement:

3. Subject and action

Describe the subject precisely and the action as continuous motion: "a golden retriever runs through shallow waves, spraying water with each stride" is better than "a dog at the beach".

4. Setting

Location, time of day, weather and a few specific background details.

5. Lighting and color

golden hour backlight, overcast soft light, neon reflections on wet pavement, candlelight, teal and orange grade.

6. Style

photorealistic, cinematic 35mm film, documentary, anime, claymation, 3D Pixar-style animation, watercolor animation.

A template

[Shot type], [camera movement]. [Subject] [continuous action] in
[setting] at [time of day]. [Lighting and color]. [Style]. [Audio, if supported].

Example:

Medium tracking shot, camera moving alongside. A cyclist in a yellow
raincoat rides through a narrow cobblestone street in Lisbon at dusk,
tires splashing through puddles. Warm street lamps reflect on the wet
stones under a deep blue sky. Cinematic 35mm film look, shallow depth
of field. Ambient sound of rain and a distant tram bell.

Model-specific notes

• Sora handles long, detailed descriptions and physical interactions well; be explicit about how objects move and collide.
• Veo can generate synchronized audio. Describe sound effects and put short dialogue in quotes: the chef says, "Order up!"
• Kling responds well to clear camera directions and supports image-to-video — describe only the motion when animating a still.
• Runway, Hailuo and Wan favour concise prompts; lead with the action and camera move.

For image-to-video, do not re-describe everything in the picture. Describe what should move and how the camera behaves: "the woman turns her head toward the camera and smiles, hair moving in the wind, slow push-in".

Common mistakes

1. Several actions in one clip — the model blends them into something strange. One action per shot.
2. Conflicting camera moves — "zoom in while orbiting and panning" usually produces jitter.
3. Vague motion words — "moving", "doing something". Use specific verbs: sprints, pivots, pours, unfolds.
4. Too many characters — each extra person increases the chance of warped faces and hands.
5. Asking for readable text or logos — most video models still render text poorly.
6. Real people or copyrighted characters — most platforms block them; describe an original character instead.

Examples to adapt

Nature

Extreme wide shot, slow crane up. Morning fog rolls over a pine forest
valley as the first sunlight breaks over the ridge. Soft golden light,
cool blue shadows. Photorealistic nature documentary style.

Product

Close-up orbit shot. A pair of white running shoes rotates slowly on a
floating concrete block while water droplets hang frozen in the air
around it. Clean studio lighting, dark gradient background, high-end
commercial style.

Animation

Wide static shot. A small robot waters a single flower on a rooftop
garden at sunset, then gently pats the pot. City skyline in the
background, warm orange light. Soft 3D animated film style.

Generate a video prompt from an idea

The AI Video Prompt Generator turns a one-line idea into a single, production-ready shot description with camera, action, setting, lighting, style and optional audio — ready to paste into Sora, Veo, Kling or any other video model.