# How to Write AI Video Prompts That Actually Work (50+ Examples)

> Learn the exact structure for writing effective AI video prompts. Includes 50+ tested examples across horror, children's, history, real estate, and more.

By Benet, founder of FrameSurfer. Published 2026-02-07, updated 2026-10-06.
Canonical: https://framesurfer.com/blogs/ai-video-prompts-guide

Most AI video prompts fail for one of two reasons. They are too vague, so the model fills the gaps with generic defaults. Or they are a pile of instructions that contradict each other. This guide gives you the shape of a prompt that works on every major tool, the five mistakes to stop making, and 51 prompts sorted by genre that you can paste in and adapt.

## The shape of a prompt that works

![The anatomy of a great AI video prompt](/blog/generated/ai-video-prompts-guide-anatomy.webp)

Write the prompt as a shot description, in this order:

**Subject → Action → Environment → Camera → Lighting → Style**

1. **Subject.** Open with the main thing on screen. "A lone astronaut", not "A beautiful cinematic scene of".
2. **Action.** The subject is always doing something continuous. "Walking slowly toward", not "standing".
3. **Environment.** What the camera can see, with something moving in the background. "A barren red plain, dust drifting across the ground."
4. **Camera.** Simple words. "Camera slowly pushes forward." Skip film-school terms.
5. **Lighting.** Light that physically exists. "Low orange dusk light", never "cinematic lighting".
6. **Style.** Last, once the scene is set. "Photorealistic", "watercolor animation", "grainy security footage".

Put together: "A lone astronaut walking slowly toward a glowing alien structure on a barren red plain, dust drifting across the ground, camera slowly pushes forward, low orange dusk light, photorealistic."

Two rules sit under all of this. Describe a scene, do not give commands: "a smiling man as the camera drifts toward him" beats "make the man smile and zoom in". And keep it to 20 to 50 words. Below that the model guesses. Above 75 it starts ignoring parts.

## Five mistakes that ruin prompts

![Common prompt mistakes and how to fix them](/blog/generated/ai-video-prompts-guide-mistakes.webp)

- **Too many actions.** "A dog runs through a park, jumps in a lake, shakes off" is three clips. Models render one coherent moment. Split it.
- **Empty adjectives.** "Beautiful", "amazing", "epic", "8k" carry no visual information. Swap in specifics: "deep orange sunset over calm water, palm trees in silhouette".
- **Contradictions.** "Dark moody lighting in a bright sunny room." Every word should push in the same direction.
- **Commands.** "Zoom in", "make it scary". Describe the result instead: "camera slowly pushes toward the door, a single bulb swinging overhead".
- **Nothing moving.** A video prompt with no motion gives you a slideshow. Add wind, water, steam, or a walking figure.

## Horror and thriller prompts

![Horror and thriller prompts](/blog/generated/ai-video-prompts-guide-horror.webp)

Horror is atmosphere, not gore. Models render fog, shadow, and slow reveals well, and refuse or botch graphic violence. Describe one moment of dread, not a sequence.

**What works:** security-camera and found-footage styles for authenticity; ordinary objects moving on their own; motion that breaks physics, like a tree swaying against the wind; a camera pulling backward so the viewer feels they are retreating.

1. A flickering hallway light reveals a shadowy figure standing motionless at the far end of a hospital corridor, grainy desaturated security camera footage
2. An abandoned Victorian dollhouse on a dusty attic floor, one tiny door slowly creaking open by itself, a single beam of light from the attic window
3. Dense fog rolling across a moonlit cemetery, a lone willow tree swaying against the wind direction, camera slowly pushes forward
4. A child's music box on a windowsill starts playing by itself at 3 AM, pale blue moonlight casting long shadows across an empty nursery
5. Rain streaking the window of a parked car on an empty road at night, a pale face appearing briefly in the rearview mirror then vanishing, seen from the driver's seat
6. An old rotary phone ringing in a condemned building thick with dust and cobwebs, the receiver slowly lifting off the hook on its own, wide shot with a soft background
7. A long dark staircase descending into a basement, a bare lightbulb swinging and throwing moving shadows on concrete walls, handheld camera slowly approaching the stairs
8. A forest trail at twilight where every tree has a carved face with hollow eyes, mist creeping along the ground, camera slowly pulling backward
9. An empty school classroom at night, chalk writing appearing on the blackboard letter by letter, fluorescent lights buzzing and flickering overhead
10. A dimly lit bathroom mirror whose reflection moves half a second behind the person in front of it, close-up on the glass, warm tungsten light

Making shorts in this genre? See [how to create viral AI horror videos](/blogs/ai-horror-videos-tiktok).

## Children's and educational prompts

![Children's and educational prompts](/blog/generated/ai-video-prompts-guide-children.webp)

Bright colors, friendly faces, gentle movement, simple compositions. Always name an animation style, because photorealistic children drift into uncanny territory. For educational clips, show a process step by step on a clean background.

**What works:** naming a studio style (Pixar, Ghibli, storybook watercolor) to anchor the look; giving objects faces; watercolor and cel animation, which hide AI imperfections; first-time discovery moments.

1. A friendly cartoon fox in a red scarf teaching the alphabet on a chalkboard in a cozy forest classroom, Pixar-style 3D animation, warm golden light
2. Colorful planets orbiting a smiling cartoon sun, each planet with a friendly face, flat illustration style on a pastel background
3. A baby elephant learning to spray water from its trunk on a sunny savanna, droplets catching rainbow light, watercolor animation
4. A group of children planting seeds in a garden, flowers growing from the soil in time-lapse, bright storybook illustration with soft outlines
5. A cheerful robot helping a child build a sandcastle on a sunny beach, waves lapping gently behind them, clean cel animation with bold outlines
6. The numbers one to ten as bouncing characters on a playground, each a different bright color, 2D animation with confetti
7. A curious kitten discovering snow for the first time, pawing at snowflakes in a backyard garden, soft pastel palette, camera gently tracking
8. A smiling sea turtle swimming past a colorful coral reef, small fish swimming alongside in formation, Pixar-style rendering with light rays through the water
9. A paper airplane folding itself step by step from a flat sheet, then flying out an open window into a blue sky, stop-motion paper craft style
10. A treehouse in a giant oak with tiny windows glowing at twilight, a rope ladder swinging gently, fireflies drifting through the scene, storybook watercolor

More in [how to create AI bedtime story videos for kids](/blogs/ai-childrens-story-videos).

## History prompts

![History prompts](/blog/generated/ai-video-prompts-guide-history.webp)

Name the century, the region, and a visual treatment (documentary, oil painting, sepia photograph). Do not ask for real historical figures. Most models restrict them or render them badly. Describe the clothing, architecture, and activity instead.

**What works:** wide shots for large scenes; architecture over characters; a named light source such as candles, windows, or fire; quiet meditative scenes, which are among the most reliable outputs.

1. A bustling 14th-century European marketplace, merchants selling spices and textiles under canvas canopies, cobblestone streets, warm afternoon light, documentary realism
2. Ancient Egyptian workers hauling limestone blocks up a sand ramp toward a half-built pyramid, vast desert, golden hour, wide shot
3. A Viking longship cutting through stormy North Atlantic waves, warriors bracing against the rain, dark clouds and lightning, camera tracking from above
4. A 1920s jazz club, musicians playing on a small stage, art deco decor, cigarette smoke drifting through amber spotlights, warm sepia tones
5. A Silk Road camel caravan crossing dunes at sunset, traders in traditional robes, long shadows across rippled sand, drone-style documentary shot
6. A Roman forum full of citizens in togas debating, marble columns and statues, sharp Mediterranean sunlight, camera walking through the crowd at eye level
7. A samurai meditating in a Zen rock garden in cherry blossom season, petals falling slowly, traditional Japanese architecture behind, wide contemplative shot
8. Medieval monks illuminating manuscripts in a candlelit scriptorium, gold leaf catching the flickering light, close-up on hands and parchment
9. An ancient Greek amphitheater carved into a hillside above the sea, actors performing far below, late golden light, sweeping aerial shot
10. An Industrial Revolution factory floor, steam-powered machinery, workers turning large metal gears, shafts of light through dirty windows, gritty documentary style

See also [the complete guide to AI history videos](/blogs/ai-history-videos-guide).

## Real estate prompts

Sell space, light, and flow. Always specify a slow camera move: a push through a doorway, a pan across a kitchen, a drone descent onto the roof. Twilight exteriors with the interior lights on are the most aspirational shot in the genre.

1. Camera slowly gliding through a modern open-plan living room, floor-to-ceiling windows over a city skyline, minimalist furniture, late afternoon sun streaming in
2. Drone slowly descending toward a Mediterranean villa with a terracotta roof, swimming pool, and manicured gardens, clear blue sky, golden hour
3. A modern kitchen with marble counters and brass fixtures, camera panning slowly across the island to reveal a bright breakfast nook, natural morning light
4. Twilight exterior of a contemporary home, warm interior lights glowing through large windows, path lights along the walkway, deep blue dusk sky
5. A sunlit bathroom with a freestanding tub under a skylight, white marble floor, green plants on the shelves, camera slowly pulling back to reveal the full room

Turning listing photos into tours? Read [AI real estate videos](/blogs/ai-real-estate-videos), or see how to [turn listing photos into a video without inventing rooms](https://vidlistings.com/blog/ai-real-estate-video).

## General purpose prompts

Your everyday toolkit for social posts, B-roll, and product clips. A static wide shot with one subtle motion, like flags or waves, is the most reliable output any model produces.

**What works:** wet surfaces, which double every light source with reflections; hands instead of faces; film grain or vintage treatments to mask artifacts; natural phenomena like auroras and storms, which models have seen millions of times.

1. A steaming cup of coffee on a wooden table by a rainy window, droplets running down the glass, warm interior light, close-up with a soft background, lo-fi mood
2. Hands typing fast on a laptop, code reflected in the typist's glasses, dark room lit only by the monitor, close-up with a soft background
3. A neon-lit Tokyo street at night after rain, reflections on wet pavement, pedestrians with umbrellas, handheld camera at street level, cyberpunk color grading
4. Time-lapse of a desert from sunrise to sunset, shadows sliding across the dunes, stars appearing as the sky darkens, locked-off wide shot
5. A golden retriever running in slow motion through shallow ocean water at sunset, water splashing around its legs, backlit by warm sun, long lens
6. Fresh fruit falling into clear water in slow motion against a clean white background, splashes and bubbles, product photography lighting
7. Northern lights dancing over a frozen Arctic lake, green and purple ribbons reflected in the ice, wide night-sky shot
8. A vinyl record spinning on a turntable, needle in the groove, warm room light, a bookshelf soft in the background, overhead close-up, vintage film grain
9. Aerial flight over a winding mountain road through autumn forest, vivid red and orange foliage, morning mist in the valleys
10. A paper boat floating down a rain gutter stream on a suburban street, leaves and petals flowing alongside, low macro angle, soft overcast light
11. A chef's hands sprinkling salt flakes onto a seared steak, steam rising, dark kitchen with one focused overhead light, extreme slow motion
12. A hot air balloon rising above a misty valley at dawn, patchwork farmland below, soft pastel sky, camera rising smoothly with the balloon
13. Close-up of an old mechanical watch with exposed gears turning, golden parts reflecting warm light, macro shot on a black background
14. A lone figure walking a misty forest path in autumn, fallen leaves covering the ground, light filtering through the canopy, camera tracking slowly from behind
15. Bioluminescent jellyfish drifting through dark ocean water, blue and green glow lighting the particles around them, macro underwater shot, slow motion
16. A row of colorful beach huts on white sand with turquoise water behind, a light breeze moving the flags on each roof, locked-off wide shot, vivid colors

## How to adapt these

Keep the shape, swap the parts. Change the subject and environment, keep the camera and lighting language. One action per clip. If your story needs three actions, write three prompts and cut between them in the editor. FrameSurfer plans scenes from a narrative description, so you can paste a paragraph and let it write the shot-level prompts, or write them yourself with the structure above. [Text-to-video AI in plain English](/blogs/what-is-text-to-video-ai) covers what happens next.

## Frequently Asked Questions

### How long should an AI video prompt be?

20 to 50 words. Shorter leaves too much to the model. Over 75 words, prompts start to contradict themselves or get partially ignored.

### Do different AI video tools need different prompts?

The shape is the same everywhere. Some models have preferences: Runway responds well to camera movement, Sora handles complex physics better than most, and FrameSurfer plans scenes from narrative descriptions, so paragraphs work.

### Can I reuse an image prompt for video?

As a starting point, yes. Then add motion. "A mountain lake at sunset" becomes "a mountain lake at sunset, a light breeze rippling the surface, camera panning slowly along the shoreline."

### Why does my video look nothing like the prompt?

Usually vague words, a contradiction, or something the model cannot do well: legible text, brand logos, or a specific real face. Cut the prompt back to subject, action, environment, camera, lighting, and style.

### Should I use negative prompts?

Only if your tool supports them, and most consumer video generators do not. Where they exist, "no text overlay, no watermark" is useful. Otherwise, make the positive prompt specific enough that the unwanted thing never comes up.

## Related Articles

- [What Is Text-to-Video AI?](/blogs/what-is-text-to-video-ai)
- [Best AI Video Generators in 2026](/blogs/best-ai-video-generators)
- [10 AI Video Ideas for TikTok That Actually Get Views](/blogs/tiktok-video-ideas-ai)