The Ultimate Guide to AI Video Prompts (With Copy-Paste Examples)

Video prompting is a different skill from image prompting. An image prompt describes a frozen moment; a video prompt has to choreograph time — what moves, how the camera behaves, and how the shot evolves from first frame to last. Most bad AI videos come from image-style prompts fed into video models.

This guide gives you a repeatable structure, a vocabulary of camera language that models actually understand, and copy-paste examples you can adapt inside the AI video generator.

The Anatomy of a Video Prompt

Every effective video prompt answers four questions, in roughly this order:

1. Subject & setting — who or what, where. One clear subject beats three competing ones.
2. Action — what happens during the clip. Pick ONE primary action; a short clip has time for one beat, not a plot.
3. Camera — how the shot is filmed: framing, movement, lens feel.
4. Style & mood — lighting, color, film grade, atmosphere.

A template you can reuse:

[Shot type] of [subject] in [setting]. [One primary action]. [One camera move]. [Lighting/mood/style].

Example:

Close-up of a barista pouring latte art in a sunlit café. Steam rises as the milk swirls into a rosetta. Slow push-in. Warm morning light, shallow depth of field, cinematic color grade.

Camera Language That Models Understand

Video models were trained on footage described with real film vocabulary. Use it — precise terms outperform vague ones every time.

Framing:

• "extreme close-up", "close-up", "medium shot", "wide shot", "aerial shot"
• "low angle" (subject feels powerful), "high angle" (subject feels small), "eye level"

Movement:

• "static camera" — often the most cinematic choice; movement in frame, camera still
• "slow push-in" / "pull-back" — builds or releases tension
• "pan left/right", "tilt up/down" — reveals
• "orbit" / "arc shot" — dimensionality around a subject
• "tracking shot" — follows a moving subject
• "handheld" — documentary energy; "crane up" — grand reveals

Lens feel:

• "shallow depth of field", "bokeh background" — subject isolation
• "wide-angle lens" — drama and scale; "telephoto compression" — tight, intimate

One move per clip. "Slow push-in" reads as intentional cinematography; "camera flies around while zooming and tilting" reads as chaos.

Motion Words Do the Heavy Lifting

The difference between a static-feeling clip and a living one is usually ambient motion — small, physical, continuous movements the model can render believably:

• People: "hair moves in the breeze", "she blinks and smiles slowly", "coat flutters"
• Nature: "waves roll in", "leaves drift down", "clouds move slowly overhead"
• Objects: "steam rises", "rain streaks the window", "neon signs flicker"
• Light: "sunlight shifts through the trees", "headlights sweep across the wall"

Stack one subject action with one or two ambient motions and the frame comes alive without overwhelming the model.

Copy-Paste Prompt Examples

Adapt the nouns, keep the structure:

Travel / lifestyle (9:16 for Reels):

Medium shot of a traveler in a yellow raincoat standing on a cliff over the ocean at golden hour. Her coat and hair move in strong wind as she looks toward the horizon. Static camera. Moody cinematic grade, soft rim light.

Product (great with image-to-video):

Close-up of a perfume bottle on wet black stone. Water droplets run down the glass as soft light sweeps across the label. Slow orbit right. Luxury commercial style, dark background, glossy highlights.

Food:

Overhead shot of a chef's hands tossing fresh pasta in a steel pan. Steam bursts upward, flames flare briefly at the edge. Static overhead camera. Warm kitchen light, rich color, shallow depth of field.

Cinematic establishing shot:

Wide aerial shot of a mist-covered mountain village at dawn. Fog drifts slowly between rooftops as lights turn on one by one. Slow crane up. Ethereal blue-hour palette, epic scale.

Character moment:

Close-up of an elderly fisherman's weathered face lit by lantern light on a night boat. He exhales slowly, breath visible in the cold air. Slight handheld movement. Documentary realism, warm-against-cold lighting.

Model-Specific Tips

• Veo responds well to full cinematic descriptions — lens language, lighting setups, and color grades all land. Use it when the clip is the hero asset.
• Seedance is a dependable all-rounder for short-form content; clean structure and explicit camera moves get the most out of it.
• Hailuo shines on expressive character motion — give it clear, physical action beats.
• Whatever the model, shorter clips are more coherent. If a moment matters, give it its own clip rather than cramming three beats into one generation.

Iterating Without Wasting Credits

1. Draft cheap, finish strong. Test the idea on a fast model, then re-run the winning prompt on a flagship one.
2. Change one variable at a time. If the motion is right but the mood is wrong, touch only the style clause.
3. Reuse what works. Keep a personal prompt library — the camera + lighting patterns that work for your niche transfer across subjects.
4. Go image-first for control. When composition matters, generate a still and animate it — see our image-to-video guide.

Common Mistakes to Avoid

• Writing a novel. Past a few sentences, extra detail competes with itself. Cut anything the viewer can't see in the shot.
• Multiple scenes in one prompt. "She wakes up, goes to the beach, then dances at sunset" — that's three clips, not one.
• Vague motion. "Dynamic", "epic movement", and "action-packed" give the model nothing physical to render.
• Forgetting the format. Prompt for the platform: vertical framing language for Shorts and Reels, wide framing for YouTube.

Open the AI video generator, start from one of the templates above, and swap in your subject. Structure first, style second — that's the whole craft.