AI Video Prompt Examples: 12 Templates for Five-Second Clips

The templates below are written prompts, not generated results. Each one is built for a single five-second shot on the verified model, and each includes a note explaining the choices so you can adapt it rather than copy it blindly.

The structure formula

Every template follows the same order:

subject + one action + one camera move + light + visual style + constraints

Put the subject first to anchor the frame, name the camera once near the end, and finish with the constraints you already know you need.

Aim for 25 to 45 words. Longer prompts add competing ideas a five-second clip cannot resolve.

Product shots

1. Perfume bottle orbit

A frosted glass perfume bottle on a dark stone plinth, a slow camera orbit from left to right, one soft key light from above, minimal luxury product photography, no text, no hands.

Why it works: a slow orbit is one of the most reliable motions at this length, and a single light source keeps the reflections consistent from first frame to last. "No hands" removes a common source of malformed fingers.

2. Cream texture close-up

A dollop of white cream on a matte ceramic tile slowly spreading, camera locked off overhead, soft diffused daylight, clean beauty product photography, no text.

Why it works: the motion is material rather than mechanical, so small imperfections read as texture. A locked camera means the model only has to animate one thing.

3. Shoe turntable

A plain white running shoe rotating slowly on a pale grey turntable, camera fixed at eye level, soft studio lighting, e-commerce product video, no logos, no text.

Why it works: the turntable supplies the movement, so the camera does not need to move at all. The logo constraint matters because small text and brand marks are where generated video usually breaks down.

Landscapes

4. Valley sunrise push-in

A quiet mountain valley at sunrise, mist drifting between pine trees, a slow camera push forward, warm low sun, cinematic natural light, no people.

Why it works: mist keeps the scene moving even when nothing else does, and a push-in reads clearly in five seconds.

5. Black sand coast

Waves rolling onto a black sand beach, the camera slowly rising and tilting down, overcast diffused light, documentary aerial style, no buildings.

Why it works: the rising camera reveals the beach instead of panning across it. Overcast light avoids the flickering highlights that hard sun can produce.

6. Forest path after rain

A narrow path through a mossy forest after rain, camera tracking forward at walking height, dappled light through the canopy, cinematic realism, no people.

Why it works: tracking at walking height gives the viewer a body in the scene. Dappled light is forgiving because small changes in shadow look natural.

Character motion

7. Hooded walker at night

A lone traveller in a hooded coat walking away down a rain-soaked street at night, camera tracking behind, neon reflections on wet asphalt, cinematic, no face visible.

Why it works: a back view avoids the face, which is the least stable part of a generated person. Walking away is a simple, repeatable motion, and wet asphalt adds detail without adding subjects.

8. Backlit dancer

A dancer silhouette spinning slowly in front of a bright window, camera locked off, strong backlight, high-contrast black and white, no text.

Why it works: a silhouette hides facial detail and rewards shape and timing. One turn is enough movement for five seconds, and the locked camera keeps the frame clean.

9. Hands and a wooden box

A pair of hands turning a small wooden box over, close-up, soft window light, shallow depth of field, no face.

Why it works: hands are difficult, so the prompt keeps them slow and gives them one object to hold. Shallow depth of field reduces how much detail has to stay consistent.

Vertical social clips

10. Iced coffee pour

A slow pour of milk into a glass of iced coffee, camera locked off overhead, bright soft daylight, close-up vertical framing, no text.

Why it works: a top-down composition survives small screens because the subject fills the frame. Pouring is a single continuous action with a clear start and end.

11. Crosswalk walk

A rainy city crosswalk seen from chest height, people with umbrellas moving away, a slow camera walk forward, cool evening light, vertical composition, no readable signs.

Why it works: "no readable signs" prevents garbled lettering, which is common in street scenes. Chest height and a forward walk make the viewer part of the crowd.

12. Headphones case reveal

A matte black headphones case opening slowly on a pale background, the camera slowly pushing in, soft studio light, vertical product video, no logos.

Why it works: the opening lid provides the action and the push-in adds emphasis without a second subject. A pale background keeps the product edge clean.

Six failure modes to design around

1. Two shots in one prompt. The model blends them into a morph. Write one continuous take per generation.
2. Too many subjects. People, objects and animals compete for attention and identities swap. Keep one hero.
3. Vague camera language. Words like "dynamic" or "epic" produce random movement. Name a move: push in, pull back, orbit, track, or locked off.
4. Conflicting light. Neon and bright daylight in the same prompt give inconsistent frames. Pick one source and one direction.
5. Asking for text or logos. Generated lettering is usually unreadable. Add "no text, no logos" unless text is the subject, and expect retries even then.
6. Fast action. Running, fighting and dance choreography smear at this length. Slow the action down and let the camera do the work.

Faces deserve their own note: close-ups of faces are the highest-risk subject. A back view, a silhouette, or a wide shot is more likely to give you a usable five seconds.

How to adapt a template

Change one element per attempt. If the product is wrong, keep the camera and lighting and rewrite only the subject phrase. If the motion is wrong, keep the subject and rewrite the camera. That is how you learn which words the model responds to.

Match the template to the frame you will publish in. A vertical composition is not a crop, so rewrite framing language when you switch ratio. Image-to-video is currently 9:16 only, while text-to-video supports both.

Finally, constraints are requests, not guarantees: adding "no text" lowers the chance of lettering, it does not remove it.

Where to take this next

Build a prompt from parts with the free AI video prompt generator, then render it in the studio. If you have not generated before, the step-by-step guide covers input modes, credits and review. Technical setups for each mode live on text-to-video and image-to-video, and current credit prices are on pricing.