Cinematic AI Video Prompts: One Unified Template + 20 Examples (2026)

A cinematic AI video prompt works best when it follows one unified order: shot and camera move, subject, action, setting, lens, light and palette, mood, then sound. Fill every slot with something the camera can actually see or hear, keep one camera move and one action per shot, and pick the model by scene: Seedance 2.5 for long multi-shot sequences, Veo 3.1 for dialogue with synced sound, Wan 3.0 for cheap drafts. Below you get the template, a shot-term cheat sheet and 20 copy-paste prompts grouped by scene type (as of November 2, 2026).

This guide is the "cinematic" layer on top of two posts we already have. If you want general prompt examples, read AI video prompts: 42 copy-paste examples. If you only need camera moves, the 18-move camera movement cheat sheet goes deeper. Here we stitch everything into one prompt that reads like a shot from a film.

What is a unified cinematic video prompt?

A unified cinematic prompt is a single paragraph that briefs the video model the way a director briefs a cinematographer. Instead of a pile of keywords ("cinematic, 4k, epic, dramatic"), it answers eight questions in a fixed order:

[Shot + camera move] + [Subject] + [Action] + [Setting] + [Lens] + [Light + palette] + [Mood / film look] + [Sound]

Google's own Veo 3.1 prompting guide uses a shorter version of the same idea, cinematography + subject + action + context + style and ambiance (Google Cloud). OpenAI's Sora 2 guide describes prompting as briefing a cinematographer and recommends one camera move and one subject action per shot (OpenAI Cookbook). Our unified template adds two slots those formulas fold into "style": a separate lens slot (because focal length and depth of field change the whole feel) and a separate sound slot (because most current video models generate audio).

Here is the template filled in once, so you can see the order. Putting the shot first also stops you forgetting it, and a missing camera direction is the most common reason a clip looks flat; Adobe's Firefly guide makes the same point (Adobe).

Here it is:

Slow dolly-in, medium close-up. A middle-aged lighthouse keeper in a wool coat lights an oil lamp, then looks out the window. Inside a cramped stone lighthouse room at dusk, a storm building over the sea. 50mm lens, shallow depth of field, background softly out of focus. Warm amber lamplight against cold blue window light, palette of amber, slate blue and charcoal. Quiet, tense, shot like a 1970s drama on 35mm film with fine grain. Sound: wind howling against glass, the click of the lamp, distant waves.

Every slot holds something visible or audible. Nothing says "beautiful" or "high quality", because the model cannot film an adjective.

Which shot terms should you use? (cheat sheet)

Use these words exactly; video models were trained on captions that use standard film vocabulary.

How do you write a cinematic prompt step by step?

1. Decide the shot first. One shot size and one camera move. If you need several moves, you need several shots (see the multi-shot section below).
2. Name one subject with two or three visible details. "A woman in a yellow raincoat with a red umbrella" holds up better across frames than "a woman".
3. Give one action, in beats. "Stops, turns, looks up at the sign" is filmable; "feels nostalgic" is not.
4. Place it. Location plus time of day plus weather in one sentence.
5. Pick a lens and focus. One focal length and one depth-of-field note.
6. Light it with sources, not adjectives. "Neon sign from the left, wet asphalt reflecting it" instead of "moody lighting". OpenAI also suggests naming three to five palette colors to keep the look stable.
7. Set the film look. An era or stock ("1990s music video, 16mm") does more than "cinematic".
8. Add sound if the model makes audio. Ambience, one or two effects, and any spoken line in quotes. Google's Veo guide uses quotes for dialogue and prefixes like SFX: for effects.
9. Set length, resolution and aspect ratio in the settings, not the prompt. Writing "make it 20 seconds" does nothing; the duration control does.

20 copy-paste cinematic prompts by scene type

All prompts are SFW and written in the unified order. The model note after each one says which model on Kenerate suits it; you can paste any of them into the AI video generator and switch models from the dropdown.

Establishing shots

1. Desert town at dawn — Extreme wide shot, slow aerial rise. A lone pickup truck drives into a small desert town, dust trailing behind it. Low adobe houses, a water tower, empty main street at sunrise. 24mm lens, deep focus. Golden-hour light from the right, long shadows, palette of ochre, rust and pale blue. Quiet western mood, 35mm film grain. Sound: wind, distant engine, a dog barking once. (Seedance 2.5)
2. Rainy neon city — Wide establishing shot, slow crane down from rooftop level to street. Crowds with umbrellas cross a busy intersection. A dense Asian city at night in heavy rain. Anamorphic lens, horizontal flares. Neon signs in magenta and cyan reflecting on wet asphalt. Melancholic, 1990s Hong Kong cinema look. Sound: rain, traffic hiss, muffled street music. (Seedance 2.5)
3. Mountain monastery — Extreme wide shot, slow push-in. Monks in red robes walk single file up stone steps toward a cliffside monastery. Himalayan peaks, clouds drifting below the buildings. 35mm lens, deep focus. Soft overcast light, palette of red, slate grey and white. Calm, documentary-epic. Sound: wind, a single temple bell. (Seedance 2.5 or Wan 3.0 draft)

Chase and action scenes

4. Rooftop run — Tracking shot from behind, handheld. A parkour runner in a grey hoodie sprints across rooftops and vaults a low wall. Dense city rooftops at sunset. 24mm lens. Hard backlight, sun flare, palette of orange and steel blue. Urgent, kinetic. Sound: footsteps on gravel, heavy breathing, wind. (Seedance 2.5)
5. Car chase at night — Low-angle tracking shot alongside a vintage red sports car racing through a tunnel, then out onto a coastal road. 35mm anamorphic. Sodium tunnel lights strobing across the car, then moonlight on the sea. 1970s heist film look. Sound: engine roar, tyre squeal, echo in the tunnel. (Seedance 2.5)
6. Market escape — Handheld medium shot. A courier in a leather jacket pushes through a crowded spice market, glancing back over his shoulder. Narrow alley, hanging lanterns, stalls of colourful powders. 35mm lens, shallow depth of field. Shafts of hard midday light through fabric awnings. Tense. Sound: crowd chatter, a vendor shouting, footsteps. (Wan 3.0 for drafts, Seedance 2.5 final)

Dialogue scenes

7. Diner confession — Two-shot, locked-off, then slow dolly-in. Two old friends sit across a diner booth; the woman says, "I'm not coming back this time." The man looks down at his coffee. 1980s roadside diner at night. 50mm lens, shallow depth of field. Neon sign outside casting red, warm practical lamps inside. Quiet, heavy. Sound: low diner hum, cup set on saucer. (Veo 3.1)
8. Detective office — Medium shot. A tired detective looks up from his desk and says in a gravelly voice, "You're late." Cluttered 1940s office, rain on the window. 50mm lens. Venetian-blind shadows across his face, single desk lamp. Film noir, black and white. SFX: rain, a ticking clock. (Veo 3.1)
9. Mission briefing — Medium close-up, slow orbit. A starship captain in a navy uniform says calmly, "We have one chance. Make it count." Bridge with glowing consoles. 85mm lens, shallow depth of field. Cool blue key light, orange console glow as rim light. Restrained sci-fi drama. Ambient noise: the hum of the ship. (Veo 3.1)

Product hero shots

10. Perfume reveal — Macro close-up, slow orbit. A glass perfume bottle on black stone, a single water droplet running down its side. Studio. Macro lens, shallow depth of field. Hard rim light outlining the bottle, soft gold reflection. Luxury commercial look. Sound: soft low synth swell. (Seedance 2.5)
11. Sneaker drop — Low-angle shot, slow push-in. A white running shoe lands on wet concrete, splashing water in slow motion. Empty city street at night. 35mm lens. Neon blue and pink backlight, mist. High-energy sports ad. Sound: impact splash, bass hit. (Seedance 2.5)
12. Coffee pour — Top-down shot, locked-off. Espresso pours into a white cup, crema forming. Marble counter. 50mm macro. Soft window light from the left, palette of cream, brown and white. Calm morning ad. Sound: pouring, gentle café ambience. (Wan 3.0)
13. Watch on the wrist — Close-up, slow rack focus from the city skyline to a steel watch on a man's wrist. Rooftop terrace at golden hour. 85mm lens. Warm sun flare, palette of gold and steel. Confident, premium. Sound: soft wind, a single tick. (Seedance 2.5; start from your own product photo for exact details)

Nature scenes

14. Whale breach — Wide shot, slow aerial pull-back. A humpback whale breaches beside a small boat. Open ocean, overcast. 24mm lens. Silver diffused light, palette of grey-blue and white. Awe. Sound: splash, wind, distant calls. (Seedance 2.5)
15. Forest morning — Slow dolly forward through a misty pine forest, a deer lifting its head. Early morning. 35mm lens. Shafts of sunlight through fog, palette of green, gold and grey. Serene nature documentary. Sound: birdsong, a twig snapping. (Wan 3.0)
16. Storm over wheat — Locked-off wide shot. A lightning storm rolls over a wheat field, the stalks bending in the wind. Midwest farmland. 24mm lens. Dark sky lit by flashes, golden wheat. Dramatic. Sound: thunder, wind through wheat. (Wan 3.0)

Sci-fi scenes

17. Arrival — Extreme wide shot, slow crane up. A huge smooth spacecraft hovers silently over a misty valley as villagers watch. Rural landscape at dawn. Anamorphic lens. Pale fog, cold light, palette of grey, green and white. Quiet awe. Sound: low drone, birds going silent. (Seedance 2.5)
18. Android awakening — Close-up, slow dolly-in. A humanoid android opens its eyes on a lab table, pupils adjusting. White lab. 85mm lens, shallow depth of field. Cool overhead light, a blue LED rim light. Clinical, eerie. Sound: servo whir, soft electronic hum. (Veo 3.1 or Seedance 2.5)
19. Mars walk — Tracking shot alongside an astronaut walking across red dunes toward a habitat dome. Mars surface, dusk. 24mm lens. Low butterscotch sky, long shadows. Lonely, epic. Sound: breathing in a helmet, wind. (Seedance 2.5)
20. Cyberpunk alley — Handheld medium shot. A courier on an electric bike weaves through a crowded alley under holographic ads. Futuristic city at night. 35mm anamorphic. Magenta and teal neon, steam from vents. Gritty. Sound: crowd, electric hum, rain. (Wan 3.0 draft, Seedance 2.5 final)

Which model should you use for which cinematic scene?

Credit prices below come from Kenerate's video model config as of November 2, 2026. Video generation needs a paid plan.

A practical workflow: draft the prompt on Wan 3.0 at 480p (a 5-second test costs 75 credits), fix the slots that are wrong, then render the final on Seedance 2.5 or Veo 3.1. Our Seedance 2.5 page lists every setting for the long-form model.

How do you prompt several shots in one clip?

Write one block per shot, each with its own timing, shot size, move and action. Google's Veo guide shows this with timestamps such as [00:00-00:02], one shot per segment. Keep the subject description identical in every block so the character stays the same:

[00:00-00:04] Wide shot, slow crane down. A woman in a yellow raincoat with a red umbrella crosses a rainy neon street. [00:04-00:08] Close-up, locked-off. The same woman in a yellow raincoat stops and looks up at a flickering sign, rain on her face. [00:08-00:12] Tracking shot from behind. The same woman walks into a small noodle bar, warm light spilling out.

Seedance 2.5 accepts up to 30 seconds in one generation, which leaves room for several shots. For the full method, read our multi-shot AI video guide.

What mistakes make AI video look un-cinematic?

• Keyword soup. "Cinematic, 8k, masterpiece, epic" gives the model nothing to film. Replace each word with a lens, a light source or a film stock.
• Two camera moves in one shot. "Dolly in while orbiting and craning up" usually produces wobble. Pick one.
• Abstract actions. "She feels lost" becomes "she stops, looks left, then right, and pulls her coat tighter".
• Negative wording. Google recommends describing what you want ("an empty desert road") instead of what you don't ("no cars").
• Settings in the prompt. Duration, aspect ratio and resolution belong in the controls.
• Changing everything between drafts. Change one slot per re-run so you learn what each slot does.

Start with prompt 1 or prompt 8 above, run a cheap draft, and edit one slot at a time. If you prefer starting from a still, generate the first frame and animate it with image to video; the frame locks the look and the prompt only has to describe motion and sound.