AI Video Prompts Library: 20 Copy-Paste Starters by Genre

An ai video prompts library is a shelf of paste-ready starters grouped by genre so you stop staring at an empty text box. Each starter below names subject, timed action, one camera path, scene, light, and optional audio in a form that travels across Kling AI, Runway Gen-4, OpenAI Sora, Google Veo, and Luma Dream Machine as of August 2026. Swap the bracket nouns. Keep the motion grammar. Generate a four-to-five-second test before you burn credits on a ten-second montage wish list. This page is a starter bank, not the SMCD pillar primer in ai video prompt structure and not a single-host dialect encyclopedia. You leave with exactly twenty copy-paste lines across ten genres, bracket variables for reuse, host retarget notes, common mistakes, and a soft path to PromptMake at https://promptmake.net/video when your brief is still messy prose. PromptMake writes motion prompt text only. It does not render MP4 files or spend credits in video hosts.

What an ai video prompts library is for

Libraries solve blank-page latency. You pick a genre that matches the clip job: product hero, talking head, nature B-roll, action beat, food macro, fashion walk, tech UI, travel establish, abstract mood, or documentary slice. You paste a starter. You replace bracket tokens with your nouns. You run a short test. You log the winner next to the genre tag so next week's shoot reuses the same camera path with a new product.

Searchers who type ai video prompts usually want lines they can paste today, not a lecture on cinematography theory. Starters still obey structure. Subject owns the frame. Action changes across seconds. Camera picks one path. Duration matches beats. The library hides that scaffolding inside readable sentences so you move fast without forgetting motion.

This article is not ai video prompt structure, which teaches the SMCD spine in depth. Read that page when you need to understand why each field exists. Stay here when you need twenty genre starters and a habit for bracket swaps. For cross-model portable worksheets, see prompt to video AI guide. For Runway shot-list grammar, see runway video prompts structure.

Who should skip a starter library: teams that only export stills from Midjourney v7 or FLUX and never open a motion UI. Who should lean in: marketers shipping weekly product teasers, creators testing image-to-video on a locked portrait, editors who need B-roll families without rewriting camera verbs every time.

How to use these starters without wrecking credits

Treat every line as a template, not a lottery ticket. Brackets mark swap zones: [product], [face nouns], [place], [color], [seconds]. Freeze camera verbs on the first pass. Change subject nouns only. If the clip fails, fix one axis: action timing, camera path, or seconds in the UI slider. Do not rewrite adjectives across five fields at once.

Default test length is four to five seconds with one action beat and one camera move. Promote to eight or ten seconds only after subject and motion hold on the short box. Image-to-video routes inherit the upload. Prepend preserve language: maintain subject, lighting, and background from reference. Follow with action and camera from the starter. Crop the still to the output aspect before upload.

Log wins in a notes row: genre tag, exact paste, host, aspect, seconds, date. Next week you swap [product] and keep slow dolly in only. That habit turns a blog library into your private kit.

Bracket swap discipline

Change one bracket family per retry: product nouns, then place, then camera move. Stacking five adjective swaps after a failed Kling AI run teaches nothing about which axis broke.

Read the starter aloud before paste. Confirm one subject, one camera path, beats inside target seconds. If the sentence sounds like two clips, split it.

Keep a swap sheet beside this library: your SKU stack, your face nouns, your brand surface colors. Public starters stay generic. Your sheet makes them repeatable.

Image-to-video preserve lines

Prepend maintain face, hair, wardrobe, and background from reference on portrait uploads before action verbs from Starter 8 or any talking-head line.

Product stills need maintain label orientation and cap color from reference when the SKU must match packaging pixels in Runway Gen-4 or Veo.

Crop the reference still to the output aspect before upload. Luma Dream Machine and Sora routes inherit framing from the upload edge.

Host dialect at paste time

Kling AI rewards labeled clauses and second stamps when you stretch beyond five seconds. Runway Gen-4 rewards shot size in the open: medium eye-level, slow dolly in. Sora and Veo prefer dense sentences with duration set in the UI, not inside prose. Luma Dream Machine reads natural sentences when duration stays short. Starter nouns stay fixed. You adjust grammar at paste time, not story facts.

Product and commercial starters

Product clips prove material, label, or a single hero move. Keep backgrounds quiet. One prop motion beats three.

Hero and macro moves

Starter 1: hero push-in. Medium eye-level shot of a [matte white bottle, blue cap, no logo] on [gray stone]. Product static two seconds, then one [condensation bead slides down the left side]. Slow dolly in only over five seconds, 85mm product feel, soft key from upper left. Quiet room, no music.

Starter 3: pour macro. Insert shot of [clear glass over white marble]. [0-1s empty, 1-4s amber liquid pours once from above, fill stops mid-glass]. Locked macro, 100mm feel, backlight through liquid. Soft studio key. Five seconds. Quiet pour ambience only.

Unbox and rotate beats

Starter 2: unbox tease. Close-up of [kraft mailer on oak desk, brand sticker visible]. Hands enter frame, [0-2s static, 2-4s tab lifts once, no full open]. Locked tripod, shallow depth on sticker. Warm desk lamp from right. Four seconds. No score.

Starter 4: rotate reveal. [Matte black headphones on turntable plinth]. Plinth rotates [15 degrees over four seconds], product stays centered. Locked camera, product spotlight, dark gray sweep background. No camera orbit. Four seconds.

Portrait and talking-head starters

Portrait motion stays subtle. Micro-expressions and gentle camera moves preserve identity on image-to-video routes.

Micro-expression beats

Starter 5: interview nod. Medium close-up of [woman, short black hair, gray blazer] against [soft charcoal backdrop]. [0-2s holds eye contact, 2-4s single slow nod]. Locked tripod, 50mm feel, soft key camera left. Quiet room tone. Four seconds.

Starter 6: smile beat. Close-up of [man, stubble, navy crewneck] by [window light]. [0-3s neutral, 3-5s brief natural smile]. Slow push-in only, shallow depth on eyes. Overcast daylight key. Five seconds. No music.

Starter 7: turn to camera. Medium shot of [creator, red hoodie] at [desk with monitor glow]. [0-2s profile, 2-5s turns head to lens once]. Gentle pan to follow head turn only, no orbit. Cool monitor fill, warm desk lamp. Five seconds.

Image-to-video identity lock

Starter 8: image-to-video preserve. Maintain face, hair, wardrobe, and background from reference. [0-2s static, 2-4s eyes widen slightly, one breath]. Locked frame, no wardrobe change. Soft key unchanged. Four seconds.

Use Starter 8 as a prefix line on any portrait still before you paste Starter 5 through 7 motion clauses into Kling AI or Luma Dream Machine.

When identity drifts, shorten duration before you rewrite face nouns. Four seconds with one micro-beat beats ten seconds of expression drift on Veo portrait routes.

Nature and landscape starters

Nature B-roll sells place mood with one environmental motion: wind, water, light shift, or cloud drift.

Coastal and forest moods

Starter 9: coastal mist. Wide shot of [dark basalt cliffs, gray ocean]. [0-3s mist rolls inland, 3-6s wave breaks once on rocks]. Slow pan left only, overcast daylight. Six seconds. Wind and surf ambient, no music.

Starter 10: forest sun shaft. Medium wide of [tall pine trunks, moss floor]. [0-4s dust motes drift in sun shaft, 4-6s leaf falls once]. Locked tripod, golden hour key from upper right. Six seconds. Quiet forest ambience.

Weather and atmosphere beats

Starter 11: desert heat. Wide of [red sand dunes, clear sky]. [0-5s heat shimmer above ridge line, 5-8s small sand plume skims crest]. Slow aerial drift forward only, harsh midday sun. Eight seconds. Dry wind bed.

Starter 12: rain window. Medium of [rain streaks on glass, blurred city bokeh beyond]. [0-5s droplets slide down glass, 5-7s lightning flash once in background]. Locked interior frame, cool blue ambient. Seven seconds. Rain on glass sound.

Action and sports starters

Action starters time one beat. Avoid montage language in a five-second box.

Track and gym beats

Starter 13: sprint pass. Low-angle medium of [runner in yellow singlet on wet track]. [0-2s crouch, 2-5s explodes into three strides past camera]. Lateral truck right at running speed, 35mm feel. Overcast stadium light. Five seconds. Footfall and crowd bed low.

Starter 15: boxing jab. Medium of [boxer, red gloves, dark gym]. [0-2s guard up, 2-4s single jab to heavy bag, bag swings once]. Handheld subtle sway, tungsten gym practicals. Four seconds. Impact thud, room echo.

Street and road motion

Starter 14: skate trick. Wide of [concrete bowl, graffiti wall]. [Skater in white tee drops in, 0-3s roll, 3-5s kickflip once, lands]. Locked wide, late afternoon sun. Five seconds. Wheel rumble, no music.

Starter 16: cycling climb. Side medium of [cyclist, teal kit on mountain road]. [0-4s pedals uphill, 4-6s shifts grip once]. Parallel truck left slow, morning side light. Six seconds. Chain and wind ambience.

Food, fashion, tech, and travel starters

The remaining genres round out a weekly content calendar without new camera religion.

Food and fashion

Starter 17: food steam. Close-up of [ramen bowl, chashu, green onion]. [0-2s static, 2-5s steam curls once from broth]. Slow dolly in, warm overhead practical, dark wood table. Five seconds. Quiet kitchen room tone.

Starter 18: fashion walk. Full-length of [model, ivory trench, city crosswalk]. [0-3s walk toward camera three steps, 3-5s stops]. Slow dolly back, soft overcast key. Five seconds. City ambience muted.

Tech UI and travel establish

Starter 19: tech UI glow. Close-up of [laptop screen showing analytics dashboard, dark office]. [0-3s cursor moves once to highlight spike, 3-5s graph animates up one bar]. Locked tripod, screen glow on face reflection optional. Five seconds. Quiet office hum.

Starter 20: travel establish. Wide aerial of [old town rooftops, terracotta tiles, church spire]. [0-6s slow forward drift over roofs, 6-8s bell tower enters center]. Gentle forward crane feel only, golden hour. Eight seconds. Distant bell and birds.

Bracket variables to reuse every week

Build a swap sheet once. Reuse it across genres.

[product noun stack]: material, color, logo rule, surface.

[face nouns]: hair, wardrobe, age band, expression beat.

[place]: room, weather, time of day.

[camera move]: slow dolly in only, locked tripod, lateral truck left, gentle pan right.

[seconds]: match UI slider to written beats.

Example swap: Starter 1 with [matte white bottle] becomes [brushed steel flask, black cap] for an outdoor brand. Camera and condensation beat stay frozen until the short test passes.

Step-by-step: from library line to host paste

Step 1: Pick genre and starter closest to your proof statement: material finish, face identity, place mood, or single action.

Step 2: Replace brackets. Read aloud. Confirm one subject, one camera path, timed beats inside target seconds.

Step 3: Choose text-to-video or image-to-video. Add preserve line on uploads. Crop still to aspect.

Step 4: Render host dialect if needed. Kling AI: label subject, motion, camera. Runway Gen-4: lead with shot size. Sora and Veo: merge into dense sentences, duration in UI. Luma: keep natural prose short.

Step 5: Generate four-to-five-second test. Score subject readable, action finished, camera followed one path. Fix one failing axis.

Step 6: Log paste next to genre tag. Promote duration only after short test holds.

When your brief is still messy paragraphs, open https://promptmake.net/video. Demand subject, timed beats, one camera path, scene, light, duration intent, and audio or silence. Guest quota is about three runs per day. Free registration raises video path quota separately from text and image.

Common mistakes with ai video prompt libraries

Mistake 1: Pasting starters verbatim without bracket swaps. Generic bottle clips teach nothing about your SKU.

Mistake 2: Stacking two starters in one box for a montage. Split across generations and edit.

Mistake 3: Changing camera verbs and subject nouns in the same retry. Isolate one axis.

Mistake 4: Ten UI seconds with one steam curl or one nod. Empty tail frames waste credits.

Mistake 5: Image caption language only: beautiful, cinematic, 8K. Add timed action or accept chaos.

Mistake 6: Skipping preserve language on image-to-video. Upload drift fights your face or label.

Mistake 7: Treating library lines as host-specific law. Retarget dialect, not story facts.

Mistake 8: No notes log. You relearn the same five-second win every Monday.

Mistake 9: Expecting PromptMake to output MP4 files. It writes motion prompt text only.

When PromptMake /video fits

PromptMake does not render MP4 files. It drafts motion briefs you paste into Kling AI, Runway Gen-4, Sora, Veo, or Luma Dream Machine. Use https://promptmake.net/video when you have a genre in mind but nouns and beats are still fuzzy. Paste a starter from this library into the tool as a seed and ask for the same fields with your product or face nouns filled.

Typical loop: rough idea, PromptMake video draft, bracket swap, short host test, log winner. Soft sell only. The tool proposes. You approve nouns and motion scope.

Spend free runs on subject lock and a four-second test, not on synonym hunts for moves you have not timed yet. After two clean clips on the same shell, save your private library row beside these public starters.

FAQ

What are ai video prompts?

Ai video prompts are text briefs that tell text-to-video and image-to-video models what to render in a short clip. Strong lines name subject, timed action, one camera path, scene, light, duration intent, and optional audio. The model invents pixels. The prompt decides what must happen in a few seconds.

How is this library different from ai video prompt structure?

Ai video prompt structure teaches the SMCD pillar framework: subject, motion, camera, duration. This library gives twenty genre starters you paste and swap. Read structure to learn why fields exist. Use this page when you need copy-paste lines today.

Can I use these starters on Kling, Runway, Sora, Veo, and Luma?

Yes. Starters are host-portable when you keep one subject, one camera path, and beats matched to seconds. Retarget grammar lightly at paste time. Confirm live model names in each UI before you automate.

Should I start from text or a reference still?

Start from a still when identity, packaging, or a face must match approved pixels. Upload carries subject. Text carries action and camera. Start from text when you explore place and motion from a blank frame.

How long should my first test clip be?

Four to five seconds with one action beat and one camera move is the default short test across Kling AI, Runway Gen-4, Sora, Veo, and Luma Dream Machine as of mid-2026. Promote length only after subject and motion hold.

Does PromptMake generate video files?

No. PromptMake at https://promptmake.net/video generates motion prompt text only. You paste into your chosen video host and spend credits there.

How many starters are in this library?

Exactly twenty copy-paste starters organized by genre: four product, four portrait, four nature, four action, and four food, fashion, tech, and travel lines. Swap bracket nouns and keep motion grammar.