By MSN Karthik, Founder, Writer & Director — Adwhyta AI Studio
I've generated over an hour of AI footage for a single 8-minute short film. Most of it went in the bin. Not because the video models were bad — because early on, my prompts were. Somewhere between clip 200 and clip 300 of making Unmasked, I stopped writing prompts like a person asking a genie for a wish and started writing prompts like a director describing a shot to a cinematographer. That one shift changed everything.
Every text-to-video tool promises to "bring your ideas to life", then hands you a blank box. This guide is what goes in the box. You'll get the Cinematic Prompt Framework we use at the studio (six blocks, always in the same order), plus 50 copy-paste prompts for AI video generation organized by genre, the camera and lighting language that video models actually understand, the technical modifiers most guides skip, and the iteration workflow that separates cinematic results from AI slop. It's written for filmmakers and professional creators, but the prompting techniques work for anyone who wants to create a video that doesn't look ai-generated.
Key Takeaways
- AI video prompts work best as shot descriptions, not keyword lists. Write what a camera would see.
- The Cinematic Prompt Framework has 6 blocks: Subject, Action, Setting, Camera, Lighting, Style — in that order.
- One subject, one action, one camera move per prompt. Stacking camera moves is the #1 cause of chaotic output.
- Different models read prompts differently: some reward paragraph-style detail, others want direct action keywords. Treat these as dialects, not different languages.
- Expect to keep 5–10% of what you generate. Iteration — changing one block at a time — is where the craft lives.
Why Filmmaker Prompts Beat Keyword Soup in AI Video Generation
How do you write a good AI video prompt? Describe a single shot the way a director would brief a cinematographer: name the subject, give it one clear action, place it in a specific setting, direct the camera and lens, define the light, and lock the visual style. Specific beats long. One shot per prompt.
Most prompts I see from beginners look like this: "cinematic, epic, 8K, masterpiece, beautiful woman walking, city, dramatic lighting, trending, hyperrealistic." That's not a prompt. That's a mood board that fell down the stairs.
The model has no idea what to prioritize, so it averages everything — and averages are exactly what AI slop looks like. Generic shot type, a random camera angle, contradictory lighting conditions, a visual style that changes mid-clip, a camera that drifts because nobody told it what to do. The AI handles whatever you leave unspecified, and it handles it blandly.
Here's the thing filmmakers already have that prompt hobbyists don't: a hundred years of shared vocabulary. Shot sizes, lens types, camera movement, color grade. "Slow dolly-in, 35mm lens, shallow depth of field, motivated window light" means something precise on a film set — and it turns out AI video generation models were trained on footage described in exactly this language. When you prompt in camera grammar, you're speaking the model's native tongue.
When we made Unmasked, our short about a Theyyam dancer that went on to win awards at festivals from Bali to Indywood, I generated roughly 1 to 1.5 hours of footage for a film that runs 8 minutes. Only 5–10% of generations were usable. The single biggest factor in whether a clip survived? Whether the prompt read like a shot description or a wish.
As I've said before: AI filmmaking is not just prompt engineering. It's very much like an animation film process. It's very tedious. But structured writing prompts is what makes the tedium converge on a film instead of a folder of pretty accidents.
For where prompting sits in the full video creation process, see our AI film production pipeline breakdown.
The Cinematic Prompt Framework: 6 Blocks of Every Prompt
Is there a formula for writing cinematic AI prompts? Yes: [Subject] + [Action] + [Setting] + [Camera] + [Lighting] + [Style]. Fill each block with one specific choice, keep the order fixed, and you get repeatable, directable output instead of random results.
The 6-Step Cinematic AI Video Prompting Framework
- Subject: Define the main focus on screen (e.g., a retro sports car).
- Action: Specify one clear movement (e.g., drifting around a coastal bend).
- Setting: Describe the location, weather, and time of day (e.g., wet asphalt, Amalfi coast, dusk).
- Camera: Direct the framing and lens details (e.g., extreme low angle, slow tracking shot).
- Lighting: Set the mood (e.g., warm sunset light, long dramatic shadows).
- Style: Define the color grading and texture (e.g., cinematic 35mm film grain, anamorphic).
Assembled, that example reads:
[Copy Prompt Example]
A retro sports car drifting around a coastal bend, wet asphalt on the
Amalfi coast at dusk. Extreme low angle, slow tracking shot. Warm sunset
light with long dramatic shadows. Cinematic 35mm film grain, anamorphic.
Read it aloud. It sounds like a shot list entry — because that's what it is.
A few rules that make the framework actually work:
- One block, one decision. Two actions in the Action block ("running and then jumping into a car") is how you get morphing limbs. Split it into two shots.
- The order matters. Models weight early tokens more heavily. Subject first means the model builds the frame — and the cinematic composition around it — from what you care about most.
- Foreground is a Setting decision. If you want depth, say what sits in the foreground ("rain-streaked glass in the foreground, subject beyond it"). Layered frames create a sense of space that flat prompts never get.
This is the exact structure behind every prompt in the 50 examples below. Learn the six blocks and you can generate the other 950 yourself.
Camera Grammar: Prompting Camera Movement, Angles, Lenses & Coverage
The Camera block is where filmmakers have the biggest unfair advantage. Models respond to real cinematography language with surprising fidelity — but only if you're specific about shot type, camera angle, and lens.
Shot sizes that models understand well:
| Term | What you get |
|---|---|
| Wide establishing shot | Full environment, small subject — great for openings |
| Wide shot | Subject full-body in context |
| Medium shot | Waist-up, the conversation workhorse |
| Medium close-up | Chest-up, more intimacy |
| Close-up | Face fills frame, emotion carrier |
| Extreme close-up | Eyes, hands, textures — detail inserts |
Camera moves that reliably translate:
- Slow dolly-in / push-in — builds tension toward the subject
- Dolly-out — reveals context, often an emotional release
- Tracking shot (or tracking shot following the subject) — lateral or follow movement
- Slow pan left/right — surveying a space
- Crane shot / aerial rise — scale and grandeur
- Handheld camera — documentary energy, nervous intimacy
- Static shot, locked-off — stillness is a choice too; say it explicitly
- Dolly zoom — the Vertigo effect; use sparingly, models overcook it
The single most important rule: one camera move per shot. "Crane down into a dolly-in that becomes a pan" produces chaos. Real cinematographers combine moves with rehearsal and a grip team; models combine them with prayer.
Lens types are the difference between video and cinema. "35mm lens" gives natural perspective. "Anamorphic lens" brings oval bokeh and horizontal flares. "Shallow depth of field" melts the background; add "bokeh" if you want the melt to sparkle. "Deep focus" keeps foreground and background sharp for staged compositions.
We go much deeper on directing generated camera work and light in AI cinematography — this section is the prompt-facing summary.
Lighting & Grade Language That AI Video Models Understand
Lighting is where prompts most often go vague — "dramatic lighting" is the "add salt to taste" of AI video. Models do dramatically better with motivated light: light that has a named source and direction.
| Vague (avoid) | Precise (use) |
|---|---|
| dramatic lighting | single key light from a window camera-left, shadows falling right |
| moody | low-key lighting, deep shadows, small pools of warm practical light |
| beautiful light | golden hour backlight, rim of warm light on the subject's hair |
| atmospheric | volumetric light through haze, visible light shafts |
| natural lighting | soft overcast daylight, no hard shadows |
A working vocabulary for the Lighting block: key light (main source and its direction), backlight (separation and rim), golden hour (warm, low, flattering), ambient/overcast (soft and even), volumetric (visible beams through haze or dust), practical sources (neon signs, lamps, firelight — name them and the model will place them in-world).
The Style block then locks the color grade and texture: film grain (organic texture, kills the plastic look), 35mm film (halation and softness), color palette (name two or three hues — "teal shadows, amber highlights" or "desaturated except for one red coat"), film noir (high contrast, hard shadows, venetian-blind patterns), documentary style (handheld realism, muted grade). A style reference does the work of an Instagram filter, but in-camera: it changes how everything is rendered, not just how it's tinted.
One studio habit worth stealing: pick one palette per project and repeat it in every prompt. Cinematic color is a decision you make once and defend everywhere. Cross-shot consistency starts in the Style block: the color grade you describe is the grade you refine in post.
Negative Prompts, Aspect Ratio & Technical Modifiers
Beyond the six blocks sits a small technical layer that most prompt guides skip — and that quietly decides whether a clip is usable.
- Negative prompts. Many tools accept a list of things you don't want: "no on-screen text, no warping, no extra fingers, no morphing." Use negative prompts to suppress recurring artifacts, not to art-direct. If you're negating style ("not ugly, not boring"), the positive prompt is what needs fixing.
- Aspect ratio. Declare it — 16:9 for cinematic and YouTube work, 9:16 for vertical, 2.39:1 where supported for a widescreen feel. Composition changes with the frame, so a prompt tuned for 16:9 often needs its Camera block rethought for vertical.
- Speed and time. "Slow motion" is the most reliable temporal modifier; "timelapse" the second. A deliberate pause reads better when you write it as action: "she stops, holds her breath, then turns."
- Motion blur. Ask for it when you want kinetic energy ("controlled motion blur on the background"), and negate it when you need crisp product detail.
- Transformations. Morphs and transitions ("the city dissolves into a forest") are their own shot type — give the transform the whole prompt rather than burying it inside a scene. And remember that a montage is an edit decision: generate its shots individually and cut them together, don't ask one generation to be five scenes.
None of this replaces the framework. It's the fine print under it.
50 Copy-Paste Cinematic AI Video Prompts by Genre
Every prompt below follows the 6-block framework. Copy, swap the blocks for your project, start creating. They're written tool-agnostic; dialect adjustments come in the next section.
Commercial & Product (1–5)
[Copy Prompt Example 1]
A matte-black wireless earbud rotating slowly on a glass pedestal, minimalist
studio void. Extreme close-up, slow orbital camera move. Single hard key light
from above with a soft fill. High-contrast product grade, subtle film grain.
[Copy Prompt Example 2]
A perfume bottle standing in shallow water, ripples spreading outward, infinite
white cyclorama. Low angle, slow push-in. Backlight refracting through the glass,
caustic light patterns. Clean editorial style, pastel color palette.
[Copy Prompt Example 3]
A running shoe landing in slow motion on wet pavement, water droplets exploding
outward, urban street at night. Ground-level static shot, shallow depth of field.
Neon-lit reflections in cold blue with one amber accent. Gritty cinematic grade.
[Copy Prompt Example 4]
A smartwatch on a runner's wrist as she checks her pace, coastal cliff trail at
dawn. Medium close-up, handheld tracking shot. Golden hour backlight, lens flare
kissing the frame edge. Warm athletic-brand grade, crisp detail.
[Copy Prompt Example 5]
A ceramic coffee mug being set down on a wooden desk beside a laptop, cozy home
office. Locked-off overhead shot. Soft window light from the left, gentle
shadows. Warm lifestyle grade, 35mm feel.
Automotive (6–10)
[Copy Prompt Example 6]
A vintage silver roadster crossing a desert salt flat, heat shimmer rising,
midday. Wide establishing shot from a distant static camera. Harsh overhead
sunlight, mirage distortion at the horizon. Sun-bleached grade, fine film grain.
[Copy Prompt Example 7]
An electric SUV driving through a rain-slick city intersection at night.
Low-angle tracking shot alongside the front wheel. Neon signage reflecting off
wet asphalt, cold blue palette with magenta accents. Glossy commercial grade.
[Copy Prompt Example 8]
A motorcycle leaning through a mountain switchback, pine forest, early morning
fog. Aerial crane shot descending toward the road. Volumetric light shafts
through the trees. Desaturated alpine palette, cinematic contrast.
[Copy Prompt Example 9]
A classic muscle car's headlights snapping on in a dark garage, dust motes in
the air. Extreme close-up on the headlight, static shot. Single practical
source — the headlight itself — flaring into the lens. Film noir contrast.
[Copy Prompt Example 10]
A convertible gliding down a palm-lined boulevard, golden hour. Rear tracking
shot following the car, 35mm lens. Long shadows stretching across the road,
warm backlight. Nostalgic Kodak-style color palette, soft halation.
Fashion & Beauty (11–15)
[Copy Prompt Example 11]
A model in a flowing red silk dress turning once, fabric billowing, brutalist
concrete stairwell. Full-body wide shot, slow-motion. Single shaft of hard
sunlight from a high window. Editorial fashion grade, deep shadows.
[Copy Prompt Example 12]
A close-up of eyes as gold eyeshadow catches the light, model slowly opening
her eyes, dark studio. Extreme close-up, locked-off macro framing. Warm key
light raking across the face. Luxe beauty grade, velvety blacks.
[Copy Prompt Example 13]
A man in a tailored charcoal suit walking toward camera through a marble hotel
lobby. Medium shot, slow dolly-out keeping pace. Soft ambient light with warm
practical lamps in background bokeh. Understated luxury grade, anamorphic.
[Copy Prompt Example 14]
A model's hand trailing through a rack of pastel garments, boutique interior,
afternoon. Close-up tracking shot following the hand. Soft window light,
gentle fill. Airy high-key grade, creamy shallow depth of field.
[Copy Prompt Example 15]
A dancer in an iridescent jacket vogueing under a single spotlight, black void
stage. Low-angle medium shot, static. Hard top light with haze, volumetric
cone. High-fashion contrast, specular highlights, fine grain.
Food & Beverage (16–20)
[Copy Prompt Example 16]
Honey drizzling onto a stack of pancakes, steam rising, rustic kitchen table.
Extreme close-up, slow-motion, macro lens. Warm morning window backlight
making the steam glow. Rich appetizing grade, golden palette.
[Copy Prompt Example 17]
A chef's hands tossing vegetables in a flaming wok, commercial kitchen at
night. Medium close-up, slight handheld energy. The wok flame as the key
light, orange glow on steel surfaces. High-contrast culinary documentary style.
[Copy Prompt Example 18]
An espresso shot pouring into a glass cup, crema swirling, minimalist café
counter. Side-on extreme close-up, locked-off. Soft daylight from a storefront
window. Warm brown palette, clean Scandinavian grade.
[Copy Prompt Example 19]
A chilled cocktail glass with condensation, a citrus twist dropping in and
splashing, dark bar. Close-up, slow-motion at the moment of the splash.
Single amber practical light from behind the bottle shelf. Moody premium grade.
[Copy Prompt Example 20]
A baker pulling a tray of croissants from an oven, flour dust in the air,
dawn light in a small bakery. Medium shot, slow push-in. Golden hour light
through a side window, visible dust motes. Warm nostalgic film look.
Documentary & Travel (21–25)
[Copy Prompt Example 21]
A tea picker's hands moving across bright green bushes, misty hill plantation,
early morning. Close-up, handheld camera, documentary style. Soft overcast
light, mist muting the background. Naturalistic grade, gentle grain.
[Copy Prompt Example 22]
A fishing boat cutting through calm water toward a rising sun, wide open sea.
Wide establishing shot from a drone hovering low over the water. Golden hour
backlight, sun path glittering on the surface. Warm documentary palette.
[Copy Prompt Example 23]
An elderly potter shaping clay on a spinning wheel, sunlit workshop, dust in
the air. Medium close-up on the hands, static shot. Single window key light
from the right, volumetric haze. Earthy palette, 35mm film texture.
[Copy Prompt Example 24]
A crowded night market street with steam rising from food stalls, vendors
calling out, tropical city. Slow tracking shot moving through the crowd at
shoulder height. Mixed practical lighting — lanterns, neon, bare bulbs.
Vibrant documentary grade, handheld realism.
[Copy Prompt Example 25]
A mountaineer silhouetted on a ridge line as clouds pour over the peak behind
her. Wide shot, locked-off, long lens compression. Backlit by low sun, rim
light on the rope and jacket. Cold blue-and-white palette, epic stillness.
Sci-Fi (26–30)
[Copy Prompt Example 26]
A lone astronaut walking across a dust plain toward a derelict ship, twin moons
overhead. Wide establishing shot, slow pan right following the walk. Cold rim
light from the moons, deep shadows. Desaturated palette, anamorphic flares.
[Copy Prompt Example 27]
A holographic interface blooming out of a technician's palm, dark control room.
Medium close-up, slow push-in. The hologram as sole light source, cyan glow on
the face. Clean sci-fi grade, subtle scanline texture.
[Copy Prompt Example 28]
A futuristic city skyline at night with layered air traffic, rain falling,
towers receding into haze with the city skyline behind them. Aerial wide shot,
slow forward drift between towers. Neon-lit facades reflecting in wet glass,
cold blue palette with magenta accents. Volumetric rain haze, dystopian grade.
[Copy Prompt Example 29]
A cryopod hissing open, frost vapor spilling out, silhouetted figure inside
stirring. Low-angle medium shot, static. Single cold key light from inside the
pod, backlit vapor. High-contrast sci-fi noir, deep blacks.
[Copy Prompt Example 30]
A robot's hand and a human hand almost touching, sterile white lab. Extreme
close-up, shallow depth of field, locked-off. Soft top light, no shadows.
Clinical white palette with a single warm skin tone accent. Minimalist grade.
Noir & Thriller (31–35)
[Copy Prompt Example 31]
A detective in a rain-soaked trench coat under a flickering streetlamp,
midnight alley. Wide shot, slow push-in. Single overhead practical light,
hard shadows, rain streaks catching the glow. Film noir, high-contrast
black-and-white.
[Copy Prompt Example 32]
A woman's face half-lit through venetian blinds as she listens at a door,
dim office. Close-up, static shot. Slatted window light raking across her
face, deep shadows. Classic noir grade, fine grain.
[Copy Prompt Example 33]
A briefcase exchange between two silhouetted figures on an empty train
platform, fog rolling in. Wide shot, long lens, slight handheld tension.
Cold sodium-vapor platform lights, silhouettes against the haze. Muted
thriller palette.
[Copy Prompt Example 34]
A rotary phone ringing on a desk, cigarette smoke curling through a shaft of
light, nobody in frame. Medium close-up, slow dolly-in toward the phone.
Single hard side light, volumetric smoke. Noir grade, heavy blacks.
[Copy Prompt Example 35]
A car's rearview mirror reflecting headlights approaching fast on a night
highway. Extreme close-up on the mirror, handheld vibration. The pursuing
headlights as the only light source, flaring. Cold desaturated grade, motion
blur on the background.
Drama & Emotion (36–40)
[Copy Prompt Example 36]
An old man at a kitchen table holding a folded letter, deciding whether to
open it, morning. Medium shot, very slow push-in. Soft window light from the
left, dust in the air. Muted warm palette, gentle 35mm grain.
[Copy Prompt Example 37]
A daughter resting her head on her mother's shoulder on a hospital bench,
end of visiting hours. Wide shot from down the corridor, locked-off.
Fluorescent overheads with one warm doorway glow behind them. Restrained
desaturated grade, long stillness.
[Copy Prompt Example 38]
A dancer rehearsing alone in an empty studio, stopping mid-move to catch her
breath, evening. Medium shot in the mirror's reflection, static. Last golden
hour light through tall windows, long shadows on the floor. Wistful warm grade.
[Copy Prompt Example 39]
Rain sliding down a train window past a young man's unfocused reflection,
countryside blurring by. Close-up on the glass, locked-off. Overcast ambient
light, the reflection soft. Blue-grey palette, shallow depth of field.
[Copy Prompt Example 40]
Two hands slowly letting go of each other across a café table, unresolved.
Extreme close-up on the hands, static, shallow depth of field. Soft side
window light. Tender muted grade, slight film grain, slow motion.
Action (41–45)
[Copy Prompt Example 41]
A parkour runner vaulting a rooftop gap between concrete buildings, overcast
city. Tracking shot following from behind, wide lens. Flat overcast light,
gritty texture. Desaturated urban grade, controlled motion blur.
[Copy Prompt Example 42]
A boxer's glove hitting a heavy bag, sweat spraying in slow motion, basement
gym. Extreme close-up at impact, locked-off. Single hanging bulb key light,
hard shadows. Gritty high-contrast grade, visible grain.
[Copy Prompt Example 43]
A horse and rider galloping through shallow river water, spray exploding,
canyon at dusk. Low tracking shot skimming the water surface. Golden hour
backlight turning the spray into fire. Epic warm grade, slow motion.
[Copy Prompt Example 44]
A climber's hand slipping then catching a granite edge, chalk dust puffing,
big wall exposure below. Extreme close-up on the hand, slight handheld.
Hard midday sun from above, deep shadow below the ledge. High-tension
naturalistic grade.
[Copy Prompt Example 45]
A sprinter exploding out of starting blocks, stadium at night. Ground-level
side-on tracking shot, brief slow motion at launch. Stadium floodlights,
rim-lit muscles, lens flare. Broadcast-plus cinematic grade, crisp detail.
Animation & Stylized (46–50)
[Copy Prompt Example 46]
A paper boat sailing down a rain gutter past glowing shop windows, stylized
3D animation. Low-angle tracking shot at water level. Warm shop light
reflecting in the stream, cool rain ambience. Pixar-adjacent look, soft
global illumination.
[Copy Prompt Example 47]
A fox spirit leaping between floating lanterns in a night sky, hand-painted
anime style. Wide shot, slow upward crane following the leap. Lantern glow as
warm key sources against deep blue night. Painterly texture, visible
brushwork in the clouds.
[Copy Prompt Example 48]
A tiny robot watering a single flower on a rooftop garden, dawn, stop-motion
style. Medium shot, static, miniature depth of field. Soft pink morning light.
Handcrafted felt-and-metal texture, gentle color palette.
[Copy Prompt Example 49]
A watercolor city dissolving into rain and reforming as a forest, one
continuous transform. Wide shot, locked-off. Diffuse even light, no hard
shadows. Watercolor bleed transitions, paper texture visible throughout.
[Copy Prompt Example 50]
A jazz trio of animated animals playing in a smoky underground club, 1950s
cartoon style. Medium shot, gentle sway in the camera like a patron's POV.
Single stage spotlight with haze, warm amber. Retro cel-animation grade,
film grain and gate weave.
Want these as a one-page reference? The prompt templates cheat sheet at the end of this article packs the framework and vocabulary into a printable page.
Model Dialects: How Veo, Sora, Kling, Runway and Seedance Read Prompts
Here's the uncomfortable truth about every "ultimate prompt guide" online, ours included: model behavior is a moving target. Workflows we built in 2024 didn't survive into 2025, and our 2025 workflows aren't what we run today. So treat this section as what it is (one studio's working notes as of mid-2026) and check each tool's current documentation before a paid render run.
That said, the dialects cluster into recognizable families:
- Paragraph-style models (Veo, Sora): These reward full, flowing shot descriptions — the 6-block framework written as connected prose. Rich scene context, sensory detail, and explicit camera language all pay off. Our examples above are closest to this dialect out of the box.
- Direct-action models (Kling, Hailuo): These respond best to clear, compressed action statements. Front-load the subject and action, keep the sentence structure simple, and don't bury the movement under adjectives. Same six blocks, tighter grammar.
- Reference-driven workflows (Seedance with omni reference; Runway's reference tools): Seedance has been our workhorse for the last six months or so, and it changes what the text prompt is for. When character, style, and setting come from reference images, the prompt's job narrows to action, camera, and timing. Less describing, more directing. Nano Banana did something similar for us on the image side — it was the inflection point that ended our tedious fine-tuning era for character consistency.
- Animation-leaning prompts: Whatever the model, stylized output improves when the Style block leads with the medium ("hand-painted anime style", "stop-motion") before any content description.
The practical workflow: write the prompt once in the 6-block framework, then translate into the target model's dialect. The blocks don't change; the grammar does.
We keep our tool-by-tool testing in the reviews cluster — start with the best AI video tools rankings or the head-to-head Veo vs Sora vs Kling vs Runway shootout, with deep dives on Veo, Sora, and Kling.
Iteration Workflow: From First Prompt to Final Shot
Nobody nails a shot in a single prompt. On Unmasked, the climax effect (maybe 15 to 20 seconds on screen) took somewhere between 150 and 200 generated clips. That ratio sounds insane until you internalize it, and then it just becomes the job, the way an animator accepts 24 drawings per second.
The iteration method that keeps this from becoming random slot-machine pulling:
- Generate a control shot. Run your framework prompt as written. This is your baseline.
- Change one block at a time. Output too flat? Touch only the Lighting block. Camera drifting? Only the Camera block. If you change three blocks between generations, you learn nothing from the result.
- Lock blocks as they work. Once the Setting and Style read right, freeze that language verbatim and iterate only on Action. Your prompt becomes a shot spec with one variable.
- Know when to stop prompting. Some flaws are cheaper to fix in post than in the prompt — a minor color shift, a stray artifact, pacing. We never ship raw generations anyway; every clip goes through cleanup, grade, and edit. Prompting gets you 90% of a shot. Craft closes the rest.
- Keep a prompt log. Every kept clip gets its exact prompt saved next to it. Six weeks later, when a client wants "that same look," the log is the look.
One more honesty checkpoint: prompt craft cannot rescue a shot that shouldn't exist. If a shot doesn't serve the story, no framework saves it. The real work still happens at the scripting table, which is why we treat AI script writing and storyboarding as the stages that decide whether prompting is even aimed at the right target. And once shots multiply, keeping faces and wardrobe stable across them is its own discipline; that's covered in our character consistency guide.
The Prompt Templates Cheat Sheet
The whole system on one page:
[Subject] + [Action] + [Setting] + [Camera] + [Lighting] + [Style] (+ negative prompts, aspect ratio, speed)
- One subject. One action. One camera move.
- Camera: shot size → movement → lens.
- Lighting: name the source, its direction, and what it touches.
- Style: grade + texture + palette, consistent across the whole project.
- Iterate one block at a time; log every kept prompt.
Full disclosure: we're an AI studio, so we're biased — but a prompt framework is a starting point, not a production pipeline. When brands and production houses come to us, they're not buying prompts; they're buying a custom multi-model stack, not a single prompt box, plus the direction, iteration discipline, and post craft wrapped around it. If that's the level you need, see our work or book a call. If you're here to learn the craft yourself: steal everything above. That's what it's for.
