PROMPT LIBRARY
Kling 4.0 prompts, each with the video it made.
Each prompt is quoted word for word from the model maker's guide or a creator's post, next to the output it produced. Open a card for the full prompt.
Kling 4.0 is Kuaishou's video model with native stereo audio. One generation runs 3 to 30 seconds, 10 by default, at 720p, 1080p or 4K, in 21:9, 16:9, 1:1 or 9:16. Its default mode, Omni Reference, takes up to 15 inputs per task: 10 images, 5 videos totalling 30 seconds, 7 elements and 3 voice profiles. There is also a Keyframes mode with up to 10 frames, and video editing on up to 5 clips. Dialogue works in nine languages, among them Chinese, English, Japanese, Korean and Spanish. The prompts below are the examples from Kling's 4.0 user guide and ones creators shared on X, each shown with its video.
Sources: Kling 4.0 user guide
Kling 4.0 vs Seedance 2.5 prompts: what we measured on 628 prompts →
Updated 2026-10-31 · 37 prompts, newest first
Other model pages: GPT Image 2.6 · Nano Banana 2.1 · Seedance 2.5 · MiniMax H3 · Wan 3.0
Fluffy cloud creature sneezes raindrops onto woman
Red Beetle drives three witches through rainy night
Hooded hero runs across battlefield toward monster army
Kia Seltos driving through a sunlit forest road
Crime boss confronts courier in dim cantina
Husband confesses job loss at kitchen table
Two men talk in a rainy cafe
Lone knight holds fortress gate against armored attackers
Dog brings ball to the wrong friend in park
Backpacker chased by jaguar into waterfall pool
Phone clip at brunch, sister steals a bite of pancake
Fake gameplay of a 5v5 tactical shooter
A perfume bottle drives a loop ramp and blasts off
A blacksmith turns iron blocks into a greatsword, first frame to last
Remake a car spot with your car and your backdrop
Felt and clay forest class: a cloud teacher counts to four
Clay astronaut dog chases a bus-sized cheese mouse, from a storyboard
Dragon launches from a peak and burns a coastal city, six shots in 30s
Edit a clip: the corgi in the snow becomes a blocky voxel corgi
Edit a clip: turn the whole video into felt
Edit a clip: put a sleeping pig in the box and change the last line
Edit a clip: recast three people walking out of a shop
Edit a clip: remove the person in front
Edit a clip: replace the subject with one from an image
Edit a clip: swap the whole world for an image
Borrow a fight's choreography, swap in your own fighters
A girl plays a recorder in her room, six keyframes
An interpreter between a Spanish tourist and a Chinese clerk
A 17th-century militia marches out and freezes into the painting
A lotus as big as the moon, and a woman in white watching it
Recreate a cliff jump stunt, but on an office chair
Skater clears a stair set in one low tracking shot
Seven-shot ski edit from one character sheet
Spy car chase, tear-away suit, straight into a banquet
Twenty-second side-tracking escape down a busy street
Embarrassed vlog from the car after waving at a stranger
Nothing matches. Try a shorter word.
Kling 4.0 prompts worth starting from
The most liked prompt in each category on this page, in full, with what makes it work. Copy one, then change the subject.
Phone clip at brunch, sister steals a bite of pancake
A 30-second single take that reads as an unedited phone clip: auto-exposure drifting, a waiter bumping the table at 9s, off-camera dialogue and café noise.
Raw vertical 9:16 video recorded on a phone's front camera (not on screen), one continuous 30-second take with no cuts. It looks like an ordinary unedited phone clip, not a film: slightly soft image, a little digital noise and compression, deep focus with the café behind her as sharp as her face, auto-exposure brightening and dimming as people pass the sunny window behind her, no colour grading. The phone sits close and a little low, looking slightly up at her; around 9 s a waiter's arm bumps the table, the phone slides and tilts, and she catches it and props it back up. Sadie, a young woman in her early 20s with long honey-blonde waves, light natural makeup, small gold hoop earrings and a cropped cream knit top, sits at a small marble table by a big sunny window in a busy brunch café. Her sister sits across from her, out of frame except for her arm and a hand with a silver ring. Behind Sadie: other tables of people talking, a counter with a glass pastry case, plants, staff moving about, chalkboards too far away to read. Around 9 s a waiter sets down two plates, a lemon ricotta pancakes and a plate of eggs on toast, plus two iced coffees in plain glasses. Around 15 s she turns the pancake plate toward the camera; around 20 s she cuts a bite, eats it and closes her eyes; around 26 s her sister's hand reaches in with a fork and steals a piece of pancake, and Sadie swats at it, laughing. [Sadie: casual young woman's voice, warm and a bit giggly, recorded on the phone's built-in mic from half a metre away]: "Okay, I promised I'd show you what we get. ... We get the same thing every single time. ... Oh, here it is! Thank you! ... Look at this. Lemon ricotta pancakes. ... Mmm. Okay. Yeah. Still the best. ... Hey! Get your own!" [Her sister, off camera: a young woman's voice, dry and amused]: "That's the tax." [Sadie, laughing]: "She does this every time." Audio as captured by the phone mic in a busy café, not studio clean: her voice close but with the room around it, the hiss and grind of an espresso machine, cups and cutlery clinking, overlapping chatter and laughter from other tables, a door bell jingling, a chair scraping, faint music in the background. No subtitles, no captions, no on-screen text. Bathroom fight against six thugs with a water weapon
Two image references lock the hero identity and the enemy look, while a timed beat sheet in 5 second blocks sets the order of the choreography. The continuity section lists what must stay fixed and states that the water weapon stays liquid.
Ultra-photorealistic Hollywood dark-fantasy action, ARRI Alexa 65 aesthetic.
HERO: Use
@image1
as exact female identity reference. Preserve face, hair and physique. Secure white towel throughout.
ENEMIES: Exactly six aggressive katana-wielding thugs based on
@image2
.
LOCATION: Luxury marble bathroom, glass shower and bathtub filled with water and rose petals.
ACTION:
0–5s: Six thugs storm in simultaneously. Hero dodges multiple sword attacks, ducks and pivots beside the bathtub. Japanese battle shouts.
5–10s: Hero performs a backbend, spinning kick and elbow strike. Bathtub water rises and follows her movements.
10–15s: Two enemies attack together. Hero briefly dissolves into shadow and reappears nearby. Their swords miss and they collide. Hero counterattacks.
15–20s: Water spirals around Hero's arm, forming a flowing liquid Indonesian golok silhouette. Never solid. She repels one attacker and sends another through the glass shower. Realistic splashes and shattered glass.
20–25s: Remaining thugs attack together. Hero performs a wall-assisted jump, three-hit aerial combination and throws an attacker into bathroom fixtures.
25–30s: An injured thug activates a red communication device: 「標的を確認した。」
Indonesian subtitle: Target terkonfirmasi.
Hero freezes airborne water and petals, turns toward him and says: 「相手を間違えたな。」
Indonesian subtitle: Kalian salah memilih lawan.
Heavy footsteps approach. Hero turns toward the entrance. End mid-action.
VISUAL & AUDIO: Dynamic tracking, continuous realistic combat, expressive reactions, synchronized Japanese dialogue, sword impacts, glass breaking and water sounds.
CONTINUITY: One Hero, exactly six enemies. Preserve identity, towel, weapons and bathroom layout. Bathtub water decreases during manipulation. Hydro Cleaver remains liquid. No extra characters, duplicated bodies, solid water weapons, robotic movements, pauses, victory pose or fade-out. Fluffy cloud creature sneezes raindrops onto woman
Timed scene blocks with one action per beat, plus an explicit line to keep the same creature and woman across all scenes, hold the character steady over 30 seconds. Cues like soft volumetric sunlight and realistic fluffy cloud texture anchor the look, and the no text, no logos line keeps the frame clean.
How am I supposed to move on after seeing this AI creature
Made with Kling 4.0
Prompt : Create a 30-second ultra-realistic cinematic fantasy video featuring a completely original miniature cloud creature in a charming countryside bedroom.
Scene 1 (0–7 seconds): A tiny floating cloud creature with fluffy white fur, pale blue eyes, tiny dangling feet, and miniature translucent wings drifts above a wooden bedside table. Soft morning sunlight enters through linen curtains while dust particles shimmer in the air.
Scene 2 (7–14 seconds): A young woman wearing a peach-colored cardigan and a cream skirt enters the room. The little cloud follows her curiously, hovering above her head. Suddenly, it sneezes and releases a tiny shower of sparkling raindrops onto her hair.
Scene 3 (14–23 seconds): The woman looks surprised before smiling and holding out her hands. The cloud becomes excited and accidentally creates a miniature rainbow inside the room. Tiny floating droplets turn into glowing bubbles that gently drift around the furniture.
Scene 4 (23–30 seconds): The cloud settles onto the woman's head like a fluffy hat. It produces one final tiny raindrop that lands on its own nose. The creature crosses its eyes trying to look at the drop, then giggles silently as a miniature rainbow appears above it.
Visual style: Original imaginary creature, realistic fluffy cloud texture, expressive blue eyes, physically believable water droplets, soft volumetric sunlight, realistic reflections, pastel color palette, detailed fabric textures, cinematic close-ups, natural character animation, smooth camera movements.
Duration: 30 seconds. Aspect ratio: 16:9. 4K cinematic quality. Preserve the same creature and woman across all scenes. No text, subtitles, logos, or watermark. Use original designs only. Red Beetle drives three witches through rainy night
Eleven numbered shots each pair a visual beat with an audio cue, and three reference images lock each witch's look across cuts. The style line names cel shading, ink outlines and film grain so the whole clip stays in one retro anime look.
A 30-second, 16:9 animated short in retro 80s/90s hand-drawn cel anime style, with bold black ink outlines, flat cel shading, soft gradient skies and subtle film grain. The palette is twilight blues and purples against warm amber window glow, with red and orange accents. The setting is a rain-slick old-town street at night with glowing street lamps. The mood is dark-comic and mischievous. Camera moves are smooth and cinematic, and the cuts are fast.
CHARACTERS (keep consistent in every shot, using the reference images)
Blonde witch (Image 3): platinum-blonde hair with bangs, blue eyes, round tinted glasses, a cream and gold satin dress. Her voice is cool, breathy and sarcastic.
Redhead witch (Image 1): wavy copper hair, freckles, green eyes, a green fur-trimmed coat. Her voice is low and scheming, with a wicked laugh.
Black-haired witch, the driver (Image 2): long black hair with blunt bangs, violet eyes, red lips, a silver choker, a black lace outfit. Her voice is husky and playful.
AUDIO
Music is a jazzy, mischievous retro score with upright bass, muted trumpet and vibraphone. It turns into a tense chase with drums and strings, then a soaring magical orchestral theme in the flight, and settles into a warm acoustic guitar and music box melody at the cabin. Ambience includes light rain, tires on wet cobblestone, crackling torches, wind and crickets. Dialogue is short and natural, with no narrator.
SHOT 1: The ride. Wide side view of a red vintage Beetle gliding down a wet night street. Cut inside to the three witches: the blonde adjusts her glasses, the redhead sits in the middle, the driver grips the wheel.
Audio: jazz groove fades in, rain and tire hiss. The blonde sighs, “Such a dull evening.”
SHOT 2: The spell. Through the windshield, a couple kisses under a lamppost. Close-up of the redhead’s envious glare as she flicks her fingers and pink magic sparkles swirl.
Audio: the music drops to a low string drone, then a harp glissando. She mutters, “Hmph... how sweet.”
SHOT 3: The pig prank. The woman’s head turns into a pig’s head and she screams while the man recoils. Cut back to the car, where all three witches burst into laughter.
Audio: a cartoon “pop,” a squeal, then a comic trombone slide and their wicked cackling. The man gasps, “What?!”
SHOT 4: The old lady. The Beetle stops at a crosswalk as an old woman with a basket crosses. Close-up of her smiling and tipping her scarf, then the driver leans out the window and sticks her tongue out.
Audio: a warm clarinet phrase and a cheeky slide whistle. The old woman says, “Bless you, girls!”
SHOT 5: The queue. Wide shot of the car parked outside a glowing grocery shop with a long line of people. The three witches strut past the crowd, heads high, and stand shoulder to shoulder in a low-angle shot.
Audio: a swaggering bassline with heel clicks and the crowd murmuring. The redhead says smoothly, “Excuse us.”
SHOT 6: The closed shop. Through the shop window, shelves of grocery bags glow warm while the crowd stares. Close-up of the door sign reading “CLOSED.” The witches then stand smiling with arms full of overflowing grocery bags.
Audio: a timpani hit as the sign swings, then a smug, triumphant jingle. The blonde purrs, “Fresh. Perfect.”
SHOT 7: The mob. An angry mob with torches and pitchforks charges down the street. The witches glance sideways, then run, the redhead wrapped in a glowing green magic aura.
Audio: the music cuts to a fast chase with war drums and stomping boots. The mob shouts, “Get them!” The redhead calls out, “Hold on to your bags!”
SHOT 8: Lift-off. The witches leap onto straw brooms in a burst of green sparks and rise above the shouting crowd.
Audio: a bright “whoomph,” a powerful whoosh, and the orchestra swelling into the flight theme.
SHOT 9: The flight. Aerial wide shot of the three witches flying over the glowing town toward purple mountains at dusk, ending on tiny silhouettes against the sky.
Audio: full orchestra with harp and choir pads, wind rushing, and the witches laughing with joy.
SHOT 10: The cabin. They land by a stone cabin and picnic table at golden sunset. The driver pours tea from a white teapot and the blonde sips with her eyes closed. Close-up of all three clinking their teacups together, and their eyes flash golden through their glasses.
Audio: the acoustic guitar theme enters with crickets, a pouring sound and a porcelain clink. They say together, “Cheers.”
SHOT 11: Final shot. Wide shot of the lone cabin on a hilltop as dusk turns into a starry night, then fade to black.
Audio: the music box melody slows and ends on a single chime as the sound fades out with the picture.
NEGATIVE PROMPT
Photorealism, 3D render, extra fingers, distorted faces, inconsistent hair or outfit colors, any non-English text or speech, watermark, blur, low quality, harsh audio, lip-sync errors, narrator voice-over. A perfume bottle drives a loop ramp and blasts off
A product spot played as a stunt show, timed in two-second beats, with an unlabeled bottle so no brand gets in the way.
Cinematic, photoreal night scene on a black stage inside a giant hangar. A towering concrete loop ramp/road with glowing red edge lights and a white center stripe dominates the background. In front of it stands a premium transparent glass perfume bottle, unlabeled, no logo, filled with pale golden fragrance, featuring elegant beveled edges, a polished metallic cap, realistic glass refraction, crisp reflections, and subtle condensation droplets. The bottle engages four small rocket nozzles under its base; soft white exhaust with bright orange cores appears. The bottle glides forward, then blasts up the glowing ramp, hugging the curve, leaving faint light trails and red neon reflections across its glass surface. At the apex it shoots off the ramp and transitions to a sleek electric-thruster flight, like a futuristic luxury flying car, then ends in a close hero shot of the perfume bottle hovering and roaring upward vertically with four bright jets, volumetric spotlights and fog below, sparks and vapor interaction, dramatic rim light, golden highlights, crisp reflections, shallow depth of field, 4K, high-end luxury perfume and car commercial look. No text, no brand, no label. Duration 9–10 seconds. Aspect ratio 9:16. Negative cues: no logos, no text, no labels, no hands, no spillage, no deformation, no duplicate bottles, no cartoon look, no melting glass, no detached cap, no liquid leakage, no extreme motion-blur smears. Beat sheet (timing guide) 0–2s: Low-angle push-in on the massive loop ramp; perfume bottle centered, engines arming beneath the base with subtle heat shimmer and glowing orange cores. 2–6s: Dynamic follow-cam as the bottle rockets up the ramp; speed ramping, neon edge lights streaking, dramatic red reflections sliding across the glass. 6–8s: Bottle leaves the ramp, switches to electric thrusters; quick cinematic orbit as it hovers midair, revealing the golden fragrance and metallic cap. 8–10s: Tight hero shot of the perfume bottle ascending vertically with four bright jets, swirling fog, volumetric spotlights, realistic vapor, and luxurious golden reflections. End with the bottle sharply focused and centered against the dark hangar. Crime boss confronts courier in dim cantina
A numbered shot list with a lens FOV and screen direction for each shot, plus an exact subtitle list, lets Kling hold one consistent scene across 11 cuts. Fixed headcount and seating rules, with every spoken line tied to a shot, keep the cast stable and the dialogue timing clear.
SCENE CONTEXT
EXACTLY 4 HUMAN CHARACTERS — NO DUPLICATES, NO EXTRAS: the boss, the courier and two silent bodyguards, all Mexican. Interior, the back room of a closed cantina in Guadalajara, Mexico, late at night. A courier nicknamed Flaco reports to his boss that the merchandise is gone: a rival family, the Mendozas, ambushed his team on the highway to Tepic and took the whole load. The boss starts completely calm while he eats; the tension rises with every line; he explodes, then drops to a cold, lethal whisper. Crime drama, real time. Only the boss and the courier speak.
CHARACTERS
ERNESTO (the boss): heavyset Mexican man, slicked-back grey hair, thick grey moustache, a thin old scar cutting through his left eyebrow, heavy-lidded eyes. Black button-up shirt with subtle dark embroidery on the yoke, sleeves rolled to the forearm, a small sauce stain on the cuff, a thick gold chain at the neck. A worn gold signet ring on his right hand. 178 cm.
FLACO (the courier): very thin, wiry Mexican man, short black hair neatly cut, thin moustache, a small fresh bruise on his left cheekbone, a faded tattoo of a small cross on the back of his right hand. Clean, pressed white short-sleeved shirt, dark jeans, worn brown leather boots, a gold wristwatch with a freshly cracked face. He looks put-together, but his nerves betray him. 182 cm.
THE BODYGUARDS: two broad, silent men in plain dark short-sleeved shirts, one with a shaved head and a thick black beard, the other with a grey buzz cut and a crooked nose. They stand against the back wall in the half-dark, arms crossed, never speaking.
LOCATION MAP
A small square table with a worn green oilcloth sits at the center under a single hanging lamp, with a slow ceiling fan above it. Ernesto sits on the right side of the table (screen-right), facing left, with a plate of carne asada, a stack of warm corn tortillas in a cloth, a small clay bowl of salsa, a fork and a full shot glass of tequila. A box of dominoes rests at the edge of the table. An empty wooden chair waits on the left side of the table (screen-left). The two bodyguards stand against the back wall behind the table, one at the left third and one at the right third of the frame, partly in shadow. In the dark background: stacked beer crates without labels, a closed swinging kitchen door, a wall of painted talavera tiles, an old calendar, closed window blinds. The eyeline axis runs left–right across the table; the camera always stays on the same side of it.
FORMAT
One scene, 11 shots joined by clean hard cuts, classic shot–reverse-shot coverage. Cuts follow the dialogue and also land on the listener's reaction. The framing gets tighter and the silences longer as the tension rises, until the explosion in Shot 9. No fades, no dissolves, no split screens.
SHOTS
Shot 1 — Wide shot, locked-off, from the far corner of the room, 63° FOV. Ernesto eats in the pool of light while the ceiling fan turns slowly above him; the two bodyguards stand motionless against the back wall. Flaco walks in from screen-left and stops beside the empty chair, standing, hands at his sides. He does not sit. Ernesto keeps eating and ignores him.
Shot 2 — Medium close-up on Ernesto, 47° FOV, eye level, static. He tears a tortilla, keeps chewing, eyes on his plate. Without looking up, he points his fork at the empty chair. Relaxed, unhurried: "Siéntate, Flaco."
Shot 3 — Medium shot on Flaco, 47° FOV, static. Only now does Flaco pull out the chair and sit down slowly, on the front edge of the seat, back straight, hands resting stiffly on his knees.
Shot 4 — Medium close-up on Ernesto, 47° FOV, static. Still eating, still not looking at him. Casual, almost friendly, mouth half full: "¿Y? ¿Cómo salió lo de Tepic?"
Shot 5 — Reverse: medium close-up on Flaco over Ernesto's out-of-focus right shoulder, 47° FOV; one bodyguard is visible out of focus in the background. Flaco rubs his palms slowly along his thighs, swallows, his eyes flick from the plate to Ernesto's face. Fast and nervous, words tumbling out: "Los Mendoza nos cayeron en la carretera, patrón. Tres camionetas. Se llevaron la mercancía. Toda."
Shot 6 — Close-up on Ernesto, 29° FOV, static. His chewing slows… and stops. A long silence. For the first time he lifts his eyes to Flaco. No expression at all. Flat, quiet, slowly, as a statement: "Toda."
Shot 7 — Reverse close-up on Flaco, 29° FOV, handheld with an extremely subtle breath-like float. He cannot hold Ernesto's gaze; his eyes drop, then come back. His jaw tightens, his breathing becomes shallow, his fingers grip his knees. Voice cracking, pleading: "Nos agarraron por sorpresa. No pude hacer nada, patrón. Se lo juro."
Shot 8 — Extreme close-up insert on the table, 29° FOV, static. Ernesto sets his fork down beside the plate with a soft click. His hand rests flat on the oilcloth; his fingers slowly curl into a fist. Silence; only the creak of the ceiling fan.
Shot 9 — Medium shot on Ernesto, 47° FOV, handheld, the camera jolting once with the impact. He slams his open palm onto the table; the shot glass jumps and spills a little tequila, the dominoes box rattles. He leans over the table toward Flaco, veins visible on his neck, face red, shouting hard: "¡La quiero de vuelta! ¡Haz lo que tengas que hacer!" Behind him, both bodyguards uncross their arms and take one step forward.
Shot 10 — Close-up on Ernesto, 29° FOV, slow push-in ending on his eyes. Still leaning in, his breathing heavy, his face half in shadow under the lamp. His voice suddenly drops to a cold, quiet whisper, each word deliberate: "O no vuelves a ver salir el sol."
Shot 11 — Two-shot in profile across the table, 47° FOV, slow lateral drift of less than half a meter to the right, ending with both faces and the table between them in frame, the bodyguards out of focus behind. Ernesto slowly sits back, smooths his shirt, picks up the shot glass and drinks the tequila in one swallow. Flaco stays frozen in his chair, breathing hard, staring at him. Hold in silence.
CHARACTER ACTING
Naturalistic film acting with a steady, rising tension. Ernesto: total, unhurried calm in Shots 1–4 (he ignores Flaco, eats, never looks at him); the first eye contact in Shot 6 is the turning point; quiet tension in Shot 8; a sudden, real explosion of rage in Shot 9 (jaw clenched, eyes wide, neck tendons tight); then an instant return to icy control in Shot 10, which is the most frightening moment. Flaco: tension held inside the body — stiff posture, hands on his knees, rubbing his thighs, swallowing, quick eye movements, shallow breathing; he flinches back in his chair at the slam. His skin stays dry. The bodyguards stay expressionless and silent throughout. Faces show living micro-expressions and natural irregular blinks.
LIGHTING
One warm 3000K hanging lamp over the table is the key source: hard top light, fast falloff, deep shadows filling the room, eye sockets partly in shadow; the bodyguards are mostly in shadow, lit only on their shoulders and the edges of their faces. The slowly turning ceiling fan above the lamp casts soft rhythmic shadows across the table and faces; the lamp swings slightly after the slam in Shot 9. A faint red neon glow from the cantina sign outside leaks through the slats of the closed blinds in the background. Light color stays identical after every cut.
AUDIO
All dialogue in natural, clear Mexican Spanish with an authentic Mexican accent and intonation. Ernesto: deep voice, casual and slow at first, flat and quiet, then a loud raw shout in Shot 9, then a cold whisper in Shot 10. Flaco: tense tenor voice, fast and breathy, cracking when he pleads. The bodyguards never speak. Each speaker has a distinct voice. Only one person speaks at a time, no interruptions or overlapping, natural pauses between lines. Lip movements synchronized with the Spanish dialogue. Ambience: creaking ceiling fan, distant refrigerator hum, faint street traffic and a dog barking far away outside, fork against the plate, chewing, the chair scraping on the floor tiles in Shot 3, Flaco's breathing, a heavy palm slam on the table, the shot glass knocking on the wood. Ambience stays below the dialogue. No background music.
SUBTITLES
Burned-in English subtitles, bottom-center inside the title-safe area, positioned 8% above the bottom edge of the frame. White sans-serif text with a thin black outline, maximum two lines, stable position and size in every shot. Each subtitle appears and disappears exactly in sync with its spoken line:
Shot 2: "Sit down, Flaco."
Shot 4: "So? How did the Tepic job go?"
Shot 5: "The Mendozas hit us on the highway, boss. Three trucks. They took the merchandise. All of it."
Shot 6: "All of it."
Shot 7: "They caught us by surprise. I couldn't do anything, boss. I swear."
Shot 9: "I want it back! Do whatever you have to do!"
Shot 10: "Or you'll never see the sun rise again."
Subtitle spelling must match exactly. No other text on screen.
STYLE
Gritty Mexican crime drama. 35mm film texture with organic grain, light haze in the air, shallow depth of field, crushed blacks, soft halation around the lamp, no digital sharpening. Palette: 55% deep black and shadow, 30% warm amber and worn green, 15% red neon accent.
POSITIVE CONSTRAINTS
Exactly four people for the entire scene: Ernesto, Flaco and two bodyguards. Same faces, hair, clothing, bruise, gold chain, watch and signet ring in every shot. Ernesto always screen-right looking left; Flaco always screen-left looking right; the bodyguards always against the back wall. Flaco stands in Shots 1 and 2 and sits down only in Shot 3, after Ernesto tells him to; from Shot 3 onward he stays seated. Exactly 7 spoken lines: 5 by Ernesto, 2 by Flaco. Ernesto stays seated; the bodyguards take only one step forward in Shot 9. Hands and fingers anatomically correct. Faces stable. The scene ends in silence with Ernesto drinking his tequila and Flaco frozen. How to write prompts for this model
Tag every input and give it one job
Kling numbers uploads in order, as @Image1 or @Video1, and asks you to state each one's role, such as @Image1: preserve the character's appearance; IDs without roles make the result less predictable. 65% of the guide's 20 examples use these tags, against 6% of creator prompts. Source: Kling 4.0 user guide · AgentCraft: Kling 4.0 vs Seedance 2.5 prompt data
Write in four parts: roles, settings, timeline, audio
The guide's structure for Omni Reference is the asset roles first, then the overall settings, then a timed sequence of events, then the audio. Its timed examples plan for 15 seconds; creators on X plan for 30, with a median of six timed segments. Source: Kling 4.0 user guide · AgentCraft: Kling 4.0 vs Seedance 2.5 prompts
Edit with Edit, not Reference
Start with Edit @Video1, say what changes, and end with keep everything else unchanged. The guide warns that writing Reference @Video1 may start a new generation from the clip instead of editing it. Its 7 edit examples have a median of 352 characters on one line. Source: Kling 4.0 user guide · AgentCraft: Kling 4.0 vs Seedance 2.5 prompt data
For keyframes, name the first and last frame
In Keyframes mode write Use @Image1 as the first frame, optionally @Image2 as the last frame, then describe what happens in between. This mode takes up to 10 images and no other reference assets. Source: Kling 4.0 user guide
State exactly how many people are in the scene
Several of the guide's examples open with a scene context line that fixes the count, such as exactly 1 dragon, no duplicates. 46% of its examples outside edits do this, against 18% of creator prompts. Source: Kling 4.0 user guide · AgentCraft: Kling 4.0 vs Seedance 2.5 prompts
Keep negative instructions short
The guide says only two negatives are really needed: no subtitles and no background music. For animation, say it should not look photorealistic. Creators write far more: 94% of their prompts here contain no, avoid or without. Source: Kling 4.0 user guide · AgentCraft: Kling 4.0 vs Seedance 2.5 prompts
Spell out subtitles if you want them
When you do want on-screen text, the guide asks for the font size, alignment, when each line appears and whether it follows the dialogue. Source: Kling 4.0 user guide
Questions about this model
What is Kling 4.0?
Kuaishou's video model, which Kling describes as a next-generation native multimodal model for video creation. It makes video with stereo sound from text, images, video clips, elements and voice profiles, and it can edit existing clips. Source: Kling 4.0 user guide
How long can a Kling 4.0 video be?
3 to 30 seconds per generation, set in one-second steps, 10 seconds by default. Kling 4.0 Flash makes 3 to 20 seconds at 720p. Video extension up to 2 minutes is listed as coming soon. Source: Kling 4.0 user guide
How many references can Kling 4.0 use?
Up to 15 inputs per task in Omni Reference: up to 10 images of 30 MB or less, 5 videos of 3 to 30 seconds with 30 seconds in total, 7 elements including up to 3 video elements, and 3 voice profiles. Source: Kling 4.0 user guide
Which languages can characters speak?
Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, plus Sichuanese and Cantonese, and accents such as British and Indian English. Source: Kling 4.0 user guide
How do I edit a video with Kling 4.0?
Upload up to 5 clips, write Edit @Video1 and the change, and add keep everything else unchanged. Avoid Reference @Video1, which may generate a new video from the clip instead. Source: Kling 4.0 user guide
Do Seedance 2.5 prompts work on Kling 4.0?
As a first draft. Creators write both the same way, about 2,300 characters with six timed segments over 30 seconds. Change how you tag inputs, write edits in one line, and try 15 seconds if 30 drifts. Source: AgentCraft: Kling 4.0 vs Seedance 2.5 prompts