01 · Subject
Who or what
The one thing the shot is about, described concretely enough to picture.
a frosted glass perfume bottle
40+ Copy-Ready Prompts · Updated 2026-08
Kling AI Prompt Guide
Good Kling AI prompts are not longer than bad ones — they are more specific about six things: subject, action, camera, lighting, style and duration. Get those right and the model stops guessing.
Kling reads a prompt as one shot description, not a story, and that single fact explains most of what follows. Everything below is written to be copied. Ten weak-versus-strong rewrites, a camera-move vocabulary, keyword banks you can paste from, and full prompts for the shots people actually need through the Kling API.
The Anatomy of a Kling AI Prompt
Six components, always in roughly this order. Drop one and Kling fills the gap with its own average — which is exactly the flat, generic look people complain about.
02 · Action
What it does
A single movement that can complete inside the clip length.
rotating slowly on wet slate
03 · Camera
Where you stand
Shot size plus one named move. Unstated, and Kling holds near-static.
macro, slow dolly in
04 · Lighting
Where light comes from
Direction and quality. This is what separates cinematic from flat.
single softbox from camera left
05 · Style
How it should look
Medium, lens, era or grade — pick two, not six.
photoreal, shallow depth of field
06 · Duration
How long it runs
Set it first; Kling caps one clip at 15 seconds.
5s
A frosted glass perfume bottle on wet black slate, rotating slowly through fifteen degrees, macro lens, slow dolly in, single softbox from camera left with a dark background, photoreal product film, shallow depth of field, 5s
One sentence, six components, no adjective stacking. This is the shape every Kling API prompt on this page follows.
How Kling Reads What You Wrote
Three behaviours explain most surprises. None of them are bugs in the Kling API — they are the model doing exactly what the words asked.
Order carries weight
Kling leans on the opening clause hardest. A style declared at the end of a long sentence competes with a scene Kling has already committed to.
One clip is one shot
Kling fits everything you asked for into the duration you set. Two scenes in a 5s prompt become one confused Kling render, not a cut.
Silence becomes default
Anything you leave out, Kling supplies from its own average: even lighting, a near-static camera, a generic grade. Specify or inherit.
Ten Real Rewrites
Weak vs Strong Kling AI Prompts
Same intent on both sides. The right column wins because it removes decisions from Kling, not because it is longer.
Kling prompt rewrites — copy the right-hand column
| Shot | Weak | Strong | What changed |
|---|---|---|---|
| Product | a perfume bottle | A frosted glass perfume bottle on wet black slate, rotating slowly through fifteen degrees, macro lens, single softbox from camera left, water beads catching the rim light, shallow depth of field, 5s | Material, surface, a bounded rotation and one light source — none of which Kling can guess. |
| Portrait | a woman looking at the camera | Close-up of a woman in her thirties beside a rain-streaked window, she turns her head toward the lens and half-smiles, cool overcast light from the left, 85mm, shallow depth of field, subtle film grain, 5s | A specific micro-action instead of a pose, plus lens and light. |
| Landscape | beautiful mountains at sunrise | Aerial drone pull-back over a fog-filled alpine valley at sunrise, pine ridges emerging through the cloud layer, warm rim light on the peaks, slow steady ascent, cinematic wide, 10s | A camera move gives Kling something to animate in an otherwise static vista. |
| Food | tasty ramen | Overhead shot of tonkotsu ramen in a black bowl, steam rising, chopsticks lifting noodles up out of frame, warm practical light from a paper lantern above, macro detail on the soft-boiled egg, static camera, 5s | Angle, one human action and a named practical light. |
| Action | a fast car chase in the desert | Low tracking shot beside a red convertible speeding across cracked desert clay, dust trail billowing behind, hard midday sun, camera trucks right matching the car speed, motion blur on the wheels, 5s | Camera height and a matched move turn speed into something readable. |
| Fashion | a model in a nice dress | Full-body shot of a model in a flowing emerald silk dress walking toward camera down a raw concrete corridor, fabric trailing in the draft, hard rim light from behind, slow dolly out, 24fps motion blur, 10s | Fabric behaviour is the whole shot, and backlight is how Kling makes it visible. |
| Interior | a nice living room | Slow dolly forward through a mid-century living room, late afternoon sun through slatted blinds striping the oak floor, dust motes drifting in the beam, warm neutral grade, wide 18mm, 10s | Light becomes the subject; the room is just where it lands. |
| Anime | anime girl in the rain | 2D anime cel style: a schoolgirl under a clear umbrella on a neon-lit crossing at night, rain streaking past, reflections rippling in the puddles, camera cranes down to eye level, saturated magenta and cyan palette, 5s | Kling inherits the medium from the first clause, so declare it before the scene. |
| Dialogue | someone talking to camera | Medium close-up of a bearded barista behind a counter speaking to camera: we roast every Tuesday morning. Warm window light from the right, espresso steam behind him, static shot, natural lip sync, 5s | A quoted line plus a static frame is what the native-audio Kling versions want. |
| Logo sting | cool logo animation | A brushed-metal logo mark forming out of swirling ink in black water, backlit rim highlight along the edges, camera pushes in slowly as the mark settles into focus, high-contrast monochrome, 3s | A physical metaphor beats the word cool every time. |
Every prompt above is written for a single Kling API request. For anything longer than 15 seconds, write one of these per shot and stitch the results.
Camera Move Vocabulary for Kling API Prompts
Kling responds to film-set language. Use the term on the left and you get the move; describe it vaguely and Kling gives you drift.
Thirteen camera moves and the phrasing Kling responds to
| Move | What happens | Write it as | Best for |
|---|---|---|---|
| Dolly in | Camera physically travels toward the subject | slow dolly in | Building focus on one thing |
| Dolly out | Camera pulls back, revealing context | slow dolly out to reveal | Reveals and endings |
| Truck / track | Camera slides sideways, staying parallel | camera trucks right | Following lateral motion |
| Pan | Camera pivots horizontally on its axis | slow pan left across the room | Scanning a space |
| Tilt | Camera pivots vertically | tilt up from the feet | Scale and reveals |
| Crane / boom | Whole camera rises or descends | camera cranes down to eye level | Entering a scene |
| Orbit / arc | Camera circles the subject | camera orbits slowly clockwise | Products, hero objects |
| Handheld follow | Loose, breathing frame trailing a subject | handheld follow behind her | Documentary energy |
| Drone pull-back | Aerial retreat and ascent together | aerial drone pull-back | Landscapes, establishing |
| Push-in POV | First-person forward movement | POV walking forward down the hall | Immersion, games |
| Rack focus | Focus shifts between two depths | rack focus from the glass to her face | Directing attention |
| Whip pan | Fast blur-through pivot | whip pan to the doorway | Energy, transitions |
| Locked off | No camera movement at all | static locked-off shot | Dialogue, product loops |
One move per clip. Two named moves in a five-second Kling prompt is the fastest way to get neither of them.
Kling Prompt Keyword Bank: Style and Lighting
Paste from these Kling keyword banks instead of reaching for cinematic a fourth time. Two keywords per prompt is plenty — stacking six cancels them out.
Lighting
Direction and quality. Name one source for Kling, not three.
golden hourblue hoursingle softbox keyhard noon sunrim light from behindpractical neon signagecandlelightovercast diffusevolumetric god raysbounce fill from white cardsilhouette backlightmoonlight through blindsfirelight flickerscreen glow on the face
Lens & film
How the camera itself shapes the image.
18mm wide35mm documentary85mm portraitmacro detailanamorphic flareshallow depth of fielddeep focussubtle film grain24fps motion blurtilt-shift miniaturefisheyelong lens compression
Visual style
Declare the medium first — everything Kling renders after it inherits.
photorealcinematic film lookdocumentary handheld2D anime cel3D animated featureclaymationstop motionwatercolour illustrationcyberpunk neonfilm noirretro VHSpastel minimalhand-drawn storyboardtech product film
Mood & grade
The colour temperature of the whole frame.
warm amber gradecool teal gradehigh contrastmuted desaturatedhigh key brightlow key moodymonochromesun-bleachedsaturated magenta and cyanearthy natural tonesclinical clean whitenostalgic faded
Negative prompts earn their place here. If a version of the Kling API you are calling accepts one, use it for what you keep accidentally getting rather than as a wish list: extra fingers, warped text, watermark, duplicated subject, jittery camera. A negative prompt full of aesthetic adjectives does almost nothing.
Copy, Paste, Adjust One Noun
Kling AI Prompts by Use Case
Six full Kling prompts, each already carrying all six components, for the shots people actually need from the Kling API. Swap the subject for yours and leave the structure alone.
Product commercial
Hero object, controlled light, one bounded Kling move
A matte-black wireless headphone resting on brushed concrete, rotating slowly through twenty degrees, macro lens pushing in, single softbox from camera right with a deep shadow side, photoreal product film, shallow depth of field, 5s
Vertical social clip
9:16 Kling shot, single subject, motion that fills the frame
Vertical 9:16 shot of a barista pouring latte art into a white cup, steam curling upward, her hands entering frame from the right, warm window light from the left, handheld follow with a slight sway, documentary look, 5s
Real estate walkthrough
Let the Kling camera carry the shot, not the subject
Slow dolly forward through an open-plan kitchen into a sunlit living room, late afternoon light through floor-to-ceiling glass, dust motes in the beam, wide 18mm, warm neutral grade, steady gimbal feel, 10s
Character close-up
One micro-expression beats a described emotion
Close-up of an older fisherman on a harbour wall, wind moving his grey hair, he squints then looks off toward the horizon, hard low sun from camera left, 85mm, shallow depth of field, subtle film grain, 5s
Anime / stylised scene
Declare the medium to Kling in the first four words
2D anime cel style: a lone traveller on a cliff path at dusk, cloak snapping in the wind, distant floating islands in the sky, camera cranes up and back to reveal the scale, saturated orange and indigo palette, 10s
Abstract brand sting
Physical metaphors render better than adjectives
Iridescent liquid metal blooming outward from a single point on a black field, surface tension rippling, camera pushes in slowly as the shape resolves into a smooth sphere, backlit rim highlight, high contrast monochrome, 3s
When a Kling Prompt Does Not Land
Match the symptom on the left to the wording problem in the middle. Almost none of these are Kling failures — they are prompt failures.
Kling symptom → likely cause → fix
| Symptom | Likely cause | Fix |
|---|---|---|
| The shot barely moves | No camera move and no verb — Kling has nothing to animate. | Add one named move from the vocabulary table and one active verb. |
| The subject morphs mid-clip | Text-only prompting on a version without subject reference. | Generate a key frame first, then animate it, or switch to a reference-capable Kling version. |
| Everything looks flat and grey | No light source named, so Kling defaults to even ambient. | State direction and quality: hard sun from camera left, long shadows. |
| The style keeps slipping to photoreal | The style keyword arrived at the end, after a realistic scene description. | Put the medium first: 2D anime cel style, before anything else. |
| Text in frame is unreadable | Too many small glyphs asked of a Kling video model. | One short word, large in frame — or add it in post instead. |
| Action is cut off half-finished | A multi-beat action in a 3–5 second clip. | Slow the action down, or move to 10s and sequence it explicitly. |
| Faces distort in motion | Fast movement plus a small face in a wide frame. | Get closer — medium or close-up — and reduce the speed of the move. |
| Output ignores half the prompt | Adjective stacking; competing instructions cancel each other. | Two modifiers per element, one camera move, one light source. Cut the rest. |
For failures that are not about wording — timeouts, Kling model choice, aspect-ratio mistakes — see the Kling API tutorials index.
Test-Drive the Guide
Try a Kling-Style Prompt on a Real Model
Click an example to load it, edit a noun or two, and render a genuine key frame. It is the fastest way to feel how much one extra clause changes, and it mirrors the first half of the Kling API workflow.
Quick start
Frame generation is real and runs on our own image model; the animation step is a guided preview.
Kling AI Prompt Questions
Four things people ask once they have written a dozen prompts and want to know what the rules really are.
One to three sentences, roughly 25 to 60 words, is the sweet spot for a Kling API request. That is enough to carry all six components without any of them competing. Past about 80 words instructions start cancelling each other out, and you will notice Kling quietly dropping whichever clause it liked least.
Several versions accept a negative prompt field, and the exact name varies by whichever Kling API provider you call. Use it for concrete artefacts you keep seeing — warped text, duplicated limbs, watermarks, camera jitter — rather than for aesthetic wishes. Negative prompts are a filter, not a second creative brief.
Kling handles multiple languages, and the 3.0 series explicitly documents mixed-language dialogue including Chinese, English, Japanese, Korean and Spanish. For the scene description itself, English film vocabulary is the safest choice because terms like dolly, rack focus and rim light map cleanly to the behaviour you want.
Because a text prompt describes a subject to Kling rather than fixing one. The reliable fix is to stop re-describing and start referencing: generate a key frame you are happy with and animate that, or use a subject-reference capable model such as Kling 3.0 Omni. Repeating the same adjectives across takes will never produce the same face twice.
Take a Prompt All the Way
You have the vocabulary and forty examples. Render a key frame from one of them, then push the whole Kling API workflow through to a finished clip.
No signup required40+ copy-ready promptsReal frame in seconds
Start Creating FreeNo signup · No credits · Runs in your browser