Kling 4.0 Prompts: How to Write Shots, Motion and Sound

Kling 4.0 prompts work as shot plans: describe the subject, what changes, how the camera moves and what the viewer hears. These Kling 4.0 prompt examples come from public creator posts and published guides, grouped by use case, with the original source linked on every entry. For specifications, supported inputs and prices, see the Kling 4.0 model page. Effect Video is a separate service from Kling AI.

Kling AI's announcement video, Meet KLING 4.0 | You Call the Shots, published September 28, 2026, gives a look at the model in under two minutes before the examples.

Creator posts below link to the original videos and credit the people who made them. Entries labeled “More prompt patterns” are public writing examples from other sites, and we did not generate or verify their results. Where X shows a paid-partnership label on a post, or the post carries the #KlingECP hashtag, the entry says so. A post without a label is not proof that nobody paid for it.

How do you write Kling 4.0 prompts?

Describe visible details first, then add the movement, shot order and sound that matter to the scene.

Kling’s AI Video Prompt Guide organizes a prompt into subject, action and motion, environment, camera movement, lighting, style and pacing, and constraints. Use those categories to check a scene you already have in mind.

For a Kling 4.0 prompt guide straight from Kling, read section 2.2 of the Kling 4.0 User Guide. It favors descriptive language over abstract terms: its example replaces a “premium look” with side backlighting, a blurred background and fine matte surface grain. Its FAQ suggests starting with one sentence about the subject, scene, action and emotion, then adding only the details that change the result. The guide lists up to 8,000 tokens of text and says longer prompts are not always better.

For asset-based work, assign each uploaded asset a role before describing the scene. Then give the overall scene and camera tone, events in order, and audio. Kling numbers assets by upload order, so check every @Image1 or @Video1 reference.

Kingy AI’s tutorial runs six Kling 4.0 tests: dynamic motion, Omni Reference, editing, multi-shot storytelling, a text-only prompt and lip sync. Sponsored by Kling. The video description says so. Several tests use features Effect Video does not offer, and the description mentions the prompts without listing them.

UGC and talking heads

Specify the recording setup, the person’s behavior and the spoken performance to make a UGC prompt concrete.

Brunch selfie

Min Choi (@minchoi), October 17, 2026. X labels the first post of this thread as a paid partnership. Full prompt.

Raw vertical 9:16 video recorded on a phone's front camera (not on screen), one continuous 30-second take with no cuts.

The prompt asks for a slightly soft image, a little digital noise, deep focus and auto-exposure that shifts as people pass the window. It also plans one accident: around 9 seconds a waiter bumps the table and the phone slides and tilts.

A summer walk from a partner’s viewpoint

あぎ (@agi_aibusi), October 16, 2026. Full prompt.

Their familiarity should come through in how she waits for him, looks at him, and closes the distance between them.

The creator says in the post that this 30-second clip came from a single text-to-video generation. The prompt tells a small story of a girl on a summer walk in Japan. She gradually notices her boyfriend’s camera and starts to include him, and the prompt describes her outfit in detail, down to a small crossbody bag.

More prompt patterns: a studio talking head

A public text example from PixVerse. Read the full prompt.

Medium close-up of a woman speaking directly to camera. Softbox key light from camera-left, subtle hair light, neutral grey studio background

The example places a softbox key light camera-left with a subtle hair light against a neutral grey background, and asks for shallow depth of field with sharp eye focus. We found no clip generated from it.

More prompt patterns: a sunscreen selfie

A public text example from each::labs, published October 7, 2026. Read the full prompt.

Handheld selfie framing, 9:16, 12 seconds. A woman in her twenties sits in her car in a parking lot, daylight through the windshield.

each::labs published its guide on October 7, before Kling 4.0 launched, and says it extrapolates from Kling 3.0 habits where Kling had published no guidance. It packs holding up the product, a line of dialogue and rubbing the cream on the back of a hand into a 12-second vertical clip. The dialogue is scripted. It is not a customer testimonial and says nothing about how any sunscreen performs.

Product ads

Assign product appearance to the reference material and describe the shot sequence separately.

A car commercial built from reference images

Murphy (@Diplomeme), October 16, 2026. The post uses the #KlingECP hashtag and does not say whether the work was paid. Full prompt.

High-octane, hyper-realistic automotive commercial for the Kia Seltos

The prompt asks for 15 two-second shots across a 30-second timeline and points to reference images for the car’s design. The post describes the result as 30 seconds at 1080p on Kling 4.0, and the “4K resolution” line in the prompt is a request.

More prompt patterns: a soda-bottle whip pan

A public text example from Bennett Heyn at fal. Read the full prompt.

After a brief hold, whip the camera horizontally to the right, turning the booths and lights into streaks of color.

The prompt takes the finished scene, including the woman and the bottle, from one supplied image and the whip pan from a reference video. It then reverses the move to land on the opening composition: the same hand position, bottle angle, expression and suspended splash. fal calls the examples on its page official Kling 4.0 results, and we did not play the video. This recipe needs a reference video, which Effect Video does not offer.

More prompt patterns: a camera-charm commercial

A public text example from Noah Collins at Pollo AI, updated October 23, 2026. Read the full prompt.

Mix playful portraits, skating, and product close-ups with smooth match cuts, VHS overlays, energetic synth music, and minimal closing branding.

The 30-second plan moves through a CRT opening, an amber tunnel, cyan skate spaces and red neon gatherings, joined by match cuts around a small camera charm. Pollo AI says it made the video with Kling 4.0 and offers Kling 4.0 in its own product. We did not watch the video.

Animation

Use the starting image or explicit visual details to define the style, then spend the remaining prompt on the action.

Comedy from an opening anime image

Kris Kashtanova (@icreatelife), October 23, 2026. The post says it was made with Kling 4 inside Adobe Firefly. Full prompt.

Create an epic Japanese anime scene that begins exactly from the opening image.

The rest of the prompt asks for a short comic situation built around what the image shows, with multiple shots, natural English dialogue, escalating comedy and an anime-style payoff, set in New York City. The image supplies the visual setup, so the prompt spends its words on the situation and the payoff.

More prompt patterns: a rooftop confession

A public text example from Emma Chen at Nereo, published October 13, 2026. Read the full prompt.

20-second anime scene, 2D hand-drawn style, school rooftop at sunset, cherry petals drifting.

The plan runs four timed beats: a boy waiting at the railing, a close-up of a girl’s nervous eyes as she opens the door, a letter handoff, and a pullback over swirling petals. Sound is wind, a distant school bell and soft piano. We did not find a finished clip for it.

Multiple shots

Give each shot a framing and an action, then state which character and scene details must remain consistent.

Four shots in an alchemist’s workshop

Marco Riccetti (@shikoba_86), October 16, 2026. Full prompt.

Shot 2, medium: she slowly pours a glowing green liquid into a round glass vial.

Four numbered shots each get a framing and an action. A character lock and the location description carry across the cuts, and the audio line lists rain, candles, glass and a whisper, with no music and no subtitles.

Five shots for a coffee delivery

Synthia (@AIwithSynthia), October 23, 2026. X labels the post as a paid partnership, and the post says it was made with Kling 4.0 on Pollo AI. Full prompt.

A busy afternoon on a real Manhattan street.

Each of the five shots gets a time range (0 to 6, 6 to 12, 12 to 19, 19 to 25 and 25 to 30 seconds), an action, dialogue and street sound. The prompt ends with realism constraints such as no narrator and no artificial-looking crowds.

A siege scene with explicit cuts

PIXEL (@itsPixieVerse), October 20, 2026. Full prompt.

CUT DISCIPLINE: 7 distinct hard cuts timed explicitly below

The prompt orders seven hard cuts and tells the model not to render the scene as one continuous take, then gives each shot a duration, camera framing and action. Pan’s tests below show why timed cuts are worth checking against the result.

Eight shots around one kitchen table

MayorkingAI (@MayorKingAI), October 17, 2026. Full prompt.

One scene, 8 shots joined by clean hard cuts, classic shot–reverse-shot coverage.

A husband confesses that he lost his job in June and has been hiding it. The prompt fixes screen positions, eyelines, clothing and wedding rings, and cuts between the speaker and the listener’s reaction.

For dialogue, Kling’s guides ask you to identify who is speaking, put the line in quotation marks, and describe the voice, pace and emotion. Kling advises minimizing negative instructions, but names no subtitles and no background music as exclusions worth writing because the model may add them by default.

Camera movement

Distinguish moving the camera body from changing the lens, and define the route a tracking shot follows.

Snap zoom from a fixed camera

Marco (@ai_artworkgen), October 23, 2026. Full prompt.

One continuous shot, no cuts, the camera body never leaves its position.

The five-second plan holds from 0 to 2 seconds, snap-zooms from 2 to 3, then settles from 3 to 5. It separates lens movement from moving the camera body.

Low tracking beside a skateboard

Min Choi (@minchoi), October 17, 2026, item 5 of the thread whose first post X labels as a paid partnership. Full prompt.

30-second continuous low tracking shot of a skateboarder in a concrete park at golden hour.

The camera stays low while the ride unfolds in six-second stages: pushing off, a weight shift before a stair jump, the landing, a hard left carve and a final wide shot. The prompt repeats the skater’s clothing and face as constants.

Follow a cat through the night

Umesh (@umesh_ai), October 15, 2026. The post calls it one of the creator’s first tries with Kling 4.0, at 1080p. Full prompt.

An amateur handheld recording in third-person perspective rapidly follows a cat through its nighttime adventures in one continuous shot, with no cuts

The route moves through a home, garden, garage, rooftops and streets, with one small disruption at each stop. To adapt it, keep one event per stop.

More prompt patterns: follow a pastry chef into a courtyard

A public text example from Fuser, published October 23, 2026. Read the full prompt.

15-20s: she shoulders open a back door into a quiet, sunlit courtyard and walks across the flagstones.

Fuser’s 30-second plan follows a pastry chef decorating a cake, carrying it through the kitchen at shoulder height, setting it on a courtyard table and walking back inside. Fuser says it generated this one on October 23 and shows the output on its page, and we did not play it. The shoulder-height route is the part to study.

Landscapes

Pair a clear camera path with one visible environmental event, then add the ambient sound.

Push toward a sea cliff

Hailey AI (@Hailey4AI), October 21, 2026. Full prompt.

Smooth forward camera move with a slight rise.

A low drone move approaches the cliff as a wave strikes, with steady grey light and wind and sea sound. This is test 6 of an eight-test image-to-video comparison using the same image and prompt on both models, at 10 seconds and 1080p. Hailey posts both results side by side, says the first result was used and nothing was re-rolled, and calls the Kling 4.0 push the smoother move. Kling 3.0 threw more spray.

A ski sequence with seven camera positions

Min Choi (@minchoi), October 17, 2026, item 3 of the same thread as the brunch selfie, whose first post X labels as a paid partnership. Full prompt.

A 30-second, seven-shot ski film edit of Kai

The full prompt assigns sections to a drone, helmet camera, pole camera, snow-mounted camera, chase skier and distant long lens. A character reference, clothing and sun direction link the views.

More prompt patterns: arrival in a coastal town

A public text example from Aerin Kim at Miraflow, published October 18, 2026. Read the full prompt.

A relaxed 30-second travel vlog opening in a whitewashed Mediterranean coastal town in late afternoon, warm golden light, handheld first-person camera, 24mm lens.

The 30-second first-person route runs from a docking ferry through stepped streets and a café terrace to a rooftop view of the harbor. Keyframes at 0, 14 and 30 seconds set the locations, and seagulls, church bells and footsteps connect them. Miraflow says its images are storyboard or keyframe plans, so they show no generated video.

Creature action

Connect the subject’s movement to the camera response and the objects it touches.

A creature charges through an alien jungle

Alin (@Alin_Reaper05), October 16, 2026. In the post, Alin calls the creature’s movement “almost perfect.” Full prompt.

the creature suddenly notices the camera and rapidly charges forward, handheld tracking camera retreats through foliage, branches break naturally, mud splashes

The prompt ties the creature’s sudden charge to the camera’s retreat and to what the charge breaks and splashes.

Effects

Describe the order of an effect, including the stillness before an impact and the debris afterward.

A bat swarm with atmosphere and sound

Shine by Nous (@Shinebynous), October 23, 2026. X labels this post as a paid partnership. Full prompt.

A colossal bat swarm blotting the sky, leathery wings in inky ebony and subtle wine-red veins.

The prompt separates the effect, location, texture and mood into labeled fields and ends with halloween music.

Weightlessness, impact and falling debris

Dr Bloodline | AI (@AiwithBloodline), October 18, 2026. The post says it was made with Kling 4.0 Flash, which Effect Video does not offer, and it uses the #klingECP hashtag without saying whether the work was paid. Full prompt.

Suddenly, everything around him enters a state of weightlessness and time freezes.

In a school corridor, people and objects hang still, a small gesture releases the impact, and then paper falls.

Transitions

Give each transition a visible bridge and specify the scene or composition it should reveal.

A portal journey that returns to the sofa

The Rogue AI (@TheRogueA1), October 17, 2026. The creator ran it on Topview at 720p and called the result “not what I was expecting,” saying the opening and ending worked, the middle did not, and a more detailed prompt might help. Full prompt.

26-second continuous cinematic sequence, photoreal, film grain, dynamic camera, seamless magical transitions between worlds. Dialogue in English, lip-synced.

A laptop portal pulls the heroine from a living room into a dark-fantasy battlefield, a flight over a city and a war zone, then drops her back on the sofa. Each jump has a visual bridge such as a light vortex, a beam of light or an impact flash, and seven reference images assign identity, costume and location roles. The original includes profanity and a caped hero costume.

First-frame video

Name the image as the first frame, preserve the details that matter and describe what happens next.

Kling 4.0 text to video prompts carry the whole scene in words, while Kling 4.0 image to video prompts hand the opening look to a photo. In Kling’s user guide the pattern is Use @Image1 as the first frame, followed by what happens next. Reference assets have a different role: the guide asks you to say whether each one defines appearance, movement or another property. Keyframes pin visuals at chosen moments, and the prompt describes the movement between them.

Animate a calf from its first frame

OpenClipsAI (@openclipsai), October 22, 2026. Full prompt.

The clip opens on the given first frame and animates it.

Before the butterfly chase and a low tracking camera, the longer prompt locks the calf’s fur, fringe, horns, muzzle and eyes. The linked post stacks Kling 4.0 and Kling 3.0 results from the same prompt and first frame in one 15-second video, and its text does not say which clip is which model. OpenClips says Kling 4.0 follows the requested camera shot, from close-up to low in the flowers.

More prompt patterns: one Lisbon street from day to night

A public text example from Molyin, updated October 17, 2026. Read the full prompt.

In between, the sun drops, the sky turns from pale blue to deep indigo, shop signs flicker on one by one

The example gives a day frame and a night frame of the same Lisbon street. It describes only what changes between them: the sun drops, the sky turns indigo, shop lights come on and a tram passes. Molyin says Kling 4.0 was not yet in its generator and its “Try this” button opened Kling 3.0, so this is a 4.0-labeled writing example with no verified 4.0 result. Effect Video does not accept an end frame.

Timing and macro tests

Compare the timing and camera moves you asked for with what you got, especially when a prompt contains exact cuts.

Pan (@sebatheepan), October 21, 2026, ran three 15-second, 1080p text-to-video tests of Kling 4.0 on Higgsfield with identical settings. The video post carries all three videos, and the three prompts are in a reply. Pan reports that cuts landed early every time: requested cuts at 5, 10.5 and 8 seconds arrived at roughly 4, 10 and 6. Pan lists unbroken takes, heavy water and mechanical physics, and object continuity across a cut as strengths, and precise cut timing, macro settling and compound camera moves as weak spots. These are one creator’s observations from three tests.

A storm surge against a sea wall

A storm surge detonates against a concrete sea wall and the entire structure shudders through to its foundation.

The prompt describes three shots with hard cuts requested at 5.0 and 10.5 seconds. Pan reports that the water kept its mass and drained under gravity without morphing, and that debris that landed stayed landed across the cut.

A gantry crane and a requested dolly zoom

A gantry crane walks a loaded shipping container across a working port at last light, and the camera never leaves it.

The prompt asks for one unbroken take that ends in a dolly zoom. Pan reports zero cuts in 15 seconds and a physical push-in instead of the dolly zoom. The prompt’s timeline ends at 10 seconds, while the post calls the test 15 seconds.

A water droplet in macro

A single droplet of water falls into a still black pool and the surface answers it.

The prompt asks for one six-second shot in 9:16. The post describes a 15-second, two-shot test, so the posted video may not match this prompt exactly. Pan reports strong surface physics in vertical, with two misses: the surface never fully settled, and a stray droplet appeared at 0:03.

Can you use these prompts on Effect Video?

Some of them: text prompts and one first-frame photo fit the generator, and several recipes on this page need more.

Effect Video’s Kling 4.0 generator takes a text prompt and one first-frame photo, and clips run up to 15 seconds. A 20-, 26- or 30-second prompt needs a shorter shot plan there. Reference videos, multiple reference images, extra keyframes, end frames, video editing and the Flash model are not offered. Supported inputs, settings and credit prices are on the Kling 4.0 model page.

In Kling’s own editing workflow, the user guide says to begin with Edit @Video1, specify the change and add “Keep everything else unchanged” when appropriate. It warns that Reference @Video1 may be treated as reference-based generation instead. That editing pattern does not apply to Effect Video.

Open-source recipes from before Kling 4.0

These three recipes were written before Kling 4.0 launched, so they are shot-plan templates with no 4.0 results behind them.

Flaq AI’s repository is titled Awesome Kling 4.0 Prompts, but the three prompt files were last changed on August 20 and September 1, 2026, before Kling 4.0 launched on October 15. The repository README, updated September 29, calls the full 4.0 model announced and Flash an early-access release. The recipes come with no results for any Kling model. They end with long “No …” lists, while Kling’s guide advises minimizing negative instructions, so try them with those lists trimmed. The full prompts follow under the MIT license, Copyright (c) 2026 Flaq AI, and the license text closes this section. The seedling-to-soup recipe needs four reference images, which Effect Video does not offer.

An ink-line courier riding a paper bird

Flaq AI (flaqai). Original recipe. Last changed August 20, 2026.

[OUTPUT]
12 seconds, 16:9, original hand-drawn 2D animation, expressive black brush line, visible paper grain, limited cel shading, palette of indigo, cream and vermilion, three shots.

[CHARACTER LOCK]
Courier Iko: adult nonbinary rider, short triangular indigo hair, cream wrap jacket with one vermilion sash, black cropped trousers, rectangular canvas mail case. Original design; preserve silhouette, face, colors and case.
Mount: long-legged blue paper bird with folded geometric wings and one cream beak.

[WORLD]
Vertical city built from stacked handwritten envelopes, suspended bridges made of red thread, cream sky with broad dry-brush clouds. No readable writing; marks remain abstract.

[TIMED SHOTS]
0–4s — Side-scrolling wide shot. Iko and the bird run left-to-right across a red-thread bridge. Animation uses strong contact poses: each foot compresses the thread, which rebounds after release. Camera pans at matched speed.
4–8s — Low front angle. A wall of loose envelopes slides down the street behind them. Iko pulls one ribbon; bird folds both wings tightly and slips through a narrow triangular gap. Speed lines remain brush strokes attached to direction of travel.
8–12s — Overhead shot rotates only 30 degrees as bird unfolds wings and glides over an ink canal. Iko opens the mail case; one glowing blank envelope rises, stamps the dark canal with a cream reflection, then returns to the case. End on a clean heroic silhouette.

[AUDIO]
Dry paper percussion, brush swishes, stretched thread plucks and rapid soft foot taps. One low wooden drum on the wing unfold. No dialogue or recognizable melody.

[CONSTRAINTS]
Consistent 2D drawing, line weight, paper texture and limited palette. Preserve original character and mount. No photorealism, 3D rendering, readable handwriting, existing uniforms, franchise symbols, extra limbs, sash color change, logo or watermark.

Four reference images from seedling to soup

Flaq AI (flaqai). Original recipe. Last changed September 1, 2026.

[OUTPUT]
12 seconds, 16:9, four-stage farm-to-table visual poem, smooth realistic transitions, no people or branding.

[REFERENCE ROLES]
Image 1 locks a tomato seedling in a terracotta pot. Image 2 locks one mature tomato cluster on vine. Image 3 locks one red tomato on a wooden cutting board. Image 4 locks one bowl of tomato soup on a linen table. Use only the central subject and its surface; ignore unrelated background objects.

[CORRESPONDENCE]
Keep the main round red form centered at 55% frame width. Growth direction is bottom-to-top; the final camera direction remains left-to-right. Use leaf, circular fruit and steam shapes as visual bridges.

[TIMED TRANSITIONS]
0–3s — Macro push toward seedling. Two leaves unfold naturally; the central leaf fills frame.
3–6s — Cut within leaf occlusion to mature vine. Camera continues the same push. One green tomato changes to red through a short time-lapse while its position and size stay fixed.
6–9s — The red tomato rotates only 20 degrees; a matching red circular wipe formed by its surface reveals the same tomato resting on the cutting board. A knife remains parked safely outside frame edge and never moves.
9–12s — Camera passes behind a rising soft steam veil from off-screen cooking; reveal finished soup bowl with one basil leaf. End on a stable overhead composition.

[AUDIO]
One continuous gentle room tone evolves from garden birds to kitchen ambience. Leaf brush, soft wooden contact and quiet simmer bridge transitions. No narration or music.

[CONSTRAINTS]
Four stages in fixed order. No liquid morph directly from raw whole tomato, impossible knife action, hands, duplicated fruit, reversing camera, text, logo or watermark. Transition occlusions must hide, not display, geometry changes.

Scallion-pancake macro and ASMR

Flaq AI (flaqai). Original recipe. Last changed August 20, 2026.

[OUTPUT]
12 seconds, 9:16, three-shot food ASMR, natural kitchen light, honest texture, no presenter face.

[FOOD AND HAND LOCK]
One round layered scallion pancake, 16 centimeters wide, pale dough with evenly distributed green scallion rings. Adult cook’s clean hands, short natural nails, dark-blue apron. Black carbon-steel pan and one flat wooden spatula.

[WORLD]
Small home kitchen beside an open window, soft morning side light, pale stone counter. Background remains quiet and out of focus.

[TIMED SHOTS]
0–4s — Top-down macro. Cook rolls the layered dough forward exactly once; visible scallion rings compress but stay inside the dough. Fine flour shifts outward under the rolling pin. Audio emphasizes wood on dough and dry flour.
4–8s — Side macro at pan height. Pancake lies flat in a thin film of oil. Edges bubble and turn golden from left to right. Wooden spatula lifts one edge to reveal crisp concentric layers, then lowers it; no flip occurs in this shot.
8–12s — 45-degree close-up on cutting board. Cook makes one clean center cut with a broad knife, then gently separates the two halves by two centimeters. Steam escapes and flaky layers open. Hold for one second on texture.

[AUDIO]
No music or speech. Rolling friction, quiet window birds, controlled oil sizzle, spatula scrape, one crisp knife contact and delicate crackle as layers separate. Audio perspective matches each macro shot.

[CONSTRAINTS]
One pancake, one pan, one spatula, one knife. Correct tool contact and adult hands. Food does not regenerate, jump position or burn. No excessive oil splash, floating steam, raw center after cooked exterior, extra fingers, text, packaging, logo or watermark.

License for the three recipes

Flaq AI repository · MIT license.

MIT License

Copyright (c) 2026 Flaq AI

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

About this page

This page combines Kling’s published guidance with public creator posts and prompt-writing examples from other sites. Every source was opened on October 24, 2026 and can change after that date. We did not generate any video for this page, and we did not play every linked video in full.