Kling AI prompts work best when they read like one shot from a film script: who is in the frame, what they do, how the camera moves and what you hear. Kling 4.0, which Kuaishou released on 15 October 2026, can make clips of up to 30 seconds in one pass according to its launch announcement, takes several reference images and adds stereo sound with lip sync, so a prompt now steers far more than a few seconds of motion.
The prompts below cover long takes, dialogue, image to video, product clips and camera moves. Each has a note on why it works and what to swap. Words in square brackets are placeholders, so replace [city] or [product name] before you generate.
The short answer
Open Kling, choose text to video or image to video, paste a prompt and replace the words in brackets. Copy the negative list into Kling’s negative prompt box, set the length shown, and change only one thing per retry.
How should a Kling AI prompt be structured?
Kling follows plain sentences and rewards a steady order. Describe one shot at a time and keep the action to one or two steps: a clip that has to fit a chase, a hug and a sunset usually rushes all three. Concrete verbs help more than adjectives, so “turns the page and looks up” beats “an emotional moment”.
An order that suits most clips:
- Shot type and framing: wide shot, medium close-up, low angle.
- Subject and setting, with one or two visual details.
- The action, in the order it happens.
- Camera movement: static, slow push-in, tracking, orbit.
- Light and style: time of day, light direction, film or animation look.
- Sound, as a separate last sentence.
Kling AI image to video prompts
When you start from a photo, Kling already knows what everything looks like, so the prompt should describe motion, not the picture. Say first what must stay the same, then what moves and how much. Small physical movements such as a blink, a breath, hair in the wind or rising steam look natural; asking a seated person to stand up and walk away often bends the body or the room.
Kling 4.0 also accepts several reference images at once, for example a person, a product and a place. Refer to them by number (“the woman from image 1”) so each picture has a clear role.
How do I add sound to a Kling video?
Put sound in its own sentence at the end, starting with “Sound:”, and name two or three things you want to hear, such as rain on a window, a kettle and distant traffic. For speech, say who talks and put the words in quotation marks. Short lines sync better than speeches, and one speaker per shot is easier than a quick back-and-forth. If you plan to add music in an editor later, write “no music” in the sound line.
14 prompts to copy and adapt
Thirty-second long take at a pottery wheel
Single continuous shot with no cuts. Close-up to medium shot of a potter’s hands shaping a tall vase on a spinning wheel, from a wet lump of clay to a finished curved neck. The camera slowly arcs a quarter turn around the wheel and rises a little at the end. Soft window light from the left, dust in the air, warm natural colors, realistic documentary look.Avoid: cuts, flicker, warped hands, extra fingers, text, watermark
One action that develops over time suits a long take far better than five separate events. If the hands warp, try 15 seconds first.
Two-line dialogue at a café counter
Medium two-shot in a small neighborhood café in the morning. A barista in a green apron slides a cup across the counter and says, “Your usual, extra hot.” The customer, a man in a rain jacket, smiles and answers, “You remembered.” Static camera at eye level, soft daylight from the front windows. Sound: espresso machine hiss, cups clinking, quiet street noise, no music.Each line names its speaker and stays short, which helps the lip sync. Swap the lines, but keep one sentence per person.
Bring a portrait photo to life
Animate my uploaded portrait. Keep the face, hair, clothing and background exactly as in the photo. The person blinks, takes a slow breath and turns their head slightly toward the camera with a small smile. A light breeze moves a few strands of hair. The camera stays still, the light does not change, realistic natural motion.Avoid: morphing face, changed eyes, warped teeth, flicker
Tiny movements keep the face stable. If it drifts, remove the head turn and keep only the blink and the breath.
Product orbit on wet black stone
Product shot of [product name] standing on a wet black stone slab. The camera orbits slowly 180 degrees around it at table height while fine water droplets run down its surface. One softbox light from above, deep black background, crisp reflections, premium commercial look. The label stays sharp and readable the whole time.Avoid: distorted label, misspelled text, extra objects, flicker
A slow orbit shows the shape without needing any action. Upload a product photo as reference so the label is copied, not invented.
Three references in one shot
Use my reference images: the woman from image 1, the red bicycle from image 2 and the canal street from image 3. Wide shot: she rides the bicycle along the canal at golden hour and rings the bell as she passes a flower stall. The camera tracks alongside her at the same speed. Warm low sun, long shadows, realistic film look.Numbering the images gives every reference a job. Use photos with plain backgrounds for the person and the bike.
Wok flames at a night market
Medium close-up at a busy night market stall. A cook tosses noodles in a wok and a short burst of flame rises from the pan, lighting his face orange. Slow push-in toward the wok, steam drifting through string lights in the background, shallow depth of field. Sound: roaring gas burner, sizzling oil, metal spatula on the wok, crowd chatter.The flame lights the face, which gives the shot drama for free. Sound lists from close to far make the mix feel natural.
Drone reveal of a lighthouse at sunrise
Aerial drone shot that starts low over a rocky shoreline and rises slowly to reveal a white lighthouse on a cliff at sunrise. Waves break white on the rocks below, sea mist hangs in the air. Smooth forward-and-up movement, wide lens, soft pink and gold sky, realistic travel film look.Starting low and rising is what makes a reveal feel earned. Replace the lighthouse with any landmark that sits high.
Your dog’s head tilt, animated
Animate my uploaded photo of my dog. Keep its fur color, markings, collar and the room exactly as they are. The dog tilts its head at the camera, its ears perk up, then its tail starts to wag. The camera stays still, soft daylight, natural animal movement.Three small actions in a clear order work well for pets. Use a photo where the whole dog, tail included, is in the frame.
Transition from a coffee cup to the night sky
Start on a close-up of a steaming coffee cup on a kitchen table at dawn. The camera pushes slowly into the dark surface of the coffee until it fills the frame, then the darkness opens into a starry night sky over a desert campsite with a small glowing tent. One continuous camera move with no hard cut, warm colors shifting to cool blue.Avoid: hard cut, flicker, text
Transitions work when a shape or color carries over: here the dark coffee becomes the sky. Try a puddle that becomes the ocean.
Slow-motion dunk in an empty gym
Slow-motion low-angle shot of a basketball player rising for a dunk in an empty school gym. Sweat drops fly off as the ball goes through and the net snaps. The camera tilts up to follow the jump. Hard overhead lights, high contrast, polished wooden floor. Sound: sneakers squeak, the rim clangs, a long echo through the empty hall.The low angle makes the jump feel higher. Say “slow motion” early, or Kling may play it at normal speed.
Rooftop dance at sunset
Full-body shot of a young man in a yellow tracksuit dancing on a rooftop at sunset, smooth hip-hop footwork that ends in a clean spin. The camera stays at chest height with a slight handheld sway. City skyline behind him, warm backlight outlining his shape. Sound: upbeat drum beat, sneakers scraping on gravel.A full-body frame keeps feet and hands in view, which is where dance clips break first. Name one finishing move.
Selfie-style product review
Vertical selfie-style video: a woman in her bedroom holds [product name] up to the phone camera and says, “Okay, this one surprised me.” She turns it so the label faces the lens. Handheld phone camera, natural window light, casual creator look. Sound: her voice, quiet room tone.One short line and one gesture keep it believable. If you use it in an ad, label it as AI where the platform asks for that.
Ladybug macro in morning dew
Macro shot of a ladybug climbing to the tip of a grass blade covered in morning dew, then opening its wings and flying off. The camera stays locked while the focus follows the ladybug. Soft green bokeh background, low sun backlight making the dew sparkle. Sound: birdsong, a faint breeze.A locked camera lets the tiny movement carry the clip. Swap in a bee on lavender or a snail on a leaf.
Anime hilltop run at sunset
Anime-style shot: a young woman with a messenger bag runs up a grassy hilltop at sunset, the wind lifting her scarf and blowing petals past her. The camera pans with her, then holds as she stops and looks down at the small town below. Hand-painted backgrounds, soft cel shading, warm orange sky.Naming the painted backgrounds and cel shading keeps the style flat and clean. Swap the town for a harbor or a train line.
Kling AI not following prompts? Fixes that help
Most misses come from prompts that ask for too much at once. Change one thing per retry so you know what made the difference.
- The action is skipped: cut the prompt to one main action and put it right after the shot type.
- Faces or hands warp: shorten the clip, slow the movement and add warped hands and morphing faces to the negative prompt.
- The camera wanders: name one camera move and add “smooth” or “steady”.
- Image to video changes your photo: begin with “Keep the face, clothing and background exactly as in the photo”.
- Speech is out of sync: shorten the line and let one person talk.
Kling AI negative prompts worth adding
Kling has a separate negative prompt box. Fill it with short items, not sentences, and leave out the word “no”, since the box already means avoid. For realistic clips a solid base is blurry, flicker, warped hands, extra fingers, morphing faces, text, watermark. Add cuts for long single takes and distorted label for products.
Keep the list short. Thirty items dilute the important ones, and no negative prompt fixes an action that is physically confusing. When a problem keeps coming back, rewrite the action instead of growing the list.
Do these prompts work in Veo or Seedance?
Mostly, yes, because the structure is plain English that other video models read the same way. Veo 3.1 and Gemini Omni also follow the sound line; for Seedance, Runway and Hailuo, leave it out. Seedance 2.5 is also announced for shots of up to 30 seconds, so the long takes carry over well. In Prompt Studio you can switch the model, and the wording and the negative list adjust to it.
Questions and answers
Is a Kling AI prompt different for image to video?
Yes. For text to video you describe the whole scene; for image to video the photo already sets the look, so you say what must stay, what moves and what the camera does. If Kling changes your photo, start with a keep sentence and ask for smaller movements.
How long can a Kling AI prompt be?
Kling’s prompt box has a length limit that has changed between versions, so trust the app over any number you read online. In practice 40 to 90 words is plenty for one shot; longer prompts tend to add details Kling ignores.
Is there a free Kling AI prompt generator?
Prompt Studio on LoschaZ is free and formats each prompt for Kling, with the negative list in its own field. It only writes the prompt: the video is generated in Kling under Kuaishou’s current plans and limits.
Which Kling AI camera movement prompts work well?
Simple named moves: slow push-in, pull back, tracking shot alongside the subject, orbit, crane up and aerial drone reveal. Add the speed, such as slow or steady, and use one move per shot.
Can Kling 4.0 make 30-second videos?
According to Kuaishou’s launch announcement on 15 October 2026, Kling 4.0 makes clips of up to 30 seconds in one pass. The lengths you can pick depend on your plan, and long clips need simpler actions than short ones.
Where can I find more Kling AI prompt examples?
Every prompt on this page is also in the Prompt Studio library, next to the other LoschaZ collections, and each one can be switched to Kling. The camera movement and image to video collections are the closest companions.
