How to Prompt Kling for Realistic Video (2026)

Realistic Kling video comes from one clear structure, not from piling on adjectives. Describe a single subject doing one action in a defined place, shot with one camera move under specific light — and let duration and aspect ratio stay as panel settings. This guide breaks down the six-part formula, shows when to start from a still, and gives 10 paste-ready prompts.

The 6-part Kling formula

Realistic shots follow this order: Subject + Action + Setting + Camera + Lighting + Mood. Keep each part tight and give the model exactly one thing to focus on for motion and one for the lens.

1. Subject. Name who or what, with a few concrete visual details. Example fragment: "a weathered fisherman in a yellow oilskin coat, grey stubble."

2. Action — one main action. Choose a single verb. Stacking actions is the top cause of warping. Example fragment: "slowly hauling a rope hand over hand."

3. Setting / place. Ground the subject in a real location. Example fragment: "on the wet deck of a small trawler, open sea behind him."

4. Camera — shot size + one movement. Pick a framing and a single move. Example fragment: "medium shot, slow dolly in." You can name a start and end point, like "pan from the horizon to his face."

5. Lighting. Be specific; vague light reads as flat and fake. Example fragment: "overcast morning light, soft and cool with faint rim light on the coat."

6. Mood. One or two words steer color and pacing. Example fragment: "tense, weathered, documentary."

Put together: "A weathered fisherman in a yellow oilskin coat slowly hauls a rope hand over hand on the wet deck of a small trawler, open sea behind him. Medium shot, slow dolly in. Overcast morning light, soft and cool. Tense, documentary mood." That is the whole prompt — no duration, no aspect ratio, no extra clauses.

For faster iteration, keep the formula in a reusable structure. Our Kling prompt cheat sheet and Kling prompt templates lay out fill-in-the-blank versions of each part.

Text-to-video vs image-to-video

For realism, prefer image-to-video whenever you can. Image-to-video is Kling's strongest mode because its 3D face and body reconstruction locks the subject's structure, which sharply reduces warping and morphing.

Use text-to-video when you have no usable reference frame, when you need full creative freedom over the subject, or for quick concept tests. Write the same six-part prompt either way.

Use image-to-video when a person's face, a product, or a brand element must stay consistent. Feed a clean, well-lit still, then describe only the motion and camera in the prompt — the image already carries the look. This is the reliable path to photoreal people. See Kling image-to-video prompts for reference-driven examples, and if you also work in Veo, compare with how to prompt Veo for realistic video.

Extra tools sharpen both modes: Motion Brush to move a specific region, Start & End Frame to control the arc of a shot, Elements to combine up to four reference images, and Video Extension to continue past the length cap.

10 example prompts

Each prompt below is complete and paste-ready. Drop it into Kling 3.5, set duration and aspect ratio in the panel, and add the negative prompt from the next section.

1. Fisherman on deck

A weathered fisherman in a yellow oilskin coat with grey stubble slowly hauls a rope hand over hand on the wet deck of a small trawler, open sea behind him. Medium shot, slow dolly in. Overcast morning light, soft and cool with faint rim light on the coat. Tense, weathered, documentary mood.

Why it works: one action (hauling), one move (dolly in), and specific overcast light give realistic skin and fabric. (16:9, 1080p, 10s)

2. Barista pouring latte art

A young barista in a black apron carefully pours steamed milk into a ceramic cup, forming a leaf pattern, at a marble café counter with a chrome espresso machine behind. Close-up on the cup and hands, static camera. Warm window light from the left, soft shadows. Calm, artisanal mood.

Why it works: a static camera plus one delicate action keeps liquid physics believable and avoids hand warping. (9:16, 1080p, 5s)

3. Runner at dawn

A woman in grey running gear jogs along an empty coastal boardwalk, ocean to her right. Wide tracking shot moving alongside her. Golden dawn light with long shadows and gentle lens flare. Fresh, energetic, cinematic mood.

Why it works: a single tracking move matched to the subject's motion reads smooth and real. (16:9, 4K, 10s)

4. Chef flipping vegetables

A chef in a white jacket tosses sliced vegetables in a steel wok over a gas flame in a busy restaurant kitchen. Medium close-up, slight handheld shake. Warm tungsten light with a bright flame flare. Fast, appetizing, energetic mood.

Why it works: handheld adds life without a second move, and one flame source anchors the lighting. (16:9, 1080p, 5s)

5. Vintage car on a mountain road

A red 1960s convertible drives along a winding mountain road with pine forest on both sides. Drone shot ascending to reveal the valley. Late-afternoon sun, warm and hazy with soft atmosphere. Nostalgic, freeing, cinematic mood.

Why it works: one ascending drone move creates scale while the car does the only action. (21:9, 4K, 10s)

6. Rain on a city window

Rain streaks slide down a café window with a blurred neon street beyond, a coffee cup steaming on the sill in the foreground. Static close-up, shallow depth of field. Cool blue night light with warm neon reflections. Moody, quiet, introspective mood.

Why it works: a static shot and shallow focus let the rain and steam carry realistic micro-motion. (16:9, 1080p, 5s)

7. Product hero: sneaker turntable

A white leather sneaker sits on a matte grey pedestal in a studio. Slow orbit around the shoe. Soft, even softbox lighting with a subtle gradient background. Clean, premium, minimal mood.

Why it works: a single slow orbit on a still subject is the safest realistic product move. Start from a product photo for best results. (1:1, 4K, 10s)

8. Child blowing bubbles

A small child in a striped shirt blows a stream of soap bubbles in a sunlit backyard, green grass and a wooden fence behind. Medium shot, slow dolly out. Bright afternoon sunlight, warm and soft with bokeh. Joyful, tender, nostalgic mood.

Why it works: one gentle action and a single dolly-out keep the face stable and expressive. (16:9, 1080p, 5s)

9. Blacksmith at the forge

A muscular blacksmith hammers a glowing orange blade on an anvil in a dark stone workshop, sparks flying. Medium close-up, static camera. Firelight from the forge, deep shadows with strong orange glow on the face. Intense, gritty, dramatic mood.

Why it works: firelight as the only source plus one repeating action produces convincing sparks and skin. (16:9, 4K, 10s)

10. Astronaut looking out a window

An astronaut in a white suit slowly turns to look out a spacecraft window at Earth below. Medium shot, slow push in toward the visor. Cool blue Earthlight on the helmet with soft interior fill. Awe, calm, cinematic mood.

Why it works: a slow push in and a single subtle turn read as weightless and real. (16:9, 4K, 15s)

Negative prompts that keep it clean

A negative prompt lists the artifacts you never want, and it noticeably raises realism. Add it in the negative field, not the main prompt.

warping, morphing, distorted face, extra fingers, deformed hands, flicker, jitter, plastic skin, waxy texture, over-saturation, blurry, low resolution, watermark, text, duplicate limbs, unnatural motion

Trim the list to what matters for your shot. For faces and hands, keep warping, morphing, extra fingers, plastic skin. For landscapes, flicker and over-saturation matter more. Pair this with a clean starting still for the best result. For more ready-made shots, browse the best Kling prompts roundup.

Mistakes to avoid

Most unrealistic results trace back to a handful of habits. Fix these first.

Stacking multiple actions. "Walks, waves, and sits down" forces the model to guess timing and it warps. Pick one action per shot.

Stacking multiple camera moves. A dolly plus an orbit plus a zoom creates drift and jitter. Use one move; name a start and end if you need direction.

Writing durations and aspect ratios in the prompt. "10 seconds, 16:9" belongs in the panel settings. In the text it wastes tokens and can confuse the model. There are no --ar flags in Kling.

Vague lighting. "Nice lighting" reads flat. Name the source, direction, and quality — "soft window light from the left."

No negative prompt. Skipping it invites the usual face and hand artifacts. Always add a short one.

Too-long paragraphs. Dense prose buries the main action. Keep it one to three short sentences following the six-part order.

For finished cinematic setups built on these rules, see our Kling cinematic b-roll prompts.