Kling 4.0 Prompt Guide: Formula, Templates & Tested Examples

Start with what would make the video usable: space for an advertising headline, a reaction after a line of dialogue, or enough action to carry a 30-second story to its ending. Kling 4.0 follows those decisions when you write them down, and makes its own when you don’t.

In our perfume test, an open-ended brief produced changing close views of a faceted bottle. A specific brief produced the complete bottle on the left of the frame, with dark space on the right for copy.

This Kling 4.0 prompt guide gives you the formula, two timing templates, and prompts we ran on the JXP Kling 4.0 studio on October 29, 2026. Each example includes its settings, what we observed, and a link to the original video.

Kling 4.0 prompt cheat sheet

• Formula: Subject + Action + Setting + Camera + Lighting + Pacing + Sound.

• Write on a clock: assign actions or shots to time ranges that fit the selected duration.

• Give references a job: say what each asset controls and what to keep.

• Describe what you want to see: favor positive descriptions, and explicitly exclude subtitles or music when you don’t want them.

• Define the ending: say what the final frame should look like.

• Review the output: compare the result with your brief, then fix the biggest mismatch first.

Want to test these prompts right away?

Try Kling 4.0 on JXP

What Kling 4.0 Changes About Prompt Writing

Kling’s Kling 4.0 launch guide lists 3–30 second generation, prompts up to 8,000 tokens, up to 10 keyframes, up to 15 reference assets, two-channel stereo audio, and output up to 4K. Longer clips need beats rather than one action, references need assigned roles, and the soundtrack needs direction.

Provider settings can differ from the model’s maximums. At the time of our tests, the JXP studio offered Text, Reference, and Keyframes modes, 720p or 1080p, 16:9, 9:16, 1:1, or 21:9, and 3–30 seconds.

The Official Kling 4.0 Prompt Formula

Step 4 of Kling’s launch guide recommends writing the prompt like shot direction, built from seven elements:

Subject + Action + Setting + Camera + Lighting + Pacing + Sound

Kling’s Kling 4.0 user guide arranges a prompt in four blocks: asset roles, overall settings, a time-stamped sequence of events, and audio. Use the seven parts to decide what to say, and the four blocks to put it in order.

Four habits make the formula work:

• Use visible details. The user guide recommends replacing terms like “premium look” with lighting, materials, and props. It also warns that negative instructions can increase errors.

• Keep names consistent. Describe a character once, then reuse the same words.

• Treat timestamps as direction. In our timed examples, the main events landed in or near their windows, but not always exactly.

• Separate camera from subject movement. A camera orbiting a still bottle is a different brief from a bottle on a turntable.

Choose the Right Workflow Before You Write

Kling’s user guide says Omni Reference and Keyframes cannot be combined in one task. The JXP studio reflects this with separate modes.

Reference prompts. Uploads are numbered in order (@Image1, @Video1). Without a role, the model guesses what an asset is for:

@Image1 defines the presenter. Preserve the face, hair, and jacket.
@Image2 is the bottle. Keep the label exactly as shown.
@Video1 sets the handheld camera rhythm only. Do not copy its location.

We did not test multiple references for this article; treat this as a template.

Keyframe prompts. In the JXP studio, Keyframes mode takes a first frame, up to eight middle frames, and an optional last frame. The images fix the moments; the prompt describes the motion between them. Describe the change, not the picture you uploaded.

Templates: 15-Second and 30-Second Prompts

Both templates follow the user guide’s order: settings, then timed events, then audio. They are planning templates, not tested generations.

15-second template

[Aspect ratio]. [Subject and setting]. [Lighting and materials].

0-5s: [Opening framing and action]. [One camera behavior].
5-11s: [Main action]. [Camera movement that follows it].
11-15s: [Payoff]. End on [specific composition]. Hold.

Audio: [Ambience and effects]. [No background music / music cue]. No subtitles.

30-second template

[Aspect ratio]. [Location, time of day, light].
Character: [identifying details]. Keep [them] the same in every beat.

0-6s: [Setup]. [Shot size, camera behavior].
6-15s: [Action or change]. [Camera movement].
15-24s: [Development]. Restate the character's key details.
24-30s: [Resolution]. End on [specific expression or composition]. Hold.

Pacing: [how the energy rises and falls].
Audio: [Ambience, timed effects, dialogue, music cues]. No subtitles.

Use one framing and one camera behavior per time range, and state whether the windows continue one take or start new shots. In our structured café test, the model added a cut we had not asked for.

Camera, Dialogue, and Sound

Dialogue. Kling’s advice is to describe expression, movement, and tone with the words. Label each speaker, quote the line, and give physical actions their own moment. Write other languages in that language and name the accent in the label. To check the result, listen: mouth movement alone cannot confirm the words.

Sound and text. Give ambience, effects, and music separate sentences, and say which matters most. In our dialogue test, the audio level between the active sections was very low, even though we asked for fridge hum and rain. Quote any on-screen text exactly and say where it sits.

Kling 4.0 Prompt Examples We Tested

How we tested: clips were generated with Kling 4.0 in the JXP studio on October 29, 2026, at 720p, with our own prompts and no prompt assistance. We kept the first completed output per brief (one task was retried after a service error, unchanged). We reviewed sampled frames across every clip. For examples 3, 5, 6, and 7 and the café pair, we also measured the audio level over time; examples 1, 2, and 4 have no audio analysis. Generation times were measured from submission to completion, checked about every 45 seconds. We did not listen-check speech, transcribe it, or score lip-sync.

1. Product Ad with Space for Copy

Settings: Text to Video · 16:9 · 5s · 720p · 25 credits

One continuous 5-second product shot of a single rectangular amber-glass perfume bottle with a matte black cube cap on wet black stone. A small white front label reads "LUMA". Keep the complete bottle on the left third of the frame throughout, leaving the right half empty and dark for advertising copy. Camera slowly pushes straight in, keeping the cap and base visible. Warm rim light from behind on the left, a soft reflection across the glass, a thin layer of drifting mist near the stone. The bottle remains stationary. Hold the final composition for the last second. Audio: quiet room tone. No background music. No subtitles.

Observed: at the inspected checkpoints, the rectangular bottle, black cap, and readable LUMA label stayed on the left, with dark space on the right. An earlier open-ended version (“A cinematic premium perfume advertisement. An amber perfume bottle on black stone. Beautiful lighting, dramatic camera movement, luxury mood.”) produced a faceted bottle and changing, cropped close views. Several instructions changed at once, so the pair shows the value of specificity, not the effect of one phrase.

2. Animate a Product from a Start Frame

Settings: Keyframes · one start frame (a still from example 1) · Adaptive aspect ratio · 5s · 720p · 25 credits

Use the uploaded start frame as the opening composition. One continuous 5-second shot. Keep the amber bottle, black cap, LUMA label, wet stone, and left-side placement unchanged. The camera stays locked. A thin ribbon of mist drifts slowly from left to right across the stone below the bottle. A narrow warm highlight travels across the glass from left to right, then settles. The bottle remains still and fully visible. Keep the right half dark and empty. Audio: quiet room tone. No background music. No subtitles.

Observed: the bottle position, label, and dark right side remained recognizable while mist moved across the stone. Our first submission with a fixed aspect ratio was rejected; Adaptive worked without changing the prompt.

3. Ultrawide Brand Film with Timed Beats

Settings: Text to Video · 21:9 · 15s · 720p · 75 credits

21:9. Stereo: water lapping against wooden hulls, gulls far off, pier planks creaking. No subtitles.

0-5s: Dawn over a small fishing harbor. Wide locked-off shot, wooden boats rocking gently on gray-blue water, mist on the surface.
5-11s: Tracking shot along the pier, following a young fisher in a yellow rain jacket as she carries a blue plastic crate.
11-15s: Close-up as she sets the crate down and smiles at the catch. Soft strings enter as the sun clears the horizon behind her.

Observed: the file came back at 1460×624. At the sampled checkpoints, the fisher entered the pier at about six seconds, the crate went down at about 12 seconds, and the smile appeared at about 14.7 seconds with the sun on the horizon. In the frames we checked, we also observed a close-up that briefly cropped her head.

4. Three-Shot Story with Hard Cuts

Settings: Text to Video · 16:9 · 15s · 720p · 75 credits

A 15-second three-shot miniature clay animation in a rainy village, with hard cuts at 5 and 10 seconds. Keep the same small yellow robot with one blue eye, red mailbox, and clay materials in every shot. Shot 1, 0-5s: wide locked shot, the robot walks up the path carrying a white envelope and stops beside the red mailbox. Shot 2, 5-10s: close-up of its hands sliding the envelope completely through the mailbox slot. Shot 3, 10-15s: medium shot, the robot steps back, a warm window light turns on behind it, and it raises one hand in a small wave. Audio: rain, soft mechanical footsteps, mailbox click. No dialogue. No background music. No subtitles.

Observed: the wide approach, delivery close-up, and medium ending appeared in order, with view changes around five and ten seconds at the sampled checkpoints. The envelope went through the slot, the window lit up, and the robot waved.

5. Two-Character Dialogue

Settings: Text to Video · 16:9 · 10s · 720p · 50 credits

16:9. A small kitchen at night, warm overhead light. Stereo: fridge hum, rain on the window. No subtitles. No background music.

Character A: an older man with gray hair in a brown cardigan, holding a white mug. Character B: his adult daughter with dark curly hair in a navy sweater.

The man sets down his mug and looks at his daughter.
[Character A: Older man, quiet voice, almost smiling]: "You were right about the job."
Immediately, the daughter laughs once, surprised, and looks away.
[Character B: Daughter, soft surprised voice]: "I usually am, Dad."

Medium two-shot, locked camera, slow push-in on the last line.

Observed: both characters matched their descriptions at the sampled checkpoints, and the man set the mug down at about 3.5 seconds. The audio showed several separate sections of sound activity; we did not listen-check who spoke or what was said. The push-in was barely visible.

6. UGC Testimonial

Settings: Text to Video · 9:16 · 10s · 720p · 50 credits

9:16. Handheld selfie framing, slight natural shake. A woman in her twenties with a messy bun and a gray hoodie sits in the driver's seat of a parked car, daylight through the windshield. No subtitles. No background music.

She holds a small white tube of sunscreen up to the camera and says in a relaxed, slightly amused voice, in American English: "Third time reapplying today, and my face still isn't greasy."
Then she squeezes a little onto the back of her hand and shows it to the camera.

Sound: muffled traffic outside, the click of the cap closing.

Observed: the file came back at 720×1280. She speaks to camera holding the tube, then squeezes product onto her hand and shows it to the lens, in the order written. We saw no obvious hand problems in the sampled frames. The cap looks closed in one frame and open in the next.

7. A Full 30-Second Story

Settings: Text to Video · 16:9 · 30s · 720p · 150 credits

16:9. A rainy city street at night. Stereo: steady rain, tires on wet asphalt, distant traffic. No subtitles.
Character: a cyclist in her twenties in a bright yellow rain jacket and a white helmet. Keep her the same in every beat.

0-6s: She waits at a red light, one foot on the curb. Close-up on her face, breath visible in the cold air. Locked camera.
6-15s: The light turns green and she pushes off. Tracking shot alongside her through slow traffic, neon reflections streaking across the wet road.
15-24s: The same cyclist in the yellow jacket turns into a narrow alley. The camera rises slowly above her, revealing rooftops and glowing signs.
24-30s: She stops at a green wooden door, takes off her helmet, and looks up at the camera. A low synth pulse that has been building drops out. Hold on her face.
Pacing: tense, then fast, then a quiet held ending.

Observed: the yellow jacket stayed consistent at all 12 sampled checkpoints, and the white helmet stayed consistent until she took it off. The tracking shot, the high view over the alley, and the ending (helmet off at about 27 seconds, then a look into the lens) appeared as written. The audio level rose through the fast section and dropped for the ending, a couple of seconds early. The opening asked for a face close-up and showed a wider shot.

Before vs. After: A Café Prompt

We ran one idea twice with the same settings (16:9, 10s, 720p, 50 credits each).

Before:

A woman drinking coffee in a beautiful cafe, cinematic, premium look, masterpiece, best quality, 4k, no blur, no ugly hands.

After:

16:9. A small corner cafe on a rainy morning. Stereo: rain on the window, a low espresso machine hiss, quiet chatter. No subtitles. No background music.

0-4s: Close-up of a woman in her thirties with short black hair and a mustard knit sweater, both hands around a white ceramic cup. Steam rises from the coffee. Locked camera.
4-8s: She takes a slow sip and looks out the rain-streaked window. Slow push-in to her face. Soft gray window light from the left, warm tungsten light from behind.
8-10s: She sets the cup down on its saucer with a quiet ceramic clink. Hold on her small smile.

Both files came back at 1280×720, so resolution followed the studio setting. Neither output showed obvious hand problems. As with the perfume pair, several instructions changed together.

Run your own comparison:

open the Kling 4.0 video generator

Common Kling 4.0 Prompt Problems and Fixes

Negative prompts. The JXP studio had no separate negative prompt field. Following Kling’s user guide, describe what you want, and add “no subtitles” or “no background music” when you need them out.

Kling 4.0 Prompt Guide FAQ

How long should a Kling 4.0 prompt be?

Long enough to settle the decisions that matter. Our structured prompts ran about 480 to 900 characters, well under the 8,000-token limit.

Do I need timestamps in every prompt?

No. A single short action can work without them. Use time ranges when the order or timing of events matters.

How long does a Kling 4.0 generation take?

In our JXP tests at 720p, the 10- and 15-second clips we timed finished within about four to five minutes of submission, and the 30-second clip within about seven and a half minutes. We checked status about every 45 seconds, so these are approximate, and queue conditions can change them.

How much does a Kling 4.0 video cost on JXP?

At the time of our tests, generation without a reference video cost 5 credits per second at 720p and 7 at 1080p. With a reference video, the rate is 4 or 6 credits per second, applied to the output length plus the input video length. Check the displayed cost before submitting.

Can I use these prompts with Kling 4.0 Flash?

Reuse the ideas after checking Flash’s duration, modes, and settings. We did not test Flash.

Start Writing Better Kling 4.0 Prompts

Decide what the clip is for, split it into timed beats, give each beat one action and one camera behavior, define the ending, and direct the sound. Then compare the output with your brief and revise the biggest mismatch.

Ready to direct your first 30-second clip?

Start creating with Kling 4.0