Kling 4.0 Complete Guide

Davicho Barona
Davicho Barona

Kling 4.0 Prompting Guide

This guide gives you 15 reusable Kling 4.0 prompt templates, ranging from simple text-to-video generations to multi-reference video recreation and advanced 30-second sequences.

Kling's official guide recommends concrete visual descriptions instead of abstract language, clearly assigning roles to reference assets, and adding only the details that meaningfully affect the result. Omni Reference can combine images, videos, voice profiles, and elements, while Keyframes are designed for locking specific visual moments to the timeline.

Every example in this guide comes from one small project, The Case. Brennan swore it was a simple pickup. Then the alarms went off, the case started glowing, and Davicho found out how much of the plan Brennan had left out. Every take was made in Luma with Kling 4.0 Omni at 720p. Open the board to see every prompt, setting, and result: The Case tutorial board.

The opening still, made with GPT Image 2.5 Sunburst at 1K and high quality. Image 1 is Brennan's sheet, image 2 is Davicho's.
The opening still, made with GPT Image 2.5 Sunburst at 1K and high quality. Image 1 is Brennan's sheet, image 2 is Davicho's.
UI recording: two character sheets into one still with GPT Image 2.5 Sunburst. VO (optional): Select Brennan's sheet, then Davicho's. Open Create Image and choose GPT Image 2.5 Sunburst. Tell it what each image is for, set 16:9, 1K and high, and create.

Before You Start: The Universal Kling Formula

For most generations, think in this order:

SUBJECT → ENVIRONMENT → ACTION → CAMERA → VISUAL DETAILS → AUDIO

For advanced prompts:

ASSET ROLES → SETTINGS → TIMELINE → CAMERA → AUDIO → CONTINUITY

You don't need every category every time.

Start simple.

Add control only where you need it.

01. Simple Text-to-Video

USE WHEN

You have an idea but no reference assets.

TEMPLATE

[SUBJECT] is in [ENVIRONMENT].
They [PRIMARY ACTION], then [SECONDARY ACTION].
[SHOT SIZE + CAMERA MOVEMENT].
[LIGHTING + MATERIAL + ENVIRONMENTAL DETAILS].
Audio: [AMBIENCE / SFX / DIALOGUE].

EXAMPLE

Brennan, a night-shift guard with a thick mustache, stands alone at a warehouse loading dock at 2 AM. He hears a clank inside the bay and slowly turns toward the half-open roll-up door.

Handheld medium close-up slowly pushes toward his face.

Cold fluorescent tubes overhead contrast with the red rotating alarm light leaking under the door. Wet concrete reflects each sweep of red.

A refrigeration unit hums behind him. Distant highway traffic. No background music.

TIP

Don't use five adjectives where one physical description will work.

Instead of:

cinematic, premium, dramatic lighting

try:

Strong side backlight creates a narrow highlight around the subject while the background falls two stops darker.

Kling explicitly recommends concrete visual descriptions over abstract terms.

02. Image-to-Video / Character Animation

USE WHEN

You have one image and want to animate it while preserving the subject.

TEMPLATE

@Image1 defines [CHARACTER / SUBJECT].

Preserve [IDENTITY / CLOTHING / OBJECT DESIGN / ENVIRONMENT / COMPOSITION].

The subject [ACTION].

Camera [CAMERA MOTION].

Maintain [IMPORTANT CONTINUITY DETAILS] throughout.

Audio: [AUDIO].

EXAMPLE

@Image1 defines Brennan and the opening loading bay.

Preserve his facial features, mustache, short dark hair at the sides, pale yellow oxford shirt, the roll-up door, and the direction of the red alarm light.

He slowly turns toward screen-right after hearing the alarm behind him, pauses, then cautiously steps backward.

The camera performs a subtle handheld push-in.

Maintain the same character identity and wardrobe throughout.

03. Character Consistency Using Multiple Images

USE WHEN

Several independently referenced characters appear together.

TEMPLATE

@Image1 defines [CHARACTER A]. Preserve [IDENTITY DETAILS].

@Image2 defines [CHARACTER B]. Preserve [IDENTITY DETAILS].

Exactly [NUMBER] characters appear throughout the sequence.

[CHARACTER A] [ACTION].

[CHARACTER B] [ACTION].

Maintain their individual identities, clothing, body proportions, and positions. Do not interchange their appearances.

EXAMPLE

@Image1 defines Brennan. Preserve his facial features, bald head, thick mustache, body proportions, and clothing.

@Image2 defines Davicho. Preserve his facial features, dark curly hair, stubble, body proportions, and clothing.

Exactly two men appear throughout the scene.

Brennan runs toward screen-right carrying the glowing steel case.

Davicho follows several feet behind him.

Maintain both identities and wardrobes throughout every shot.

@Image1: Brennan. One sheet per person: full front without the head, full back, and a face close-up.
@Image1: Brennan. One sheet per person: full front without the head, full back, and a face close-up.
@Image2: Davicho, in the same three-panel layout.
@Image2: Davicho, in the same three-panel layout.
Take 1, shot 1 (0 to 3 s): the roll-up door bursts with sparks and they run.
Take 1, shot 1 (0 to 3 s): the roll-up door bursts with sparks and they run.
Take 1, shot 2 (3 to 6 s): both identities hold in the handheld medium shot.
Take 1, shot 2 (3 to 6 s): both identities hold in the handheld medium shot.
Take 1, shot 3 (6 to 10 s): behind the van, the case still glowing.
Take 1, shot 3 (6 to 10 s): behind the van, the case still glowing.
UI recording: Kling 4.0 Omni with two reference images. VO (optional): Select both sheets in order and open Kling 4.0 Omni. They load as image one and image two. Give each a role, time your shots, set the length, and create.

04. Character + Environment Reference

USE WHEN

One reference establishes the person and another establishes the world.

TEMPLATE

@Image1 defines the character. Preserve [DETAILS].

@Image2 defines the environment. Preserve [LAYOUT / ARCHITECTURE / LIGHTING / PROPS].

Place the character naturally inside the environment.

The character [ACTION].

Camera [CAMERA DIRECTION].

EXAMPLE

@Image1 defines Davicho. Preserve his face, curly hair, stubble, navy tee, light blue jeans, white sneakers, and body proportions.

@Image2 defines the loading bay. Preserve the roll-up door, pallet stacks, yellow bollards, red alarm lights, and wet concrete floor.

Place Davicho beside the parked white delivery van.

He edges along the van toward the open bay door, watching for movement inside.

A wide 28mm camera tracks sideways with him.

05. Reference an Action or Performance

USE WHEN

You want characters from your images to perform choreography from an existing video.

Kling specifically supports separating character appearance references from action references.

TEMPLATE

@Image1 defines Character A.

@Image2 defines Character B.

Use @Video1 only as the reference for [ACTION / CHOREOGRAPHY / PERFORMANCE / TIMING].

Replace the performers in @Video1 with the characters from @Image1 and @Image2.

Preserve: action order, body mechanics, interaction timing, direction of movement, relative character positions

Do not carry over [UNWANTED VIDEO ATTRIBUTES] from @Video1.

EXAMPLE

Use @Video1 only as the reference for the sprint choreography and escape timing.

Use @Image1 and @Image2 as the identity references for Brennan and Davicho.

Preserve the order of strides, the stumble, the slide behind cover, the spacing, and the timing from @Video1.

Do not copy the original runners' clothing, faces, or environment.

06. Reference Camera Movement

USE WHEN

You love the cinematography of an existing shot but want completely different content.

TEMPLATE

Use @Video1 only as the reference for camera movement, camera path, framing changes, and shot pacing.

@Image1 defines [SUBJECT].

@Image2 defines [ENVIRONMENT].

Replace the original subject with @Image1.

Replace the original environment with @Image2.

Preserve the camera's: starting position, movement direction, acceleration/deceleration, distance from subject, framing progression, timing

EXAMPLE

Use @Video1 only as the reference for camera path, movement speed, shot pacing, and framing transitions.

Replace the original vehicle with the white delivery van from @Image1.

Replace the environment with the empty industrial road outside the warehouse from @Image2.

Preserve the reference video's orbit, rapid pull-back, low pass, and final push-in.

Kling's official guide specifically demonstrates this type of camera-reference workflow.

07. Composition Reference

USE WHEN

You already have exactly the framing you want.

TEMPLATE

Use @Image1 as the composition reference.

Preserve: camera height, viewing angle, subject placement, foreground/background relationship, horizon position, negative space, approximate scale relationships

Replace [ELEMENT] with [NEW CONTENT].

The subject [ACTION].

EXAMPLE

Use @Image1 as the composition reference.

Preserve the low camera height, the centered roll-up door, the foreground pallet stack, the dark ceiling, and the position of the alarm light.

Place Brennan and Davicho from @Image2 and @Image3 in the doorway.

Both men sprint out of the bay toward camera.

08. Storyboard-to-Video

USE WHEN

You have storyboard panels or a storyboard sheet.

TEMPLATE

@Image1 defines [CHARACTER / STYLE].

Use @Image2 as the director-approved storyboard.

Follow the storyboard's: shot order, framing, camera paths, screen direction, action progression, character blocking, emotional progression

Generate the finished [LIVE ACTION / ANIMATION / STYLE] scene represented by the storyboard.

Do not render: storyboard panels, borders, annotations, arrows, labels, notes, sketch textures

TIP

Treat the storyboard as instructions, not as something that should appear in the output.

Kling's own guide recommends explicitly telling the model to reproduce the finished scene represented by a storyboard rather than recreating the storyboard itself.

09. Multi-Shot Cinematic Sequence

USE WHEN

You need precise pacing across several shots.

Kling 4.0 supports generations from 3–30 seconds, and multi-shot generation is enabled by default.

TEMPLATE

SCENE CONTEXT

Exactly [NUMBER] characters / subjects.

[LOCATION + CONTEXT]

Duration: [X seconds]

Visual style: [STYLE]

SHOT 1 — 0:00–0:03

Framing: [SHOT]

Camera: [CAMERA]

Action: [ACTION]

SHOT 2 — 0:03–0:06

Framing: [SHOT]

Camera: [CAMERA]

Action: [ACTION]

SHOT 3 — 0:06–0:09

[...]

CONTINUITY

Preserve [IDENTITY / DIRECTION / ENVIRONMENT / LIGHT] between all shots.

AUDIO

Dialogue: [...]

Ambience: [...]

SFX: [...]

Music: [...]

TIP

For complex sequences, give each shot one primary idea.

Don't ask Shot 2 to perform six separate narrative events in three seconds.

Take 1 above uses templates 03 and 09 together: two identity references and three timed shots inside one 10 second Kling 4.0 Omni generation.

10. One Continuous Long Take

USE WHEN

You absolutely do not want Kling to introduce cuts.

TEMPLATE

One continuous [DURATION] master shot.

No cuts, no angle changes, and no scene transitions.

All framing changes occur through physical camera movement.

The camera begins [STARTING POSITION].

It [MOVEMENT 1] as [ACTION].

Then [MOVEMENT 2] as [ACTION].

Finally it [MOVEMENT 3] to reveal [ENDING].

Maintain continuous geography, lighting, character identity, object permanence, and screen direction throughout.

EXAMPLE

One continuous 15-second master shot.

No cuts, no angle changes, and no scene transitions.

The camera begins three feet behind Brennan at shoulder height.

It tracks behind him as he slips through the side door, arcs clockwise around his left side as he enters the bay, then pushes past his shoulder toward the glowing case on the shelf.

All framing changes occur through physical camera movement.

11. First + Last Frame Transformation

USE WHEN

You know exactly how the shot starts and ends.

Kling's Keyframes workflow can use first frames, last frames, first-and-last frames, and intermediate keyframes.

TEMPLATE

@Image1 is the exact first frame.

@Image2 is the exact final frame.

Between them, show [TRANSFORMATION / ACTION / PROCESS].

0–X seconds: [STATE A → PROCESS]

X–Y seconds: [INTERMEDIATE STATE]

Y–END: [FINAL TRANSITION → STATE B]

The transformation must occur [PHYSICALLY / ORGANICALLY / MECHANICALLY / MAGICALLY].

Preserve [IDENTITY / COMPOSITION / OBJECT POSITION / BACKGROUND] where appropriate.

EXAMPLE

@Image1 is the exact first frame.

@Image2 is the exact final frame.

Show Brennan and Davicho sprinting out of the loading bay, ducking behind the van, and opening the glowing case.

The latches flip one at a time and the lid lifts, the light growing over their faces.

Avoid a simple dissolve between states.

The end frame for take 3, made with GPT Image 2.5 Sunburst from the opening still and both sheets.
The end frame for take 3, made with GPT Image 2.5 Sunburst from the opening still and both sheets.
Take 3 lands exactly on the end frame. With reference images attached, Kling treated the start frame as a strong reference rather than an exact first frame, so cut the first second if the opening has to match.
Take 3 lands exactly on the end frame. With reference images attached, Kling treated the start frame as a strong reference rather than an exact first frame, so cut the first second if the opening has to match.

12. Multiple-Keyframe Sequence

USE WHEN

You know several important visual moments and want Kling to construct the motion between them.

Kling 4.0 supports up to 10 keyframes.

TEMPLATE

@Image1 = keyframe at 0 seconds

@Image2 = keyframe at [X seconds]

@Image3 = keyframe at [Y seconds]

@Image4 = final keyframe at [Z seconds]

0–X seconds | Image1 → Image2

[Describe action + camera transition.]

X–Y seconds | Image2 → Image3

[Describe action + camera transition.]

Y–Z seconds | Image3 → Image4

[Describe action + camera transition.]

Maintain [IDENTITY / WARDROBE / ENVIRONMENT / SPATIAL CONTINUITY] between all keyframes.

BEST USES

Multiple keyframes are especially useful for controlling:

  • entrances
  • exits
  • transformations
  • major poses
  • camera-distance changes
  • environmental changes
  • story beats
  • composition changes

13. Video Editing / Targeted Change

USE WHEN

You have footage that is already mostly correct.

The operative word is:

EDIT.

TEMPLATE

Edit @Video1.

Change [TARGET] to [DESIRED RESULT].

Preserve: character identity, body movement, performance, background, camera movement, composition, lighting, shadows, duration, timing, audio

Keep everything else unchanged.

EXAMPLE

Edit @Video1.

Replace the dented steel case Brennan carries with a greasy cardboard pizza box, still glowing orange through the seams.

Preserve both men's faces, hair, and wardrobe, their movement and performance, the loading bay, the van, the camera movement, framing, lighting, shadows, timing, and the original audio.

Keep everything else unchanged.

Take 4: the steel case becomes a glowing pizza box. Faces, camera, timing, and audio come straight from take 1.
Take 4: the steel case becomes a glowing pizza box. Faces, camera, timing, and audio come straight from take 1.
UI recording: Kling 4.0 Omni in Edit mode. VO (optional): Select the take and open Kling 4.0 Omni. It loads as video one. Switch Video Mode from motion to edit, name the one change, list what stays, and create.

REMEMBER

“Reference @Video1” ≠ “Edit @Video1.”

Reference = learn from it.

Edit = modify the footage itself.

Kling says targeted video edits can modify characters, expressions, actions, cameras, styles, and backgrounds while preserving other parts of the footage.

In Luma, this is the Video Mode picker on Kling 4.0 Omni. Motion follows the reference video. Edit changes the reference video itself.

14. Character / Product Replacement + Creative Recreation

USE WHEN

You want to borrow the structure of an existing successful video.

TEMPLATE

Use @Video1 as the reference for: narrative pacing, shot order, shot duration, composition, camera movement, action timing, presentation structure

Replace [ORIGINAL SUBJECT] with @Image1.

Replace [ORIGINAL ENVIRONMENT] with @Image2.

Preserve the replacement subject's identity and design throughout.

Maintain the spatial relationships and timing of @Video1.

Do not carry over [UNWANTED CONTENT].

PRODUCT VERSION

Replace the product in @Video1 with the glowing steel case from @Image1.

Preserve the reference video's background composition, hand interaction, camera movements, lighting, shadows, shot sequence, pacing, and duration.

Keep the product from @Image1 visually consistent throughout.

WHY IT MATTERS

Kling calls this workflow Creative Recreation and specifically positions it for areas such as social content and performance advertising.

15. Dialogue + Native Audio Scene

USE WHEN

Dialogue, ambience, sound effects, or vocal performance matter.

TEMPLATE

CHARACTERS

@Image1 defines [CHARACTER A].

@Image2 defines [CHARACTER B].

ACTION

[Describe physical performance.]

DIALOGUE

[Character A], [VOICE / DELIVERY]:

“[EXACT LINE]”

[Character B], [VOICE / DELIVERY]:

“[EXACT LINE]”

AMBIENCE

[ENVIRONMENTAL AUDIO]

SFX

[SOUND EFFECTS]

MUSIC

[MUSIC DESCRIPTION / NO BACKGROUND MUSIC]

EXAMPLE

Davicho turns toward Brennan.

Davicho, American English, slightly out of breath, dry understated delivery:

“Please tell me that wasn't your plan.”

Brennan pauses before replying.

Brennan, lower male voice, calm and completely serious:

“It was most of the plan.”

Ambience: distant alarms, light wind, debris settling.

SFX: small pieces of metal hitting the pavement.

No background music.

Take 2: both lines land as written, in an 8 second Kling 4.0 Omni generation that starts from the opening still.
Take 2: both lines land as written, in an 8 second Kling 4.0 Omni generation that starts from the opening still.

Kling generates two-channel stereo and its guide recommends putting spoken dialogue in quotation marks when specifying language or speech. Direct audio input currently functions as a voice-profile reference rather than an uploaded soundtrack; ambience, effects, and music are generated from prompt direction.

BONUS: The “Director Prompt” Master Template

When the generation really matters, combine everything:

ASSET ROLES

@Image1 defines ___; preserve ___.

@Image2 defines ___; preserve ___.

@Video1 controls ___ only.

Do not inherit ___ from @Video1.

SCENE CONTEXT

Duration:

Aspect ratio:

Location:

Characters:

Visual medium:

SPATIAL MAP

Character A begins ___.

Character B begins ___.

The exit is ___.

Primary direction of movement is ___.

CONTINUITY

Exactly ___ characters.

Maintain ___.

The same ___ remains present throughout.

Never reverse screen direction.

SHOT PLAN

SHOT 1 — [TIME]

Framing:

Lens:

Camera:

Action:

SHOT 2 — [TIME]

Framing:

Lens:

Camera:

Action:

Continue as needed.

VISUAL CHARACTERISTICS

Lighting:

Color:

Materials:

Atmosphere:

Texture:

AUDIO

Dialogue:

Voice:

Ambience:

SFX:

Music:

EXCLUSIONS

No subtitles.

No background music.

Add additional exclusions only when truly necessary.

Five Rules That Fix Most Kling Prompts

1. Give every reference a job.

Don't just write:

@Image1 @Image2 @Video1

Write:

@Image1 = character. @Image2 = environment. @Video1 = camera motion only.

Asset-role clarity is one of Kling's strongest official recommendations.

2. Describe evidence, not adjectives.

Instead of:

expensive luxury hotel

try:

Floor-to-ceiling windows, polished dark stone floors, brushed brass fixtures, cream upholstered furniture, indirect warm architectural lighting.

3. Describe what SHOULD happen.

Instead of:

Don't show his face

try:

Keep the character in rear three-quarter view throughout.

Kling specifically advises minimizing negative instructions.

4. Scope your references.

The most powerful phrase in multimodal prompting may be:

“Reference ___ only.”

Use @Video1 for camera movement only.

Use @Image2 for wardrobe only.

Use @Image3 for composition only.

5. Start short. Earn complexity.

Kling's own guidance says longer prompts are not automatically better.

Start with:

subject + scene + action + emotion

Then add:

camera → timing → references → continuity → audio

only where additional control is actually necessary.

The Rule to Remember

Tell Kling what every asset means, what every subject does, where the camera goes, what must stay consistent, and what we should hear.

Everything else is optional.

The goal isn't to write the longest prompt.

The goal is to remove the right ambiguities.

See every prompt, setting, and result from this guide on The Case tutorial board.