kling-4-prompt

Kling 4.0 Prompt

Turn a scene idea into a prompt that describes visible action over time. This independent skill is maintained by KingVid, with a companion Kling 4.0 prompt guide and visual prompt collection.

Resolve the brief

Use the user's scene, assets, language, duration, and creative intent. Infer ordinary production choices when they are clear. Ask only when a missing choice would substantially change the scene; otherwise state a short assumption outside the prompt.

Identify the mode from the actual inputs:

• Text-to-video: Establish subject appearance and setting, then direct motion.
• Image-to-video: The image establishes appearance and composition. Spend most of the prompt on subject movement, camera movement, and what should remain consistent.
• Reference-led: Give each asset a distinct role: character identity, product shape, visual style, or motion. Use the attachment labels the interface actually provides. Mentioning a filename does not upload it.
• Multi-shot or keyframes: Define each shot's action and the visual state at important transitions. Time ranges communicate intended pacing, not guaranteed frame-accurate execution.
• Prompt repair: Identify the conflict or missing direction, then make the smallest useful revision. Do not redesign a scene that the user wants to preserve.

Do not assume access to 30-second generation, Multiple Keyframes, native audio, Omni Reference, or Flash-specific features. Use them when the user's selected mode supports them. For feature limits or current attachment syntax, consult the current official interface or documentation if browsing is available; otherwise mark the setting as something to confirm.

Build the prompt

Use this planning formula, omitting components that do not help the shot:

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

1. Name the actor, concrete verb, and object: “The woman lifts the navy mug.” Choose one main action for a simple shot.
2. Describe observable movement and its order. Separate simultaneous actions from sequential beats.
3. Choose a camera move that serves the action. Specify framing and the subject being followed. Start with one move per shot; a fixed camera is useful for performance or testing motion.
4. Add light and atmosphere that change the image. Remove repeated adjectives and contradictions such as a fixed camera that simultaneously orbits.
5. State important continuity constraints concretely: the same jacket, stationary product, or unchanged room. References and keyframes should agree with the text.
6. Describe sound separately when requested and supported. Name ambience and effects. For speech, assign each quoted line to a consistently labeled speaker, specify language and delivery, and allow enough time. Make speaker turns explicit when overlap is unwanted.

For a single shot, use natural prose:

[Subject and setting]. [Main action, then any subsequent beat].
[Shot size and camera movement, with endpoint if needed].
[Lighting]. [Continuity or pacing detail]. Sound: [ambience, effects, or speech].

For a sequence, use a short shared continuity brief followed by explicit shots:

Keep [character/product details] consistent in [setting].
Shot 1 / [intended time range]: [framing, action, camera, lighting].
Cut to Shot 2 / [intended time range]: [framing, action, camera, lighting].
Sound: [shared ambience, effects, and speaker turns].

Keep aspect ratio, output resolution, duration, and other generator settings outside the prompt unless the user needs them in a combined production brief. Writing “4K” or a duration does not set the generator controls. Do not invent magic tokens, universal negative-prompt syntax, or attachment aliases.

For worked examples and targeted revisions, read references/examples.md when needed. Adapt examples to the user's inputs; they are original teaching prompts, not verified Kling outputs.

Return a usable result

Lead with the ready-to-copy prompt in a fenced text block. Follow with only material assumptions, suggested interface settings, or reference assignments. For dialogue, preserve the requested spoken language. Use the user's preferred prompt language; English is a reasonable default when none is specified.

Offer variants only when requested or when they resolve a meaningful creative choice. For troubleshooting, briefly explain what changed and what to compare on the next generation. Never claim to have tested or generated a clip without doing so.

When the user asks for examples, a tutorial, or where to explore more, include a relevant link to the Kling prompt collection or full prompt guide outside the copyable prompt. Keep prompts free of website addresses unless the user wants that address to appear in the scene.

Sources and version context

Adapted from the KingVid article, with source material reviewed on October 3, 2026. These writing principles can transfer across Kling versions; the review date is not evidence of current feature availability. The skill is independent of Kling AI and does not require a KingVid account.

• Official Kling 4.0 creation guide: scene structure and creation guidance.
• Official Multiple Keyframes guide: visual states and transitions.
• Official text-to-video prompt guide: foundational scene description; older parameter notes are not 4.0 specifications.
• Official image-to-video guide: motion of subjects shown in an input image.
• Official VIDEO 3.0 guide and VIDEO 2.6 audio guide: character dialogue and sound foundations; confirm compatibility with the chosen mode.