Kling 3.0 Omni Prompt Guide: References and Multi-Shot Templates

The most useful Kling 3.0 Omni prompt is not the longest one. It is the prompt that gives each type of direction a clear job. Reference media should define identity or motion. The main prompt should define the scene. A storyboard should define the edit.

When the same instruction is repeated everywhere, the model receives noise instead of control. This guide gives you a clean structure for the Kling 3.0 Omni generator and copy-ready templates for common workflows.

A reliable five-part prompt

Write the main prompt in this order:

1. Subject: identify the person, product, creature, or environment.
2. Action: describe visible behavior with concrete verbs.
3. Setting: define the physical space and meaningful background activity.
4. Camera: state framing, movement, perspective, and lens character.
5. Light and finish: describe light direction, contrast, texture, and genre.

Template:

Example:

Prompting with a character Element

The Element images already describe the person's face and appearance. Do not spend half the prompt rewriting those details. Refer to the Element consistently and direct performance, scene, and camera.

Template:

Example:

Use two to four images that show the same appearance. Front, three-quarter, and side views are more useful than four nearly identical crops.

First-frame and last-frame prompts

When you upload role-based frames, acknowledge the transition without over-describing the images.

Template:

Example:

Avoid asking for a final composition that contradicts the uploaded last frame. The image is the stronger spatial constraint.

Feature video prompting

A feature video should describe movement or timing, while the prompt describes how that movement is reinterpreted.

Template:

Example:

Be specific about what to borrow. “Use this video” leaves the model to decide whether you mean subject, motion, camera, or setting.

Multi-shot prompt structure

For custom multi-shot, the top-level prompt defines global continuity. Each shot prompt defines only that beat.

Global prompt

Shot 1 · 4 seconds

Shot 2 · 6 seconds

The durations must add up to the total generation duration. On our launch preset, that total is ten seconds.

Three copy-ready storyboard templates

Product reveal: 3s + 4s + 3s

Global: Preserve the product from Element 1, including exact proportions, material, and visible branding. Controlled studio lighting and premium commercial finish.

Shot 1: Extreme macro detail moves across the product surface as condensation forms.

Shot 2: Match cut to a wider three-quarter view while the product rotates slowly above a reflective surface.

Shot 3: Settle into a centered pack shot. A narrow highlight reveals the logo and then stops.

Character hook: 5s + 5s

Global: Preserve the person from Element 1 and the same location, wardrobe, weather, and time of day.

Shot 1: Medium rear tracking shot as the character walks through a crowded market, then suddenly stops.

Shot 2: Cut around to a close-up as they turn toward the camera with a subtle realization. Background motion continues naturally.

Travel transition: 3s + 3s + 4s

Global: One traveler, one continuous journey, natural documentary camera, realistic motion.

Shot 1: Wide shot as the traveler runs toward a departing ferry.

Shot 2: Tight handheld detail of their hand catching the rail while the ferry begins moving.

Shot 3: Cut to a calm profile on deck as the city recedes behind them, camera slowly drifting outward.

Common mistakes

Contradictory camera instructions

“Locked camera” and “fast orbit” cannot both be true for one shot. Assign one camera behavior to each beat.

Too many identity descriptions

If an Element defines the person, focus the prompt on action and continuity. Repeated physical descriptions may introduce accidental variation.

Shot prompts that repeat the global prompt

Keep global constraints at the top. Shot fields should be short enough to scan as an edit decision.

Reference overload

Use either ordinary image URLs or role-based images for one task, not both. Choose the representation that matches your intent. A first frame, Element, and feature video can be powerful together, but only when each supplies different information.

Asking silent output to perform dialogue

The free launch preset has no audio. You can direct visible speaking or reaction, but do not evaluate it as generated dialogue. Native audio will be a credits setting when that system launches.

A final pre-generation checklist

• Is the subject defined by text or by a reference, but not ambiguously by both?
• Does every uploaded image have a clear role?
• If using an Element, are all views the same subject and appearance?
• If using feature video, did you say what motion to borrow?
• Do custom shot durations total ten seconds?
• Does each shot advance the sequence?
• Are camera instructions physically compatible?
• Is the desired output compatible with 720P and no audio?

Start with a two-shot sequence before attempting six beats. Once the continuity works, add complexity deliberately. You can put these templates to work in the Kling 3.0 Omni AI Video Generator, or compare the workflow in our Kling 3.0 Omni vs Kling 3.0 guide.