Guides

How to write prompts for Kling

A Kling prompt is a scene description built on the formula subject, motion, scene, camera, lighting and atmosphere, and in newer versions it also carries asset roles, timestamps and an audio block. Kling responds better to plain sentences and one clear action than to long lists of adjectives.

Updated

Open Kling

The core formula

The official Kling guide builds a prompt from six parts. The first three are essential and the rest add control. Write short sentences and keep the action small enough to happen within the length of the clip.

  • Subject: who is in the frame, appearance, clothing, pose
  • Motion: what the subject does, one simple action
  • Scene: where it happens, foreground and background
  • Camera: shot size, angle, lens and movement
  • Lighting: morning light, sunset, light rays through fog, neon
  • Atmosphere: foggy night, quiet morning, tension

Camera language and camera movement

Kling separates camera language from camera movement. Camera language covers framing and optics: close up, wide angle, low angle, shallow depth of field. Movement covers how the camera travels: horizontal, vertical, zoom, pan, tilt, roll.

Give the camera one specific command instead of a vague "cinematic". Combined moves, such as pulling back while rising, should be spelled out and tested on a short clip.

  • "Close up, the camera slowly pushes in on the hero’s face"
  • "Low angle, the camera rises up along the building facade"
  • "Wide shot, the camera smoothly orbits the hero from left to right"
  • "First person view, the camera moves forward down a corridor"

Start and end frames

This mode builds a transition between two uploaded images. Kling recommends frames on the same theme that are as similar as possible, because a big difference causes a shot change instead of a smooth transition.

In the prompt, describe what happens between the frames, not what is in them. For example: "The woman picks up the cup from the table and raises it to her lips, static camera".

Motion Control: transferring movement

Motion Control takes your photo plus a reference video with movement and makes the person in the photo repeat that movement. The output length matches the reference. These are the developer’s requirements for the inputs.

  • A reference video of 3 to 30 seconds, one continuous shot with no cuts and no camera movement
  • One person in the reference, with the whole body and head clearly visible
  • Moderate, steady motion. Real people work best
  • The photo framing matches the video: half body to half body, full body to full body
  • Enough space around the person in the photo for big movements
  • By default orientation follows the video. If you choose to follow the image instead, you can set camera movement in the prompt

Kling 3.0 and 4.0: shots, audio and assets

Kling 3.0 supports multi shot: the model can plan the cuts itself, or in custom mode you set the number of shots and the duration of each. Label shots ("Shot 1", "Shot 2") and pace events with times such as "at the 4th second". Tie each line of dialogue to a character and add a tone.

In Kling 4.0 assets are numbered in upload order, and each needs a role such as appearance, action pacing or camera path. The guide suggests four blocks: asset roles, overall settings, a timestamped sequence of events and audio. It discourages negative instructions, except for turning off subtitles and background music.

  • "Image 1 sets the hero, image 2 sets the kitchen. From 0 to 4 seconds the hero walks in and puts a bag on the table. From 4 to 8 seconds the camera moves closer and the hero smiles. Audio: footsteps, rustling bag, no music"
  • "Shot 1: wide shot, a barista sets a cup on the counter. Shot 2: close up, steam rising from the cup. The barista says calmly: “Your cappuccino”"

Prompt mistakes in Kling

Most failed clips come from an overloaded prompt rather than from the model itself.

  • An action with no subject. The model fills the gap and often returns a static clip
  • Hard physics such as a bouncing ball or a throw, which Kling itself calls difficult
  • Exact counts. Kling struggles to keep the number of objects in a frame
  • A description far from the source photo, which triggers a cut or shot change
  • Long lists of things to avoid instead of a description of what you want

How to do it in Solistra

The Solistra catalog includes Kling 4.0, Kling 3.0, Kling 3.0 Omni, Kling Motion Control and other versions. Duration, resolution and price are shown on each model card.

  • Open Video and choose the Kling version you need
  • Upload a photo, a pair of frames or a reference video for Motion Control
  • Write the prompt using the formula and pick a camera preset
  • Check the credit price on the button and run it

Frequently asked questions

What language should I write Kling prompts in?

Solistra accepts prompts in Russian or English. For spoken lines, Kling 3.0 speaks Chinese, English, Japanese, Korean and Spanish, and translates dialogue in other languages into English.

How long are Kling videos?

Kling 3.0 generates 3 to 15 seconds and Kling 4.0 generates 3 to 30 seconds in a single run.

What is Kling Motion Control?

A mode that transfers a person’s movement from a reference video onto the person in your photo. The reference should be one continuous shot with one person, framed full body or half body.

Do I need a negative prompt for Kling?

The Kling 4.0 guide recommends describing what you want instead of what to avoid. The exception is asking to drop subtitles or background music.

Sources

  1. Text to Video prompt guide · Kling AI
  2. AI Camera Control guide · Kling AI
  3. Start and end frames · Kling AI
  4. Motion Control user guide · Kling AI
  5. Kling VIDEO 3.0 model user guide · Kling AI
  6. Kling 4.0 model user guide · Kling AI