Write professional cinematic prompts for the Kling AI video generation model.
Core Principles
Prompt engineering comes after directing/screenwriting/editing — first determine what story the shot should tell, then write the prompt. A good-looking shot without drama is just wallpaper.
Kling Model Differences
Kling 3.0 (pro/standard)
- Native multi-shot: generate up to 6 consecutive shots at once
- Native voiceover + lip sync:
[Character A: 台词内容] - Timbre control: you can specify a character's voice style
- Up to 15 seconds of continuous output
- Better suited for: multi-shot narratives with dialogue and lip sync
Kling 2.6 Pro
- Element Binding: precise control over the movement of elements in the frame
- Motion Brush: specify motion direction and intensity element by element
- Motion Transfer: use a reference video to drive a still image
- Dedicated negative prompt field
- Better suited for: character consistency across multiple social media videos (no dialogue), Motion Brush, lower cost
Kling V2.5 Turbo
- Fast generation
- Suitable for rapid prototyping and low-cost batch generation
Prompt Structure
Text-to-Video Template
[主体描述] — 角色的外貌、服装、动作
[场景描述] — 环境、时间、氛围
[运动描述] — 主体的运动方向和方式
[镜头描述] — 景别、角度、运动
[灯光氛围] — 光源方向、色温、情绪
Example:
一位穿红色汉服的年轻女子，长发飘逸。
场景：黄昏时分的古寺庭院，银杏叶飘落。
运动：她缓缓转身，衣袂随风飘动，落叶环绕。
镜头：中景 → 缓慢推近至特写，低角度仰拍。
灯光：金色逆光透过银杏叶，暖色调，柔和的丁达尔效应。
Multi-Shot (Kling 3.0) Template
[Character A: 角色名] — 外貌、服装、性格关键词
镜头1 [广角，平拍]: 环境建立 — 地点、时间、氛围
镜头2 [中景，跟拍]: 角色动作 — 具体行为和表情
镜头3 [特写，慢动作]: 关键细节 — 情绪高潮
Example:
[Character A: 小白] 白色长毛猫，琥珀色眼睛，优雅。
镜头1 [广角，静止]: 雪后的日式庭院，清晨薄雾。
镜头2 [中景，水平跟拍]: 小白从走廊跳下，在雪地上留下脚印，尾巴竖起。
镜头3 [特写，慢动作60fps]: 小白的爪子踩入雪中，雪花飞溅，阳光折射微光。
Image-to-Video Template
Based on the reference image, describe onlythe parts that should move：
参考图提供：角色外观和场景。
提示词只描述运动：
— 主体运动：头发被风吹起，眼睛缓缓眨动
— 环境运动：背景烛光微微摇曳，窗帘轻摆
— 镜头运动：从特写缓慢拉远至中景
Motion Brush (Kling 2.6 Pro) Description
元素运动清单：
— [前景主体]: 向前行走，速度为步行
— [背景树木]: 轻微左右摇摆，幅度小
— [水面]: 涟漪向外扩散，频率慢
— [云层]: 从左向右缓慢移动
Negative Prompts (Kling 2.6 Pro)
负向提示词：画面抖动、变形、闪烁、模糊、不自然的面部扭曲、手指变形、文字乱码
Camera Language Reference
| Shot Size | English | Use |
|---|---|---|
| Extreme Long Shot | Extreme Long Shot | Establishing environment |
| Long Shot | Long Shot | Full character + relationship to environment |
| Medium Shot | Medium Shot | Character action display |
| Close Shot | Close-up | Facial expression |
| Close-Up | Extreme Close-up | Emphasis on key details |
| Camera Movement | Effect |
|---|---|
| Dolly In | Progressive focus |
| Dolly Out | Reveal the environment |
| Tracking | Follow the subject |
| Pan | Horizontal sweep |
| Tilt | Vertical sweep |
| Handheld | Realism |
Lighting Design
| Style | Descriptors |
|---|---|
| Golden Hour | warm golden backlight, long shadows, soft diffusion |
| Blue Hour | cool blue ambient, twilight sky, city lights |
| Rembrandt Lighting | Rembrandt lighting, dramatic chiaroscuro, triangle patch |
| Neon Aesthetic | neon glow, colored rim light, wet street reflections |
| Natural Light | diffused window light, overcast soft shadows, practical light |
Rhythm and Duration
- 5 seconds: single action, simple emotional expression
- 10 seconds: action + reaction, short dialogue
- 15 seconds: complete narrative arc (Kling 3.0 multi-shot)
Quick checklist
- Does the prompt include specific motion descriptions? (not just a static image)
- Is the shot language clear? (shot size, angle, movement)
- Is the lighting direction specified? (light source, color temperature)
- Do multiple shots have character consistency anchors?
- Is the Motion Brush specifying motion element by element?
- Are the negative prompts filled in (Kling 2.6 Pro)?
Reference
- Kling official documentation: https://app.klingai.com
- smixs/visual-skills: https://github.com/smixs/visual-skills