- Section: Code assistant
Description
_(no description)_
Downloaded resources
62-higgsfields-image-video-prompts.pdf— Ultimate Prompting Guide for Video Generation Models.pdf (ok (150041 bytes))
The Definitive Guide to Prompting Video Generation Models for Kling 0.1, Kling 2.6, Veo 3.1, and LTX Pro
This guide teaches you how to write prompts that work with all major video models. The goal is simple: tell the model exactly what you want, in the order it needs to understand it, without overwhelming it.
- Start with the Core Idea (1–2 sentences)
Start your prompt by helping the model understand the overall context:
● What’s happening?
● Where is it happening?
● What’s the mood of the scene?
Example: “A wide aerial shot of a Lamborghini racing through a busy city at sunset, pursued by police cars with flashing lights.”
This sets the scene so the model understands the world, tone, and action.
- Describe Camera Movement Clearly
Video models pay more attention to camera movement than anything else. Always define:
● Starting position (aerial, close-up, over the shoulder) ● Movement path (glide forward, rotate, zoom in, orbit)
● Final position (inside the car, front angle, behind the character)
Example: “The camera follows the car from above, then glides forward and gently pushes through the windshield into the driver’s cabin.”
Camera instructions make videos feel intentional—not random.
- Define Characters with Simple, Strong Details
Keep character descriptions short but clear:
● age range
● gender
● overall look
● expression or emotion
● what they’re doing
Example: “Inside the car is an attractive man in his late 20s, holding the steering wheel with calm confidence. He briefly looks at the camera.”
Models don’t need essays — they need simple visual cues.
- Add a Secondary Character or Object (If Needed)
When showing more than one person, keep the details minimal:
Example: “In the passenger seat, a beautiful blonde model sits beside him, watching the road.”
Avoid describing faces in too much detail. These models have difficulty when you specify too many human features.
- Build the World Around the Scene
Describe only the parts that matter:
● time of day
● setting
● atmosphere
● lighting
● weather (if it contributes to the mood)
Example: “The city glows with neon signs and warm sunset light reflecting off the car’s surface.”
This gives the scene personality without cluttering it.
- Explain the Lighting and Mood
Lighting is an important cue for realism. Keep it simple:
● soft cinematic lighting
● golden-hour glow
● dramatic shadows ● bright neon reflections
Example: “Cinematic lighting with warm highlights and deep shadows gives the sequence a dramatic, high-budget look.”
- Define Each Character's Movement
Small actions work best:
● turn their head
● look at the camera
● smile
● grip the steering wheel
● lean slightly
Example: “The man turns his head toward the camera, then looks at the model in the passenger seat.”
This adds life without overwhelming the model.
- Keep Visual Style Instructions Simple
If you want a specific look, choose ONE of these and keep it clean:
● realistic
● cinematic
● warm, film-like ● stylized anime
● smooth commercial style
● dramatic action style
Example: “The video should look cinematic and realistic, with a smooth, high-budget film style.”
Don’t stack too many styles — it confuses any model.
- Set the Tone for Movement and Energy
Tell the model how intense the movement should be:
● slow and elegant
● fast and intense
● steady camera
● shaky action camera
● drone-style tracking
Example: “The movement is smooth and fast-paced, emphasizing the intensity of the chase.”
This helps the model choose the right type of movement.
- Avoid these common prompt mistakes
❌ Too many adjectives ❌ Trying to describe every facial detail ❌ Providing three conflicting styles ❌ Overloading with camera instructions ❌ Mixing atmospheres ("dark and dramatic, but bright and cheerful") ❌ Describing actions that require multiple cuts Video models excel at continuous movement, not edited sequences.
- The perfect prompt structure
(Universal template) Here’s a structure you can share with your community—simple, effective, and works with all models:
🎬 Universal video prompt template [Scene Setup] One clear sentence explaining the main idea and atmosphere.
[Camera Movement] Tell the model where the camera starts, how it moves, and where it ends.
[Characters] Briefly describe each character: appearance, emotion, and small actions.
[Environment] Define the setting: city, forest, desert, futuristic interior, etc.
[Lighting & Style] Choose a visual style and describe the lighting simply.
[Motion Tone] Explain whether the scene is smooth, fast, intense, or calm.
- Example using the template
(This matches your Lamborghini scene.)
“A wide aerial shot of a bright yellow Lamborghini racing through a downtown grid at dusk, chased by police cars with flashing lights. The camera tracks the car from above, then glides forward and seamlessly pushes through the windshield into the cabin. Inside is a handsome man in his late 20s, gripping the wheel with calm confidence and briefly looking toward the camera. In the passenger seat sits a beautiful blonde model, focused on the road. The city glows with warm dusk light and neon reflections. The scene is cinematic, fast-paced, and smooth, with dramatic lighting and high-detail realism.”
- Model-specific tips
Kling 0.1 (older version) ● Keep prompts short.
● Too many details cause flickering.
· A camera movement works better.
Kling 2.6 (newer version) ● Can handle longer prompts and complex camera movements.
● Excellent with faces and movement.
● Responds well to cinematic lighting instructions.
Veo 3.1 ● Very strong at realism and soft details.
● Keep the prompt balanced—not too long.
● Excellent with warm lighting and natural movement.
LTX Pro ● Ideal for stylized and slightly dramatic styles.
● Keep scene descriptions clean.
● Good for fast-moving action sequences.