Lesson 5. The art of writing video prompts — structure and technique
Writing video prompts is fundamentally different from writing image prompts. Video has the dimension of time: you must direct movement, the camera, and the pacing. In this lesson, you will master the professional structure of a video prompt.
Topic breakdown
The main difference between a video prompt and an image prompt is the dimension of time. In an image, you describe a single moment. In video, it's a sequence: what happens, the motion, where the camera looks, how long it lasts. This requires a completely different approach.
Professional video prompt structure: 1) Subject — who or what; 2) Action — what they are doing; 3) Environment — where; 4) Camera — angle and movement; 5) Lighting/Mood — emotion; 6) Technical specs — quality, style, duration.
The biggest mistake is writing it like an image prompt. 'Beautiful sunset over mountains, golden hour' — good for a photo, insufficient for video. For video, you need: 'Camera slowly pans across mountain range during golden hour, clouds drifting, birds flying in distance' — you must include motion.
Another critical rule: simplicity. AI video models currently create 5-10 second clips. In this timeframe, show 1-2 distinct actions. 'A person walks, then sits, then opens a book' is too much. 'A person reads a book, slowly turning the page' is perfect.
What you'll learn
- Know the 6 elements of video prompt structure
- Describe action clearly and concisely
- Apply camera commands correctly
- Convey mood and atmosphere through words
- Understand the difference between photo and video prompts
- Adapt your prompt for different tools
Deep dive
The professional video prompt formula: [Shot Type] + [Subject] + [Action] + [Environment] + [Camera Movement] + [Lighting/Mood] + [Technical Parameters]. Example: 'Medium shot of a barista pouring latte art in a cozy cafe, camera slowly zooms in, warm morning light through window, cinematic color grading, 5 seconds'. This formula always works.
Adapting prompts for different platforms: Kling understands short, keyword-heavy prompts well (like Midjourney). Runway accepts more detailed, cinematic descriptions. Google Veo understands natural language — you can write as if speaking to a human. Pika responds well to very simple prompts.
Camera command vocabulary: pan left/right (horizontal turn), tilt up/down (vertical), dolly in/out (pushing in/pulling out), orbit (rotating around), crane up/down (vertical lift/drop), handheld (shaky effect), static. Knowing these words lets you direct the video like a filmmaker.
Mood and atmosphere words: cinematic, dreamy, energetic, melancholic, dramatic, peaceful, nostalgic, futuristic, cozy, dark, ethereal, raw. Each gives the AI a different color palette, lighting, and pacing. 'Cinematic' yields high contrast, 'Dreamy' makes it soft and slightly blurred.
Ready video prompt template
Copy and adaptClose-up of hands typing on a mechanical keyboard, shallow depth of field, soft blue LED backlight reflecting off the keys, smooth camera drift to the right, lo-fi aesthetic, 5 seconds
Why it works
Shot type: 'Close-up of hands' — clearly specifies the scale of the shot
Action: 'typing on a mechanical keyboard' — simple, repetitive action — easy for AI
Visuals: 'shallow depth of field, soft blue LED backlight' — elements of a professional look
Camera: 'smooth camera drift to the right' — slow, minimal camera movement
Duration: '5 seconds' — specific timing
Practice
- Write a single scene in 3 ways: 1) subject only, 2) subject + action, 3) all 6 elements — compare the results
- In Runway Camera Control, write a scene from 5 different camera angles and generate the videos
- In Kling, change the 'mood' words (dramatic vs peaceful vs energetic) — analyze how this affects the output
- For your specific niche (business, education, creative), write 5 video prompts and select the best one
Common mistakes
- Writing a prompt without movement — instead of 'beautiful city at night' write 'camera glides through city streets at night'
- Too many actions — 5 seconds = 1-2 actions optimally; 3+ decreases quality
- Forgetting to specify the camera — without a camera command, the AI decides, and the result is unpredictable
- Failing to specify duration — write '5 seconds' or '10 seconds', otherwise the platform uses its default
Lesson FAQ
Is it mandatory to write video prompts in English?
Yes, currently all major tools (Kling, Runway, Veo, Pika) understand English prompts best. Prompts in other languages degrade the quality of the result.
How long should the prompt be?
Optimally 15-40 words. Too short (5 words) and the AI improvises heavily. Too long (60+ words) and the AI gets confused. 2-3 sentences is ideal.
Will the exact same prompt yield the same result every time?
No. AI video generation is a stochastic process; it will vary slightly every time. If you need an exact result, use a seed (if available) and image-to-video.