Guide for all skill levels

How to write a AI video prompt

A step-by-step guide with examples. Learn to write prompts that produce stunning results.

💡 Tip: write in your own language!

On Sixio you don't need to know English. Just describe the idea in any language, and the AI will do the rest:

  • Will enhance — turns a simple description into a professional cinematic script
  • Will translate — auto-translates to English for the AI
  • Adapts — matches the length and style to the chosen model (VEO, SORA, Kling, etc.)

5 elements of a good prompt

Every video prompt has 5 key elements. You don't have to use all of them — but the more details, the better the result.

1

Scene (Where?)

Setting and atmosphere

"An abandoned factory, moonlight streams through the broken windows"
2

Subject (Who/What?)

Main subject in frame

"A young woman in a red dress, dark hair flowing in the wind"
3

Action (What is happening?)

Motion and dynamics

"Walks slowly along the wall, touching it with her fingertips, turns toward the camera"
4

Style (How does it look?)

Visual aesthetic and mood

"Cinematic, dark color palette, high-contrast film-noir lighting"
5

Camera (What angle?)

Camera movement and shot type

"Camera slowly pushes in, medium shot transitions to close-up, shallow depth of field"

Examples: bad prompt → good prompt

See how AI enhancement turns simple descriptions into detailed cinematic scripts.

Nature
✗ BADYour input

«A beautiful sunset»

AI enhances ↓
✓ GOODAfter AI enhancement

«A golden sunset over a Mediterranean coast. The sun touches the horizon, painting the sky in warm orange-pink tones. Waves gently roll onto the sandy shore, reflecting the last rays. The camera slowly rises, revealing the panorama. A light breeze sways the grass on the cliff. Ambient sound of the surf and seagull cries.»

Cinema
✗ BADYour input

«A person walks down the street»

AI enhances ↓
✓ GOODAfter AI enhancement

«A young man in a long dark coat walks through nighttime Tokyo. Neon signs reflect off the wet asphalt after the rain. The camera follows him at shoulder height, slightly swaying. Passers-by are blurred in the background, focus on the hero. Cinematic color grading in cool blue tones. City sounds: distant car hum, footsteps splashing in puddles.»

Advertising
✗ BADYour input

«Coffee ad»

AI enhances ↓
✓ GOODAfter AI enhancement

«Close-up: hot espresso pouring into a white ceramic cup. Thick caramel crema forms on the surface. The camera slowly pulls back, revealing a cozy café with warm morning light. Coffee beans scattered on a wooden table. Steam rises in the rays of sunlight. Soft jazz soundtrack.»

Sci-Fi
✗ BADYour input

«Space»

AI enhances ↓
✓ GOODAfter AI enhancement

«The camera flies through Saturn's rings. Billions of icy particles sparkle in the light of the distant Sun. The giant planet fills half the frame, its banded atmosphere slowly rotating. Milky Way stars in the background. Epic orchestral music. Cinematic sci-fi quality, IMAX style.»

HappyHorse 1.0 — detailed guide

HappyHorse 1.0 — the 2026 premium flagship. Tolerates DENSE detailed prompts, but density must do real work. Use the 6-block structure.

6-block prompt structure

  1. 1. Scene and time — WHERE and WHEN.
  2. 2. Subject — who/what is in the frame: scale, pose, gaze.
  3. 3. Action and motion — what moves and by how much (movement budget).
  4. 4. Camera language — lens, DOF, camera movement, angle.
  5. 5. Light and texture — direction, quality, time of day.
  6. 6. Audio — ambient, foley, dialogue in quotes.
Photorealistic mode

Add the word "photorealistic" + "pores", "fabric wear", "available light". Anti-cues against AI-look: "no glamorization".

Movement budget

Meter: "no more than 5% push-in", "slight breathing". Keeps the frame away from AI-drift.

R2V — character1..N

Up to 9 references. In the prompt: "character1 jogs through forest. character2 floats behind her".

Native joint audio+video

Describe sound in the prompt. Lip-sync supports many languages — do NOT translate dialogue.

Full HappyHorse 1.0 guide

Tips for every model

Ready-made prompt templates

Copy the template and replace the data in [brackets] with your own. AI enhancement will polish the rest.

🎬 Cinematic scene

[Setting], [time of day]. [Character with appearance details] [performs action]. Camera [movement type], [shot size]. [Lighting style], [color palette]. Atmosphere of [mood].

Example:

An abandoned metro station at night. A girl in a white dress stands at the edge of the platform, looking into the dark tunnel. The camera slowly pushes in, medium shot. Flickering neon light, cool blue tones. An atmosphere of mystery and solitude.

📱 Product ad

Close-up: [product] on [surface/background]. [Action with product]. Camera [movement], revealing [context]. [Lighting]. [Brand colors]. [Ad style].

Example:

Close-up: a perfume bottle on black marble. A water drop rolls down the glass. The camera pulls back, revealing a luxurious interior. Soft side lighting, golden highlights. Minimalism, luxury style.

🌍 Nature and landscape

[Location], [time of day/season]. [Natural elements] [their movement]. Camera [type: drone/panorama/timelapse]. [Lighting]. [Nature sounds].

Example:

Norwegian fjords, dawn in June. Mist hangs over mirror-still water, mountains reflected on its surface. A drone flies low above the water. Golden rays pierce through the clouds. Silence, the splash of water, a distant bird call.

🎵 Music video

[Performer/character] in [location]. [Movement/dance]. [Visual effects]. Camera [dynamic movement]. [Color palette]. [Music genre] video style.

Example:

A dancer in an empty warehouse with high ceilings. Contemporary dance, smooth arm movements. Dust particles in the spotlight beams. The camera orbits around her. High-contrast shadows, warm orange tones. Contemporary music video style.

How AI prompt enhancement works on Sixio

On Sixio you don't have to be a prompt expert. Our AI enhancement system turns any plain-language description into a professional cinematic scenario:

1️⃣

You write

A simple description: "a kitten playing with a ball of yarn"

2️⃣

AI enhances

Adds details: lighting, camera, textures, atmosphere, sound

3️⃣

Translates

Auto-translates to English and adapts to the model

Frequently asked questions

How do I write a prompt for AI video generation?
A good AI-video prompt has 5 elements: 1) Scene — where the action happens, 2) Subject — who or what is in the frame, 3) Action — what is happening, 4) Style — visual aesthetic, 5) Camera — camera motion and angle. Write in Russian — the AI will automatically translate and enhance the description.
Do I need to write the prompt in English?
Нет! На Sixio можно писать промпты на русском языке. AI автоматически переводит их на английский перед генерацией. Более того, AI улучшает простое описание до профессионального кинематографического сценария. Вы также можете использовать кнопку "Перевести" для ручного перевода.
How long should the prompt be?
Depends on the model. VEO 3.1 and SORA 2 work better with long, detailed prompts (200–500 words). WAN 2.5 is limited to 750 characters in Russian. Grok Imagine works well with short prompts. AI enhancement automatically adapts the length to the chosen model.
How do I improve generation quality?
Используйте AI-улучшение промпта — оно добавляет профессиональные детали: освещение, движение камеры, текстуры, атмосферу. Также указывайте конкретные детали вместо абстрактных: не "красивый закат", а "золотой закат над морем, солнце касается горизонта, тёплые оранжево-розовые тона".
How do AI prompt-enhancement models differ?
On Sixio there are 3 prompt-enhancement models: ChatGPT (fast, cheap), Gemini Pro (highest quality, more expensive), and Grok (experimental, cheap). All three create cinematic descriptions, but Gemini Pro gives the most detailed and precise result. See current prices on the 'Pricing' page.
Can I use my own prompt without AI enhancement?
Да! Вы можете отправить свой промпт на генерацию напрямую, без AI-улучшения. Это полезно, если вы опытный пользователь и точно знаете, какое описание нужно. Просто нажмите "Перевести" вместо "Улучшить", или введите промпт на английском.
Which prompts work best for different models?
VEO 3.1 — cinematic scenes with sound and dialogue details. SORA 2 — realistic physics and timing. WAN 2.6 — detailed descriptions with minimal censorship. Kling 3.0 — multimodal scenes with audio. Grok — short, vivid descriptions. LTX-2 — technical details, 4K style. Seedance — budget descriptions with aspect ratio specified.
How do I write a prompt for image-to-video?
Для Image-to-Video промпт описывает действие, которое должно произойти с объектами на фото. Не описывайте сам объект (модель его видит), описывайте движение: "девушка поворачивает голову и улыбается", "камера медленно отъезжает, открывая панораму". Для Kling Motion загрузите референсное видео с нужными движениями.

Ready to create your first video?

Free credits on sign-up — try it right now

Start generating

We use cookies to operate the service, keep your session, and collect anonymous statistics. See our Privacy Policy.