VODOOHow to write an AI video prompt: a template that works on any engine
By VODOO AI · · 6 min read
Most AI video disappoints for one reason: the prompt left the engine guessing. "A coffee ad" gives an engine nothing to decide with, so it decides everything for you: the cup, the room, the light, the camera, the pace. Sometimes it guesses well. Usually you pay for three tries before it does.
A good prompt is not long or clever. It answers six questions a director would ask before shooting. Here is that template, the mistakes that quietly waste credits, a before-and-after, and prompts you can copy into any engine.
The six-part template
Answer these in order, one short phrase each. Together they make a prompt of 40 to 80 words, which is the range most engines read best.
- 01SubjectWho or what, specifically. Not a product, but a matte white ceramic mug with a chipped rim.
- 02ActionOne clear thing that happens. Steam rises, she turns, the door swings open.
- 03SettingWhere, and what is around it. An oak table by a rainy window in a small cafe.
- 04LightThe single biggest lever on mood. Soft morning daylight, golden hour, neon at night.
- 05CameraOne move and a framing. Slow push in, close-up, handheld at eye level.
- 06Feel and formatThe mood and where it runs. Calm and premium, vertical for Reels.
A matte white ceramic mug of black coffee on an oak table by a rain-streaked cafe window; steam curls up slowly. Soft grey morning light from the window, warm lamp glow behind. Slow push in from a close-up, shallow depth of field. Calm and quiet. Vertical 9:16.Make this in VODOO →
Before and after
A coffee ad
- After
- A barista pours latte art into a white cup on a marble counter, morning sun through the window, slow push in, warm and inviting, vertical
A woman walking in the city
- After
- A woman in a camel coat walks toward camera down a wet street at dusk, neon reflections, tracking shot at waist height, confident, wide 16:9
Show our sneakers
- After
- White leather sneakers step off a curb into a puddle in slow motion, water splashing up, low angle at ground level, bright overcast light, bold and energetic
| Before | After |
|---|---|
| A coffee ad | A barista pours latte art into a white cup on a marble counter, morning sun through the window, slow push in, warm and inviting, vertical |
| A woman walking in the city | A woman in a camel coat walks toward camera down a wet street at dusk, neon reflections, tracking shot at waist height, confident, wide 16:9 |
| Show our sneakers | White leather sneakers step off a curb into a puddle in slow motion, water splashing up, low angle at ground level, bright overcast light, bold and energetic |
Nothing on the right is fancy. Each one simply answers the six questions, so the engine spends its effort on making the shot instead of inventing it.
Five mistakes that waste credits
"No people", "without text"
- What happens
- Engines tend to draw what they are told to leave out
- Fix
- Describe what is there: an empty street, a plain wall
Several actions in one clip
- What happens
- They blur together, or only the first happens
- Fix
- One action per shot; cut a sequence from several shots
A sentence of on-screen text
- What happens
- Long words come out misspelled
- Fix
- Six words or fewer, or add the text as a caption after
Vertical and wide at once
- What happens
- The engine picks one, maybe not yours
- Fix
- Choose the format you need first
"8k, masterpiece, ultra realistic"
- What happens
- Little or nothing
- Fix
- Name the light and the lens instead
| Mistake | What happens | Fix |
|---|---|---|
| "No people", "without text" | Engines tend to draw what they are told to leave out | Describe what is there: an empty street, a plain wall |
| Several actions in one clip | They blur together, or only the first happens | One action per shot; cut a sequence from several shots |
| A sentence of on-screen text | Long words come out misspelled | Six words or fewer, or add the text as a caption after |
| Vertical and wide at once | The engine picks one, maybe not yours | Choose the format you need first |
| "8k, masterpiece, ultra realistic" | Little or nothing | Name the light and the lens instead |
Describe the motion, not just the scene
A still image prompt and a video prompt differ in one way: the video needs a verb. Say what moves, how fast, and how it ends. Strong verbs do most of the work: drifts, swings, settles, pours, lands, catches.
A silk scarf slips off a hook and drifts down through a sunbeam, turning slowly, and settles in a soft fold on a wooden floor. Dust in the light, quiet room. Static camera at floor level, shallow depth of field, slow motion.Make this in VODOO →
If you start from a photo, describe only the motion and what must stay the same. The picture already holds the look. See image to video AI for more.
Pick the first frame before you pay for the clip
Video engines charge by the second, and the first frame decides the look of the whole clip. A cheaper move is to make a few stills of the opening shot first, keep the best, and animate that one.
- 01Write the briefUse the template above.
- 02Make three stillsAs written, closer and wider: three framings of the same shot, for a fraction of a clip's price.
- 03Animate the one you likeThe clip now starts from a picture you chose, not one the engine guessed.
In VODOO this is one button: "Try three stills first" appears under any video brief in Create, priced next to what the clip would cost.
When you can't find the words
Most people know what they want long before they can describe it. If your brief is a few words, VODOO's "Help me describe it" asks the questions that kind of video needs, with answers to tap: how long, the mood, the light, where it runs. It then writes the full brief for you to edit.
Candle shop ad, 15 seconds, warm and inviting, candlelight, Instagram vertical. Becomes: a hand-poured soy candle in an amber glass jar flickers on a linen-draped shelf in a small shop at dusk; the camera pushes in slowly as the flame steadies, warm candlelight against cool blue window light. Inviting and calm, 15 seconds, vertical 9:16.Make this in VODOO →
Using these prompts in VODOO
Paste any prompt here into Create. VODOO picks the engine that suits the shot, shows the price before anything runs, and keeps track of which engines get it right first time. 300 free credits a month, no card. For engine-specific tips, see Kling AI prompts, Veo 3 prompts and Seedance prompts.
Questions people ask
- How long should an AI video prompt be?
- Usually 40 to 80 words. Long enough to name the subject, the action, the setting, the light, the camera and the mood; short enough that the engine reads all of it.
- What is the best AI video prompt template?
- Subject, action, setting, light, camera, and feel and format, in that order, one short phrase each. It works on every major engine because it answers what a director would decide before shooting.
- Why does my AI video not follow my prompt?
- Usually too much at once, or a negative like "no people". Keep one action per clip, describe what is there instead of what isn't, and choose one camera move.
- Should I write a different prompt for each engine?
- The template works everywhere. Some engines have strengths worth leaning on, such as Kling for weight and motion and Veo for sound, but the six parts stay the same.