Check your video prompt. Fix it in seconds.
Pick your AI video tool and paste any prompt. The doctor checks it against the 10 golden rules, highlights the words causing trouble, and writes a corrected version you can copy straight into Veo, Sora, Runway, Kling or Manus.
- Choose your tool
- Paste your prompt
- Get the fixed version
The 10 golden rules
1. Name your subject precisely
Video models fill vague words with random guesses. Material, color, size and brand details keep the subject the same from the first frame to the last.
Instead of: a product on a table
. Write: a white ceramic coffee mug with a matte glaze and a thin gold rim
.
2. Give it one clear action
A clip lasts 5–8 seconds. One action fits; three actions turn into a blur of half-finished movements.
Instead of: she walks in, sits down, drinks her coffee, then waves goodbye
. Write: she lifts the cup slowly and takes a first sip
.
3. One place, one time of day
Two locations or two times of day in one shot force the model to morph between them or pick one at random.
Instead of: a desert and a neon city, at sunset and at night
. Write: a rain-slicked neon city street at night
.
4. Make the light match the time
Light sets the mood more than any other word. It has to agree with the time of day, or the shot looks fake.
Instead of: midday sun with moonlight and neon glow
. Write: warm low-angle backlight during golden hour
.
5. Choose one camera move
Models handle one smooth move well. Stacking pans, zooms and orbits causes jitter and warping. Name the shot size too (wide, medium or close-up), so the model knows how close to start.
Instead of: pan left, then zoom in, then orbit around her
. Write: slow push-in toward the subject, ending on a close-up
.
6. Describe the lens, not "cinematic"
Lens and focus words like 85mm, shallow depth of field or anamorphic change the look far more than "cinematic, 8k, masterpiece" ever will.
Instead of: cinematic, 8k, ultra HD, masterpiece, award-winning
. Write: shot on an 85mm lens, shallow depth of field, focus locked on the subject
.
7. Keep it 5–8 seconds
One move reads best in 5–8 seconds. Longer clips drift, and many tools cap a single clip around 5–10 seconds. Generate 8 seconds, then extend.
Instead of: 15 seconds with a whole story arc
. Write: 8 seconds, one continuous shot, no cuts
.
8. Frame it for where it will play
For 9:16 reels, keep the subject in the middle third so captions and app buttons never cover it.
Instead of: (no format at all)
. Write: 9:16 vertical, subject in the middle third, top and bottom 20% kept clear
.
9. Write the exact words
Veo and Sora generate sound and speech. Give the exact line in quotes, about 2.5 words per second, or the model invents its own. With no line, still name the sound: ambience, music or effects.
Instead of: a voiceover about the product
. Write: voiceover says: "Your morning, made slower."
.
10. Keep text out of the frame
Video models still warp letters: logos melt and signs turn into made-up words. Keep the frame clean, leave room for the text, and add the logo or title in your editor, where it stays sharp and on-brand.
Instead of: a coffee cup with our logo and "Good morning" written on it
. Write: a plain white coffee cup, subject offset to one side, leaving clean space for a headline or logo
.