Reusable AI-video prompt template
Give every instruction one job.
A useful AI-video prompt is easier to test when subject, action, camera, timing and preservation rules are separated instead of compressed into one cinematic sentence.
The copy-ready template
SUBJECT: [Who or what is visible, including the few identity details that must remain stable.]
SETTING: [Location, background and lighting state.]
ACTION: [One observable movement with a clear start and end state.]
CAMERA: [Shot size, angle, lens feel and one camera behavior.]
TIMING: [Clip length and when the action occurs.]
FRAMING: [Aspect ratio, subject placement and any required negative space.]
CONTINUITY: [Features, geometry, wardrobe, labels, counts or scale that must not change.]
NEGATIVE DIRECTION: [A short list of the failures most likely to ruin this shot.]
How to fill it without overloading the model
1. Describe only what affects the shot
Identity needs a few stable anchors, not a biography. For a person, that may be hair, wardrobe and one facial feature. For a product, it may be shape, material, cap, label orientation and scale. Add a detail only when you would reject a result for changing it.
2. Choose one primary action
“Moves naturally” is not an observable event. Replace it with a physical change: a wrist rotates until a label faces camera; a subject turns their head once; a droplet travels down one side of a bottle. If the action needs two clauses joined by “and”, consider a second shot.
3. Keep camera and subject motion separate
A slow dolly toward a stationary subject is different from a subject walking toward a locked camera. Writing them as separate fields exposes conflicts before generation. One primary camera verb is usually easier to evaluate than a stack of orbit, zoom, pan and handheld direction.
4. Turn timing into an interval
Words such as “slowly” and “quickly” are relative. A short interval makes the desired sequence testable: hold for one second, rotate over two seconds, then settle for the final second. Keep the schedule realistic for the clip length.
5. Preserve the success criteria
Continuity direction is not decoration. It names what must survive every frame: the same face, the same clothing, stable product geometry, one object, unchanged logo orientation or consistent background architecture.
Filled example: vertical product hero shot
SUBJECT: A clear rectangular perfume bottle with a matte black cap and centered cream label, standing on dark stone.
SETTING: Minimal charcoal studio background, soft rim light from camera left, controlled reflection beneath the bottle.
ACTION: The bottle remains stationary while one narrow light sweep travels from left to right across the glass.
CAMERA: 85 mm close-up at label height, locked tripod, no orbit, no zoom and no focus breathing.
TIMING: 5-second clip; hold for 1 second, complete the light sweep over 3 seconds, hold the final frame for 1 second.
FRAMING: 9:16 vertical, bottle centered with clean negative space above the cap.
CONTINUITY: Preserve bottle proportions, cap shape, label position, glass edges, reflection direction and product scale throughout.
NEGATIVE DIRECTION: Warped packaging, mirrored or changing label, extra objects, camera shake, flicker, melting edges, changing background.
Thirty-second preflight
- Can you name the single primary action?
- Does the camera have one clear behavior?
- Can the action fit inside the stated duration?
- Are framing and aspect ratio explicit where they matter?
- Does continuity protect the details you would reject if changed?
- Are negatives specific to this shot rather than generic quality words?
Try the structure