CLIP 1 PROMPT
A person matching the reference selfie stands on an empty city street at night, facing the camera in a waist-up shot. They raise a red 2 card and a blue 4 card toward the lens.
WHY IT'S WRITTEN THAT WAY
"Matching the reference selfie" locks the face; "waist-up" keeps the framing tight enough for the cards to read on a phone; "toward the lens" is what sells the illusion that the camera is about to be covered.
CLIP 2 PROMPT
Continue the exact same shot. A hand swipes across the lens. When the hand clears, the reference car is parked naturally behind the person, who reacts with a subtle smile.
HOW TO USE IT
Feed your selfie as the character reference and the car photo as the vehicle reference. Keep the night setting and camera framing identical across both clips — any drift shows up instantly.
WHY ONE PROMPT RARELY WORKS
The swipe is doing two jobs at once: covering the camera and hiding a scene change. Split into setup + reveal, the model gets a clear beat for each, and the continuity holds.
COMMON FAILURE CASES
The car's colour drifting from the photo, a hand that never quite covers the lens, and a reveal that happens too early. If one appears, regenerate the second clip rather than the whole thing.