Crazy Frog AI prompts for the invisible moped video

The best Crazy Frog AI prompt describes four things: the rider squatting in mid-air with nothing underneath, a chaser on a police bike, the place, and the order of the shots. Copy the base prompt below, swap the setting for your own, attach your photo as the reference image, and set the video to 16:9 and about 10 to 15 seconds.

If you'd rather skip prompting altogether, Crazy Frog AI builds the prompt for you from your photo and the scene you pick.

The base prompt

14 seconds, 16:9, comedic action-movie chase on a concrete highway overpass at golden hour.
The person in the reference image rides an invisible moped: seated in a low squat with nothing underneath,
hands gripping invisible handlebars, leaning into turns, hair and clothes blown back by the wind,
gliding forward at speed as if a scooter were there.
Right behind, a motorcycle police officer in a white helmet gives chase on a hovering police bike
with flashing red and blue lights, pointing and shouting.
Cuts: a side tracking shot of the rider, a low front angle on the rider's determined face,
a close-up of the chaser, then the rider jumps a ramp, flips in mid-air still in the squat pose,
lands cleanly and speeds away alone.
Handheld camera, motion blur, sharp recognizable faces.

Why it works:

• "Nothing underneath" and "invisible handlebars" stop the model from adding a real scooter.
• "Low squat" gives the pose that makes the meme readable in one frame.
• The shot list keeps the five beats of the trend in order instead of one long tracking shot.
• "Sharp recognizable faces" pushes the model to keep the likeness from your photo.

Scene versions

Replace the setting and the chaser in the base prompt. These are the settings Crazy Frog AI uses:

• Night market: "a crowded night market street at night, glowing paper lanterns, steam from food stalls, wet pavement reflections", chased by "a city patrol officer in a reflective vest"
• Shopping mall: "a bright shopping mall atrium with glass railings, escalators and potted palms", chased by "a mall security guard in a navy uniform"
• Office: "an open-plan office with rows of desks, monitors and swivel chairs under fluorescent lights", chased by "a stressed manager in a shirt and tie holding a coffee mug"
• Desert: "an empty two-lane desert highway between red rock mesas under a hard blue sky", chased by "a highway patrol trooper in a campaign hat"
• Ski slope: "a snowy ski slope with pine trees and a chairlift, powder spraying in the sun", chased by "a ski patrol member in a red jacket with a white cross"
• Moon base: "a lunar base on the surface of the moon, grey dust, domes and Earth hanging in the black sky", chased by "an astronaut police officer in a white space suit with a badge"

All 12 scenes are ready to use, with no prompt, on the Crazy Frog AI generator.

Prompt for a friend as the cop

Attach two reference images, rider first, then the friend. Change the chaser sentence to:

Right behind, the person in the second reference image, dressed as a motorcycle police officer
with a white helmet, gives chase on a hovering police bike with flashing red and blue lights.

Keep the outfit in the prompt. Without it, most models dress your friend in whatever they wore in the photo.

Prompt for remaking the original clip

If your model accepts a reference video, you can keep the exact cuts of a clip and swap the people:

Recreate the reference video shot by shot: keep its cuts, camera movement, timing and actions.
The rider on the invisible moped is the person in the first reference image.
The chaser is a motorcycle police officer in a white helmet.
Keep the faces, hairstyles and outfits recognizable. Natural, smooth motion.

Only use reference videos you have the right to use.

Fixes for common problems

A moped or scooter appears

Add "there is no vehicle, empty space under the rider" and put "moped, scooter, motorcycle under the rider" in the negative prompt if the model has one.

The rider stands up

Repeat the pose in the flip and landing: "flips in mid-air still in the squat pose, lands in the squat". Models tend to straighten people out during action beats.

The face drifts away from the photo

Use a sharper, front-facing photo, add "sharp recognizable face" and keep the clip short. Longer clips give the face more time to drift.

The chase feels slow

Add motion words to both characters: "gliding forward at speed", "leaning into turns", "motion blur". A handheld camera also adds energy.

Too much happens at once

Cut the shot list to three beats (side shot, chaser close-up, getaway) on models that only render 5 to 10 seconds.

Or skip the prompt

Crazy Frog AI was built so you don't have to tune any of this. Upload a photo, pick a scene, and the chase, the chaser and the shot list are handled for you. See how to make a Crazy Frog AI video for the full walkthrough.

Make your Crazy Frog AI video