Viral Pet Fridge Heist AI Video Prompt: Your Cat, Caught on Fridge Cam Stealing Cake

The funniest pet videos on the internet right now are shot from inside a refrigerator. The door opens at 2 a.m., a guilty little face leans in, and an entire tray of tiramisu disappears spoonful by spoonful — all “caught” by a fridge camera that was never really there. It’s AI, it takes one photo of your pet, and viewers are watching these on loop.

The comedy lives in the details: the fixed security-camera angle, the serious thief expression, and that empty-tray ending. Miss any one of them and the video turns into a generic pet clip. The prompt below builds all three in, and it works for cats, dogs — even a person playing the snack thief.

What you need

  • One clear photo of your cat or dog — face straight at the camera is perfect, since the fridge cam stares right back
  • An AI video tool with image-to-video (Kling, Runway, Veo or similar)
  • The Instagram or TikTok editor for a trending funny audio at posting time

No pet? The same prompt works with a photo of yourself or a family member caught mid-snack — the human versions do surprisingly well because the guilt reads bigger.

The master prompt

Paste this with your pet’s photo. Keep the camera instructions exactly as written — the locked-off fridge angle is what sells the joke.

Create a funny photorealistic video from the uploaded photo, vertical 9:16, about 12 seconds, shot from a fixed hidden camera inside a refrigerator looking out. Use ONLY the pet from the uploaded photo — same face, same fur colour, same markings. On the shelf sits a full tray of tiramisu dusted with cocoa. Scene 1 (0–4s): the fridge door opens in the dark, cool blue fridge light spills in, the pet leans into frame with wide, guilty eyes and looks straight at the camera. Scene 2 (4–9s): the pet holds a small spoon in its paw and eats spoonful after spoonful of the tiramisu, glancing at the camera between bites, cocoa dust on its nose. Scene 3 (9–12s): the tray is empty except for crumbs, the pet freezes with the spoon mid-air staring into the lens, then the fridge door slams shut to black. Camera: completely fixed inside-fridge POV, slight security-camera feel, no camera movement. Strict rules: use only the uploaded photo’s pet — no face change, no markings change, no extra animals, no people, no readable brand labels, no text overlays, no watermark, no lip-sync, no dialogue.

Short version for free tools (when credits or characters are tight):

Animate the uploaded photo as a fridge-cam video: fixed camera inside a fridge, the pet sneaks in at night and eats a whole tray of tiramisu with a tiny spoon, guilty face at the camera, ends on the empty tray. Vertical 9:16, funny and photorealistic. Same pet as the photo only — no face change, no extra animals, no text, no watermark.

How to do it

  1. Pick the guiltiest photo you own. Wide eyes and a straight-on stare render best. A blurry side shot gives the model nothing to act with.
  2. Upload it as the reference image in your video tool’s image-to-video mode.
  3. Paste the master prompt and set 9:16 and roughly 12 seconds. Shorter than 10 and the three-beat joke — peek, eat, caught — gets squeezed.
  4. Check the paws and the tray. Spoons in paws are the hardest render in the video. If the spoon melts or multiplies, regenerate before you settle.
  5. Swap the dessert if you like. Chocolate cake, cheesecake or a bowl of strawberries all work in the same sentence — just keep the “tray” wording so scene 3’s empty ending still makes sense.
  6. Add trending audio in the Instagram or TikTok editor. The AI tool exports silent video; the sneaky, tiptoe-style sounds this trend rides on get added when you post, and using the same audio is how these videos find each other’s audience.

Fixes for the usual problems

The camera moves around. Models love to “help” with a dramatic push-in. Repeat “camera locked off, zero movement, fixed security camera” — twice if needed. The stillness is the joke’s straight man.

My pet’s face melts while eating. Eating motions are hard on faces. Slow the middle scene down in the prompt — “slow, small bites” — and the face usually holds together.

The fridge looks like a shop shelf. Add household specifics: “home fridge with milk bottles, jars and leftovers on the shelves.” Generic fridges render like adverts; cluttered ones render like evidence.

The ending got cut. If your tool caps at 10 seconds, drop scene 2 to (4–8s) and protect the empty-tray slam — never sacrifice the punchline to save the set-up.

More prompts people are copying right now

The hub refreshes every day — these are the fastest movers this week:

Frequently asked questions

Can I make this video with a person instead of a pet?

Yes. Use a clear front-facing photo of yourself or a willing family member and keep the prompt identical — just say “the person from the uploaded photo” instead of “the pet.” The 2 a.m. snack-thief guilt lands just as well.

Which dessert works best in the prompt?

Tiramisu is the classic — the cocoa dust shows every missing spoonful on camera. Chocolate cake and cheesecake read well too. Avoid anything pale or liquid; the camera needs to see the tray emptying.

Why does my spoon look weird?

Paws plus cutlery is genuinely hard for video models. A smaller spoon, slower bites, and one regeneration usually fix it. If it keeps failing, switch the pet to licking the tray edge — rude, but renders clean every time.

How long should the fridge heist video be?

Ten to thirteen seconds. Under ten, the three beats crush together; over thirteen, the fixed camera starts to feel static and faces drift. Twelve is the sweet spot this trend settled on.