🎨 Seedream 5.0 Pro — ByteDance's flagship image model: up to 2K, best-in-class typography & multi-reference editing!

Try It!
Logo

Caught at the Window AI

Upload a photo and the camera starts across the street at night, then creeps in on one lit apartment window. Inside, you're at the open fridge stuffing a big slice of cake into your mouth. You freeze mid-bite, turn and glare straight into the lens, chew once and yank the curtain shut. Vertical 9:16 with night-city sound and a comic sting.

Why Make Your Caught at the Window Video Here

The busted-at-the-window joke, starring you, from one photo.

  1. 01

    A Real Spy-Zoom Shot

    The clip starts wide across the street and slowly zooms into one window, like a telephoto lens, so the reveal feels like you've genuinely been spotted.

  2. 02

    Your Face, Your Glare

    Wan 3.0 keeps your face from the photo, so the person frozen with a mouthful of cake and staring down the camera is clearly you.

  3. 03

    Comic Timing in the Sound

    Quiet night-city ambience, a tense string pluck the moment you notice the lens and a curtain swish at the end. Original audio, no speech, no copyrighted song.

  4. 04

    Vertical and Ready to Post

    Every clip renders in 9:16 at 5 to 15 seconds, in 480p or 720p, or 1080p for Pro members, so it fits TikTok, Reels and Shorts without cropping.

How to Make a Caught at the Window AI Video

Three steps from a selfie to getting busted at the window

  1. 01

    Upload Your Photo

    Use a clear, well-lit photo of yourself with your face toward the camera. Add a friend's photo in the optional slot if you want to get caught together.

  2. 02

    Set Length and Quality

    Choose 5 to 15 seconds (5 is the default) and a resolution: 480p, 720p, or 1080p for Pro members. There is no prompt to write; the street, the zoom, the cake and the curtain are built in.

  3. 03

    Generate and Download

    Hit generate and download a vertical MP4 with sound. The default 5-second 720p clip costs 65 credits, the price is shown before you start, and failed renders are refunded automatically.

Guide

Caught at the Window AI: The Complete Guide

TL;DR

Caught at the Window AI turns one photo into the window zoom prank: the camera creeps from across the street into a lit apartment window, catches you secretly eating cake at the fridge, and you freeze, glare into the lens and yank the curtain shut. Vertical 9:16 with comic sound, no prompt needed.

What Is the Caught at the Window Trend?

The caught at the window trend is a short comedy format. The camera starts far away, zooms slowly into one window, and finds someone in the middle of something a little embarrassing. They realise they're being watched, stare at the camera and close the curtain.

Filming it for real means a second person across the street with a long lens, a lit room at night and perfect timing, and pointing a camera into other people's windows is not something to do. As an AI effect, none of that is needed: you're the only person in the joke, and it's all generated.

This template is made with Wan 3.0. Your photo gives the face of the person at the fridge; the building, the zoom, the cake and the sound come from the template.

What Happens in the Video?

  1. Across the street: a city at night, with the camera looking at a large apartment building.
  2. The zoom: the camera slowly and sneakily zooms toward one lit window.
  3. The crime: inside, you stand at an open fridge stuffing a big slice of cake into your mouth.
  4. Busted: you freeze mid-bite and slowly turn to stare into the lens with an annoyed look.
  5. Curtain: you chew once, reach up and yank the curtain shut.

The audio is generated with the clip and timed to the joke: quiet night ambience during the zoom, a tense comic string pluck when you notice the camera and a curtain swish at the end. There's no speech.

The warm light inside the flat against the dark building is what draws the eye to your window, and your face comes from your photo throughout.

Why Does the Window Zoom Joke Work?

It's built on suspense. The slow zoom makes viewers wonder what they're about to see, and the payoff is small and relatable: someone eating cake alone at midnight.

The glare is the punchline. Freezing mid-bite and staring straight at the camera turns the viewer into the one who's been caught snooping, and the curtain closing ends it on a beat.

It's also easy to read without sound. A building, a window, a face with cake: anyone scrolling gets it in a couple of seconds.

Caught at the Window Ideas

  • The diet joke: post it on day two of your new healthy eating plan.
  • The midnight snack confession: a caption about what you really do after everyone's gone to bed.
  • Caught together: add a friend or partner in the second slot and you both get busted at the fridge.
  • Birthday leftovers: post it the day after a party, when the cake is suspiciously smaller.
  • Stitch bait: post it with a caption asking what the camera should catch you doing next.

How to Get the Best Result

  • Face the camera: the big moment is you staring into the lens, so a front-facing photo works best.
  • Keep your face clear: no sunglasses, masks or hair across your eyes.
  • Use even light: a sharp daytime or indoor selfie holds up better than a dark or blurry one.
  • Skip heavy filters: smoothing and face-shaping filters can make the clip look less like you.
  • One person per photo: for two people, use the second slot instead of a group photo.
  • Your photo or one with permission: never a celebrity, a neighbour or anyone who hasn't agreed.

What If the Result Isn't Right?

  • You don't look like yourself: try a sharper, front-facing photo with your face uncovered and no filter.
  • Your friend looks like you: give each person their own photo in a separate slot instead of one shared photo.
  • The glare is too subtle: a 10-second clip gives the freeze and the stare more time to play out.

Test a new photo at 480p for 35 credits before rendering the final version at 720p or 1080p.

Which Length and Resolution Should You Pick?

Billing is per second: 7 credits at 480p, 13 at 720p and 28 at 1080p, which is for Pro members. The default 5-second 720p clip costs 65 credits, the exact price is shown before you start, and failed renders are refunded automatically.

Five seconds keeps the joke tight. A longer clip slows the zoom down and builds more suspense before you're caught.

ResolutionLengthCreditsBest for
480p5 s35Testing your photo
720p5 s65Most posts (the default)
720p10 s130A slower, sneakier zoom
720p15 s195Maximum suspense
1080p5 s140 (Pro)The sharpest short clip
1080p15 s420 (Pro)The sharpest full-length clip

How to Post It on TikTok, Reels and Shorts

  1. Download the MP4: vertical 9:16 with sound, no cropping needed.
  2. Set up the joke in text: a line like 'me at 1 a.m. on day two of my diet' over the zoom does most of the work.
  3. Keep the sting: the string pluck lands on the moment you notice the camera. If you add a trending sound, keep that moment audible.
  4. Name the trend: mention 'caught at the window' in the caption so people searching for it find your post.
  5. Label it as AI: use the platform's AI label, so nobody thinks a real window was filmed.

Keep the joke about yourself. Don't caption it as if it shows a real neighbour or a real flat.

Caught at the Window AI FAQ

It's a comedy format where a camera zooms from far away into someone's window and catches them doing something they'd rather keep private. They notice, stare at the camera, and shut the curtain. The joke is the slow zoom and the look of being completely busted.
It's night in the city. The camera starts across the street and sneaks toward one lit apartment window. Inside, you're at an open fridge eating a big slice of cake in secret. You freeze mid-bite, turn and glare into the lens, chew once and yank the curtain shut.
Upload a photo of yourself, pick a length and a resolution, and press generate. The building, the zoom, the cake, the sound and the timing all come from the template, so you don't write a prompt.
A clear, front-facing photo of one person in good light, with nothing covering the face. The best moment in the clip is your glare into the lens, so the face matters more than the outfit.
Yes. Add their photo in the optional second slot and you're both caught at the fridge. Only upload a friend's photo if they're happy to be in it.
Not in this template. The cake, the fridge and the curtain are fixed so it works without a prompt. For a different scene, write your own description in a general image-to-video tool.
Clips run 5 to 15 seconds in vertical 9:16, and 5 seconds is the default. You pay per second: 7 credits at 480p, 13 at 720p and 28 at 1080p (Pro members), so the default 5-second 720p clip costs 65 credits. If a render fails, the credits are refunded automatically.
Yes. Wan 3.0 generates quiet night-city ambience, a tense comic string pluck when you notice the camera and a curtain swish at the end. There's no speech and no copyrighted song; add a trending sound in the app if you want one.
Only your own photo, or a photo of someone who has agreed to be in the video. Don't use a photo of a celebrity, a neighbour or anyone else without their permission, and don't present the clip as real footage of a real window.
No. The building, the window, the camera and the cake are all generated from your photo, and nobody is filmed. It's a harmless joke, so label it as AI when you post.

AI Pricing Plans

Flexible AI Pricing

Pay-as-you-go credits or subscription plans. No hidden fees, cancel anytime.

One Time supports crypto payment (BTC, USDT, ETH, 350+) · Monthly billing

Free

Try before you buy

Free
32points
  • Up to 3 videos
  • Up to 32 images
  • Multi-Model Support
  • Text to Video
  • Image to Video
  • Video to Video
Show more (+11)
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support

Pro

Popular

Elevate your AI experience

29.99USD / 1 Month
800
800 credits/mo
  • Up to 80 videos
  • Up to 800 images
  • 3 tasks(Parallel Tasks)
  • Multi-Model Support
  • Text to Video
  • Image to Video
Show more (+12)
  • Video to Video
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support

Lite

Start your AI journey

19.99USD / 1 Month
300 credits/mo
  • Up to 30 videos
  • Up to 300 images
  • 3 tasks(Parallel Tasks)
  • Multi-Model Support
  • Text to Video
  • Image to Video
Show more (+12)
  • Video to Video
  • Consistent Character
  • AI Animation Generator
  • Templates & Effects
  • AI Video Enhancers
  • Interactive Community
  • Faster Generation Speed
  • No-watermark Outputs
  • More Camera Movement
  • Private Video Visibility
  • Copy Protection
  • Priority Support
  • Card
  • Crypto accepted (BTC, USDT, ETH, 350+)

Privacy preferences

We use essential storage to run the service. Optional analytics, advertising and convenience features stay off until you choose to enable them. Privacy policy