1. Choose the scene and add your photos
There are three scenes. The hot tea prank is a 17-second cafe scene: someone sneaks a hot spoon into the tea, the lead jumps, mops up and the clip ends on a facepalm. Spaghetti night and Burger business are 15-second dinner scenes starring one pet. For the tea prank you also choose a cast mode: You (one person photo), You + friend (two person photos), 1 dog (one pet photo) or 2 dogs (two pet photos). Use one clear subject per photo, facing the camera, with the full face, both ears and the upper chest in frame. The shorter side must be at least 300 pixels. On a computer you can drag a photo straight from Finder or paste it with ⌘V; on a phone, tap the photo box. Check the crop before continuing and replace a blurred photo rather than expecting the model to invent detail it cannot see. For two-photo tea modes, photo 1 becomes the lead on the left and photo 2 the neighbor on the right.
2. Preview, sign in and generate
For spaghetti and burgers, the studio creates two watermarked still images to choose from. Review the hands, food, expression, likeness and clothing before choosing; no extra visual check or automatic quality retry runs. The clean original of the still you pick is reused as the first frame of your video. The tea prank skips this step: your photo goes straight into a fixed 17-second motion template. Press Generate video. Sign in with Google or with your email so the result is attached to your account; your photo, scene and preview stay saved in the browser through sign-in and checkout. If you have no video left, choose a one-time pack: $9.90 for one, $19.90 for three or $29.90 for six. There is no subscription. Generation usually takes 5–20 minutes. You can leave the page; a status item shows progress and My videos keeps the job.
3. Download and share the MP4
The finished file is a 720p, 9:16 vertical MP4: 15 seconds with food sounds for the dinner scenes, 17 seconds with the scene’s original reactions and laughter for the tea prank, never with added music. A delivery email links to the result. Download it within 24 hours after delivery; after that the private file expires and cannot be restored. On a phone, use the download or share option of your browser; a website cannot add files to your photo library by itself.
The four tea prank cast modes
You puts one person in the lead seat next to a fixed supporting dog. You + friend uses two person photos: photo 1 is the lead on the left and photo 2 is the neighbor on the right. 1 dog puts one pet in the lead seat next to the same supporting dog, and 2 dogs seats two of your pets side by side. In every mode the cast wears complete hoodies, the hands come from their own sleeves, the camera operator’s hand brings in the spoon and the clip ends on the same fixed golden dog facepalm. Choose the mode before you upload, because it decides how many photos the studio asks for.
Photo checklist
Good: daylight, eye level, one subject, sharp eyes and nose, both ears visible, plain background. Avoid: group shots, sunglasses, hats that hide the ears, strong beauty filters, heavy motion blur, tiny subjects in a wide shot and screenshots of other videos. Only upload photos you have the right to use, and only photos of people who agreed to appear in the video.
If something goes wrong
Use one subject per photo and select the matching cast mode. Review both preview images yourself before choosing. If a paid generation fails for a technical reason, it is retried once; if it fails again, or moderation declines the request, your video entitlement is restored once so you can try another photo. Unknown submissions are checked before any new paid request is placed.
AI generation vs a CapCut hand overlay template
This tool creates the animal, arms, setting and movement together, so the hands belong to your pet and react to the food or the tea. A hand-overlay template instead places a hand graphic or separate footage over an existing clip. Overlays are quick and free but stay flat; generated scenes take longer and cost a little, but move as one picture. Results, controls and timing differ between tools, so do not assume every template works the same way.