How to choose photos for an AI video
Start with a sharp reference that shows the subject’s face and distinctive features. Use separate photos for separate people, keep their order consistent, and choose a body view when the scene depends on legs, paws or dance poses. Review the photo requirements in the template form, then save your inputs and request a generation quote.
Choose a reference for the actual scene
A face-focused transformation needs visible eyes, hair and facial contours. A flying-broom scene or dance also depends on body proportions, so a full-body reference helps with preparation. A tightly cropped selfie cannot show an outfit or leg position that is outside the frame. Avoid heavy beauty filters and repeated screenshots that remove facial detail.
For pets, preserve the features you recognize
Show ears, muzzle, eye color and coat markings. Leave room around a rabbit’s ears or a dog’s paws so the preparation crop does not remove them. Our pet bridge sample uses a lop-eared rabbit, Ramalama uses a West Highland terrier, and La Calabaza uses an orange cat. Each is one example, rather than proof that every breed works in every scene.
For two people, assign the roles before generating
Keep one person per reference and label the roles in order. Heavenly Reunion assigns the first portrait to the traveler and the second to the waiting adult. The two-person Thriller example binds each identity to its own dancer, beginning left and right. Screen position can change during a turn; that should not change who the dancer is.
Check detail before using a small portrait
Use JPG, PNG or WebP files up to 12 MB each, at least 300 pixels wide and tall. The horror-call template needs a prepared scene showing the adult, corded phone and open doorway. A small face crop cannot supply those scene details.
Judge the result beyond the first frame
Compare the face before and after a turn, the pet’s markings through a costume change, and hands or paws near props. Watch the opening and ending as well as the most attractive frame. A clear input helps judge identity, but it does not guarantee a particular output.
Questions before you create?
Can I upload these photos here now?
Template forms let you preview reference photos locally, sign in to upload and save a private draft, and review a generation quote. Current availability is shown in the form.
Should I send a group photo for a two-person scene?
Use one adult per photo. Assign the photos to the labeled roles so each person stays attached to the intended character; group photos are not an input option for these templates.