Best Photos for a Cafe POV Video

The scene has exactly two roles: the person already sitting on the left, and the customer who walks in from the right. Upload one photo for each.

Two separate photos give the model two separate identities. A single image with both people forces it to guess which face belongs to which role, and the swap usually lands in the wrong order.

The clip is about posture: someone perching on a table, then sliding down onto a bench. A half-body or full-body photo gives the model the shoulder and hip information it needs to make that movement believable.

A tight head crop still works as a reference, but the body is invented, and invented bodies tend to sit awkwardly.

A face lit from one side reads as two different faces across the clip. Prefer soft, even light with the person facing the camera.

Avoid backlit window shots, hard flash, and heavy beauty filters. Filters smooth away the exact details that keep a likeness consistent.

The scene ends in an embarrassed pause. A neutral or relaxed expression gives the model room to act that out; a big grin or a surprised face fights the script.

Keep hands away from the mouth and avoid sunglasses. Both hide the parts of the face that carry the reaction.

JPG, PNG or WebP, up to 8 MB per file. Larger browser uploads are converted and downscaled on the client before they reach the model, so a modern phone photo is fine.

Both images are moderated before generation. A rejected photo returns an error and is not charged.