Start with one clear, recent portrait of each person, facing the camera in even light. Keep the face unobstructed and avoid beauty filters. Additional views of the same person can help describe their appearance, but neither more files nor a head-turn video guarantees an accurate likeness.
LOVESTORY AI Editorial · Product documentation, not independent testing. Our facts and limitations
Make the face easy to interpret
Use a separate portrait for each person. A well-lit face with visible eyes and a natural expression is more useful than a distant group photo. Avoid sunglasses, strong colored light, deep shadows and photos cropped through the chin or forehead.
Crop distracting surroundings while retaining the full head and a little space around it. Check the actual selected image after cropping. A higher pixel count cannot recover a face that was out of focus when photographed.
Add views that provide new information
If available, add a gentle three-quarter view and another natural expression of the same person. Prefer photos taken at a similar age and with the appearance you want represented. A collection mixing very different ages, hairstyles and filters can be ambiguous.
Assign every file to the correct named person before continuing. A beautiful photo of both people is not a substitute for clear individual references when the interface asks for each person separately.
Video is an option, not a requirement
When recording reference video, hold the camera steady in soft light and turn the head slowly from front toward each side. Stay in frame and pause briefly at each angle. Do not share identification documents or sensitive background details.
In the web flow, the original reference clip stays in the browser; only the still frames you select and confirm are uploaded. This is not identity verification. A portrait upload remains available if recording is inconvenient.
Know what the preview proves
The text preview checks story direction; it does not demonstrate final facial resemblance or motion quality. Animated interpretation may differ from the source photograph. Character consistency must be assessed across the finished scenes, not inferred from a single attractive still.