Which photos make a better AI rap video?
Light, framing and a clear cast: choose reference photos that give the model a useful starting point.
Make the face easy to read
Choose a recent portrait with even light and a face that is clearly visible. Avoid sunglasses, heavy filters and deep shadows. A model has fewer clues when most of a face is covered. A waist-up photograph can also communicate clothing and posture, while a tight crop mainly describes identity.
Give each person enough room
For two people, separate portraits make the cast easier to describe. If you use a shared photo, choose one where both faces are similarly large and neither overlaps the other. Very crowded group shots leave too much room for ambiguity. Name each person’s role in the topic if it matters who starts the verse.
Check size and permission before upload
The studio accepts JPEG, PNG and WebP images up to 10 MB each. Use an image between 301 and 5,999 pixels on each side, with an aspect ratio wider than 0.4 and narrower than 2.5. Crop a very wide group photo first. Upload only images you own or have permission to animate, and ask every identifiable person before putting them in a synthetic performance.
Review the result as a new performance
Reference images guide generation; they do not lock every facial detail or expression. Watch the finished clip, especially faces, hands and mouth movement. If a face is inconsistent, try a cleaner portrait and a simpler camera direction. A short scene with two visible people is easier to review than a complex sequence of cuts.