Why two separate photos work better than one
The generator reads each face from its own picture and keeps that person on their side of the mic for the whole clip. When both people share one image, the faces are smaller and the model has to separate them, which is where features start to mix and the two people get mixed up between cuts.
The photos do not need to be taken together, on the same day or in the same light. Two ordinary selfies are enough. If the only good picture you have is a photo of the two of you, crop it twice so each person has their own image.
Who should be photo 1, and who photo 2?
Give photo 2 to whoever should steal the show: they rap most of the clip and get the close-up. Photo 1 opens with the hook and spends the rest of the clip reacting, which is often the funnier role for the friend who never takes anything seriously.
If you plan to make two clips, swap the photos the second time. Posting both versions side by side is a common way people use the trend.
What makes a good photo
One person (or one pet) in the frame, face clearly visible, front-facing or a slight angle, chest-up or closer. Even daylight beats a dark room. Remove sunglasses, masks and anything covering the eyes, and skip heavy beauty filters.
What you wear in the photo is what you wear in the clip, shoes included on the wide shot. Use photos you own or that the other person agreed to share. Photos of minors, celebrities and public figures are not allowed.
Duo ideas for the 58th Floor
Best friends: the classic version, one confident hook and a friend who takes over the verse. Partners: an easy anniversary post. Siblings: settle a family rivalry as a studio face-off. A parent and grown-up child: the surprise of seeing your dad take the verse. Coworkers: a launch-day clip for the team channel, if everyone agreed to it.
You and your pet also works: put your dog or cat in either slot and it raps at the mic with its real paws.