Butterfly of Chance: A Cinematic AI Music Video

Objective
Make an AI music video built entirely on the one thing the tools are worst at: two people in real physical contact. No spectacle, no set-piece to hide behind - just a couple who have to read as a couple, touching, dancing in hold, staying the same two people from the first frame to the last. If the contact breaks, the film breaks. That was the brief we set ourselves.

Approach
Two threads, intercut and paced to the song's slow ~70 BPM halftime feel. The verses and the bridge stay intimate and near-still - the couple close, candlelight, a room that feels lived in. The choruses open into a contemporary partnered dance: lifts, dips, an embrace held through a turn, two bodies moving as one line rather than two figures placed side by side. A single reference set keeps both the couple and the world consistent across every look - the black robe, the cream dress, the dark feathered gown - so three and a half minutes and several wardrobe changes still read as one story, one place, one pair of people. The hard work is invisible when it lands: hands that stay hands through a hold, faces that don't drift shot to shot, a lift where two generated bodies actually take each other's weight. That is the direction. It is also the argument for what an AI music video can be when it is built rather than generated.