Up to four people join a room and spawn as avatars in a theatre. The host starts the show, a 3D Michael Jackson dances the full four minutes and fifty-seven seconds locked to the real audio, and everyone's camera snaps to a director cam that reproduces the 1983 broadcast's thirty-seven cuts in order.

The motion is captured from the broadcast footage itself rather than animated by hand. That was not the first plan: a fully hand-authored choreography was built, locked to the beat, and thrown away for being ten times worse, the roughness of real capture carries the character. What shipped is the capture with three surgical repairs: the root is anchored only where tracking is confident, gaps are filled from a motion graph of the same performance, and the moonwalk glides on an authored path while the legs stay captured.

Because video and audio from the source run to the same length, the two stay in step across the whole song, and the animation follows the audio playhead rather than the frame clock, so a slow machine drops frames instead of drifting out of time.