Window Dance AI prompts and reference video

Recreate the complete performance of Video 1 with the replacement person. Image 1 supplies the replacement identity; Image 2 is the approved appearance preview. Video 1 alone determines the performance and timeline: retain the entire clip, its camera path, zoom, framing, dance gestures, gaze changes, facial reactions and ending at their existing times. Keep the building, room, window, glass, speaker, lighting and all foreground occlusions. Use Image 2 only to guide appearance, not as a first frame or as a new camera view. Do not invent a new dance, camera move, shot, curtain action or ending. Do not extend, shorten or retime the source. Keep the replacement recognizable throughout turns and occlusion. No extra subjects, captions or watermarks. Output silent video: the unchanged reference soundtrack is added separately. Do not generate music, voices or effects. Replace only the sole indoor dancer with the adult from Image 1, using the face, hair, skin tone, build and outfit shown in the uploaded photo and approved preview. Remove the source dancer's identifying features. Replace the whole performer consistently, rather than placing a new face on the original person's body.