Rumpelstiltskin AI: Star in the Viral Tiptoe Dance

Create a new 15-second cinematic comedy scene from these STILL IMAGES. There is no input video. Reconstruct a coherent little tiptoe dance through the shot plan below; approximate the choreography rather than inventing unrelated action.

CAST — keep exactly two distinct people throughout:
DANCER: the adult person in @Image1, recognizable face and hair, transformed into a miniature performer wearing a black patterned tuxedo, white shirt and black bow tie.
SPECTATOR: an adult blonde woman in a cream dress, seated on a hay bale.
@Image1 is ONLY the dancer in every dancer shot. The spectator is a different person and must never inherit @Image1 face.
The dancer is approximately knee-high relative to the seated spectator, with an adult face and compact body proportions, not a baby. Both characters occupy the same physical space. Maintain their faces, hair, clothes, scale and screen positions across edits.

REFERENCE MAP:
@Image2: environment only: close dark timber barn, tall warm straw piles and a narrow straw-covered dance floor.
@Image3: feet insert only: black pointed shoes, raised heels, bent knees and overlapping ankles on straw. Do not use this picture as a face or body identity.

SHOT 1 — 0–3 seconds, medium-wide introduction:
Camera at the miniature dancer eye level. Dancer stands screen-left on the narrow floor; spectator sits screen-right facing them. Warm side light, dark wooden posts behind, layered hay bales, shallow depth of field. Dancer leans forward with flexed knees, elbows bent and hands lifted near the chest, smirks, raises one heel, makes two tiny alternating toe steps, then briefly raises an index finger near the mouth. Spectator follows with their eyes.

SHOT 2 — 3–5 seconds, spectator reaction close-up:
Hard cut to a tight frontal three-quarter shot of the SAME seated spectator. Clearly show the assigned spectator face and hair. Eyes widen and one hand covers the mouth in surprised amusement. Keep the same clothing and light direction. The performer stays off camera; do not transform the spectator into a dancer.

SHOT 3 — 5–8 seconds, dancer medium shot:
Hard cut back to the SAME dancer face and tuxedo, with the spectator shoulder softly blurred at the right foreground. Knees remain bent; perform a compact heel-up shuffle: left foot crosses inward, weight returns right, then two short tiptoe bounces. Shoulders bob, head tilts, elbows open outward. Spread both arms briefly to shoulder height with a smug smile; jacket tails swing. Keep feet touching the ground, no generic walking, large leaps or spinning.

SHOT 4 — 8–10 seconds, low feet detail:
Cut to a low close-up of the dancer's black pointed shoes, lower trousers and jacket tails against straw. Show raised heels, ankles crossing and uncrossing, a small sideways toe slide, then weight transferring to the other foot. Preserve the dancer scale. No faces or extra legs; two feet only, plausible knees and clear contact shadows.

SHOT 5 — 10–13 seconds, return to the two-person composition:
Hard cut to the original screen positions. The dancer does two small rhythmic toe bounces, leans toward the seated spectator, then opens the arms and holds a triumphant mischievous pose. The spectator remains seated, delighted and incredulous, hand lowering from mouth. Clearly retain both assigned faces without changing either person's role. A light shower of golden straw sparkles may appear around the dancer near the finish.

SHOT 6 — 13–15 seconds, final spectator reaction:
Cut to the SAME assigned spectator's reaction close-up. They glance toward the tiny dancer, smile in disbelief and hold the expression for the end. Never restore a different face in this last shot.

Use purposeful hard cuts, stable lenses and consistent geography. Keep a realistic fairy-tale comedy look, visible facial detail and warm muted color. No extra characters, cloned bodies, face morphing, costume changes, floating shoes, new room, subtitles, logos, camera orbit or lip-sync. Generate visuals only; backing music will be added separately.