Mudar fundo de vídeo com IA: 1 take vira 10 vídeos com Seedance 2.5 Edit + Claude (com os prompts)

The same take, but the world around her transforms. Her performance never changes. RULE 0 - THE WOMAN IS 1:1 WITH THE SOURCE VIDEO (overrides everything). The source video is a full-body performance capture. In every frame her head, face, lips, hair, shoulders, torso, arms, hands, fingers and legs are in exactly the same position as in the same frame of the source. Second N of the output = second N of the source. Nothing re-timed, sped up, slowed down, frozen, skipped or re-acted. Same woman: same face, long straight dark-brown hair with light-brown highlights, same makeup, same small earrings and rings, not beautified. In her right hand (frame-left) she pinches a small black wireless microphone between thumb and index finger near her chin, exactly as in the source, for the whole clip. Lips in perfect sync with the original audio. Keep the original speech audio. THE PERFORMANCE TRACK (follow it): she stands in the same spot for the whole clip, talks to camera and gestures with her free left hand exactly as in the source. Around 2-3 s she turns her body to her left and back to camera, exactly when the source does. TIMELINE (seconds from the start of this clip), one continuous take, no cuts: 0-2.2s: the same office and the same outfit as the source (black sleeveless V-neck mini dress over a white long-sleeve collared shirt, black belt with a gold buckle), medium-wide shot from her front-left. 2.2-3.5s: while she turns, the whole office transforms around her into the summit of a snowy mountain: jagged white peaks and a sea of clouds behind her, deep fresh snow under her feet, bright cold sunlight, light snowflakes in the air, a soft wind moving the ends of her hair. 3.5-5.8s: she stays on the snowy summit, still in her original outfit; the camera starts a slow arc from her left toward her front. 5.8-7s: her outfit becomes a white puffy winter down parka reaching mid-thigh with a fur-trimmed hood resting on her shoulders, black thermal leggings, white snow boots, and a cream knit beanie; her hair stays long and loose. 7-9.3s: she stays on the snowy summit in the winter outfit, medium shot from the front. Smooth motion-control feel, level horizon, no shake. FAILURE: the transformation happening before 2.2 s; the room staying an office after 3.5 s; the outfit changing before 5.8 s; any gesture missing, early, late, mirrored or done with the other hand; hands idle where the source moves; lipsync off; face changed or beautified; hair cut or tied up; the microphone disappearing or changing hands; warped hands or extra fingers; cuts; camera shake; new people; film gear; added text, captions or logos; music.

The same take, relocated to a snowy mountain summit for the whole clip. Her performance never changes. RULE 0 - THE WOMAN IS 1:1 WITH THE SOURCE VIDEO (overrides everything). The source video is a full-body performance capture. In every frame her head, face, lips, hair, shoulders, torso, arms, hands, fingers and legs are in exactly the same position as in the same frame of the source. Second N of the output = second N of the source. Nothing re-timed, sped up, slowed down, frozen, skipped or re-acted. Same woman: same face, long straight dark-brown hair with light-brown highlights, same makeup, same small earrings and rings, not beautified. In her right hand (frame-left) she pinches a small black wireless microphone between thumb and index finger near her chin, exactly as in the source, for the whole clip. Lips in perfect sync with the original audio. Keep the original speech audio. FROM THE FIRST FRAME: she stands on the summit of a snowy mountain: jagged white peaks and a sea of clouds behind her, deep fresh snow under her feet, bright cold sunlight, light snowflakes in the air, a soft wind moving the ends of her hair. She wears a white puffy winter down parka reaching mid-thigh with a fur-trimmed hood resting on her shoulders, black thermal leggings, white snow boots and a cream knit beanie; her hair stays long and loose. THE PERFORMANCE TRACK (follow it): she stands in the same spot, talks to camera and gestures exactly as in the source. In her left hand she holds a small tube low by her hip at the start and lifts it and waves it exactly as in the source. TIMELINE (seconds from the start of this clip), one continuous take, no cuts: 0-3.3s: medium shot from the front, slow push-in; the object in her left hand is a small tube. 3.3-4.4s: the tube in her left hand transforms into a stainless-steel thermos flask with a red cap, a little steam rising from it; she keeps holding and moving it exactly like the tube in the source. 4.4-11.2s: slow arc of the camera from her front toward her right side (frame-left), medium shot, she stays on the snowy summit with the thermos. Smooth motion-control feel, level horizon, no shake. FAILURE: any office, shelves or indoor wall visible; the outfit changing; the thermos switching hands or disappearing after 4.4 s; any gesture missing, early, late, mirrored or done with the other hand; hands idle where the source moves; lipsync off; face changed or beautified; hair cut or tied up; the microphone disappearing or changing hands; warped hands or extra fingers; cuts; camera shake; new people; film gear; added text, captions or logos; music.

The same take, re-lit. The camera stays fixed; the ONLY thing that changes is the lighting of the room. RULE 0 - THE WOMAN IS 1:1 WITH THE SOURCE VIDEO (overrides everything, including the light). The source video is a full-body performance capture. In every frame her head, face, lips, hair, shoulders, torso, arms, hands, fingers and legs are in exactly the same position as in the same frame of the source. Second N of the output = second N of the source. Nothing re-timed, sped up, slowed down, frozen, skipped or re-acted. Same woman: same face, long straight dark-brown hair with light-brown highlights, same makeup, same small earrings and rings, not beautified. Same outfit: a black sleeveless V-neck mini dress over a white long-sleeve collared shirt (the white hem shows below the dress), black belt with a gold buckle. In her right hand (frame-left) she pinches a small black wireless microphone between thumb and index finger near her chin, exactly as in the source, for the whole clip. Lips in perfect sync with the original audio. Keep the original speech audio. THE PERFORMANCE TRACK (follow it): she stands in the same spot, talks to camera and gestures with her free left hand exactly as in the source. Around 9-10 s she turns her head and shoulders to her left and looks back over her shoulder, exactly when the source does, then turns back to camera. CAMERA: fixed medium shot (head to knees) from slightly to her right side (frame-left), for the whole clip, no camera movement, no cuts. LIGHT TIMELINE (seconds from the start of this clip): 0-8.8s: the office as in the source, natural daylight mixed with the warm LED strip. 8.8-10.3s: it becomes night: the room goes dark, cold moonlight falls on her from the side. 10.3-12s: almost too dark, she is barely visible. 12-16.4s: a warm orange lamp light turns on behind her and from the side, glowing on her hair and lighting her face in warm amber tones. 16.4-18s: the light shifts to cool blue tones. 18-19.4s: far too blue, a strong saturated blue cast over everything. 19.4-21.9s: the light returns to the balanced natural daylight of the source. THE ROOM: the same office as the source: light oak built-in shelves up to the ceiling with framed plaques (one green), plants and skincare products, a backlit pink onyx stone wall panel, pink tulips, an acrylic cube with gold spheres, a white ceramic flower vase, a long slatted oak sideboard, light wood floor. Parts the original lens never saw continue the same office in the same style. FAILURE: any gesture missing, early, late, mirrored or done with the other hand; hands idle where the source moves; lipsync off; face changed or beautified; hair changed; outfit changed; the microphone disappearing or changing hands; the camera moving; cuts; new people; lamps or film gear appearing in frame; added text, captions or logos; music.

The same take, re-photographed by a different camera on every beat. The ONLY thing that changes is the viewpoint. RULE 0 - THE WOMAN IS 1:1 WITH THE SOURCE VIDEO (overrides everything, including the camera). The source video is a full-body performance capture. In every frame her head, face, lips, hair, shoulders, torso, arms, hands, fingers and legs are in exactly the same position as in the same frame of the source. Second N of the output = second N of the source. Nothing re-timed, sped up, slowed down, frozen, skipped or re-acted. Same woman: same face, long straight dark-brown hair with light-brown highlights, same makeup, same small earrings and rings, not beautified. Same outfit: a black sleeveless V-neck mini dress over a white long-sleeve collared shirt (the white hem shows below the dress), black belt with a gold buckle. In her right hand (frame-left) she pinches a small black wireless microphone between thumb and index finger near her chin, exactly as in the source, for the whole clip. Lips in perfect sync with the original audio. Keep the original speech audio. THE PERFORMANCE TRACK (follow it): for the whole clip she stands in the same spot, talks to camera and gestures with her free left hand exactly as in the source, at the same moments. RULE 1 - THE CAMERA, with hard camera changes exactly on these marks (seconds from the start of this clip): 0-1s: medium shot from her left side (frame-right), at eye level. 1-2s: top-down high angle from directly above and in front of her, looking down at her. 2-4.2s: extreme close-up of her face, filling the frame from eyebrows to chin, lips clearly in sync. 4.2-9.1s: close-up on her two hands while they gesture (the microphone hand and the free hand), then slowly pull back to show her face and hands together in a medium close-up. Level horizon, no shake. THE ROOM: the same office as the source: light oak built-in shelves with framed plaques (one green), plants and skincare products, a warm LED strip under the shelf, a backlit pink onyx stone wall panel, pink tulips, an acrylic cube with gold spheres, a white ceramic flower vase, a long slatted oak sideboard, light wood floor. Parts the original lens never saw continue the same office in the same style. Same warm light. FAILURE: any gesture missing, early, late, mirrored or done with the other hand; hands idle where the source moves; lipsync off; face changed or beautified; hair changed; outfit changed; the microphone disappearing or changing hands; warped hands or extra fingers; camera shake or stutter; new people; film gear; added text, captions or logos; music.