Prompt👇
Use the uploaded reference video as the EXACT VIDEO REFERENCE. I will upload THREE separate reference images, one for each person: Reference Image 1 = PERSON 1, Reference Image 2 = PERSON 2, Reference Image 3 = PERSON 3. Create a new video that follows the uploaded reference video as closely as possible. IMPORTANT — PRESERVE THE THREE PEOPLE: Use the three uploaded reference images as the identity source for the corresponding persons. Do NOT change, replace, merge, beautify, redesign, or reinterpret their faces. Keep each person’s exact facial identity, face shape, hairstyle, hairline, skin tone, body proportions and overall appearance consistent with their reference image. Each person must remain clearly recognizable as the same person from their uploaded reference image. REFERENCE VIDEO MATCH: Recreate the reference video’s exact sequence of shots, timing and pacing, camera movement, camera distance, camera height, camera angle, framing, composition, subject positioning, walking and movement direction, hand gestures, head movements, body movements, facial expressions, interactions between people, transitions between shots and overall visual rhythm. Do NOT create a different interpretation of the scene. Do NOT add new actions. Do NOT remove actions. Do NOT rearrange the sequence. Follow the reference video’s visual choreography and timing as closely as possible. PERSON CONSISTENCY: When Person 1 appears, use ONLY Reference Image 1 as the identity reference. When Person 2 appears, use ONLY Reference Image 2 as the identity reference. When Person 3 appears, use ONLY Reference Image 3 as the identity reference. Do not accidentally swap identities between shots. Do not mix facial features between the three people. Do not generate generic replacement faces. CAMERA: Match the reference video’s handheld smartphone-camera feeling, perspective, camera height, lens look, framing, movement, push-ins, pull-backs, close-ups and wider shots as closely as possible. Preserve the same vertical 9:16 composition. MOTION: Reproduce the same natural human movements shown in the reference video. Keep movements realistic and physically consistent. No exaggerated AI movements. No robotic motion. No unnatural body deformation. Hands and fingers must remain anatomically correct. VISUAL STYLE: Photorealistic live-action footage. Natural human skin texture. Realistic lighting and shadows. Realistic outdoor/environmental lighting. Natural motion blur. Natural depth of field. Real smartphone-camera image quality. No CGI. No 3D-rendered appearance. No cartoon look. No plastic AI skin. No artificial beauty filter. ENVIRONMENT: Preserve and recreate the same environment, background arrangement, atmosphere, lighting conditions, perspective and overall visual mood visible in the reference video as closely as possible. Do not unnecessarily redesign or replace the environment. IDENTITY LOCK: The reference images control WHO the people are. The reference video controls HOW they move, where they appear, the camera behavior, shot sequence, interactions and timing. Do not change the identity of any person just because the pose, camera angle, lighting or expression changes. FINAL OUTPUT: Generate one continuous vertical 9:16 photorealistic live-action video that closely reproduces the uploaded reference video with the three uploaded people replacing the corresponding people in the reference. The result should look like the same video was recreated with these three exact people, not like a new AI-generated interpretation. Maintain strict identity consistency throughout the entire video and preserve the reference video’s choreography, timing, camera behavior, composition, movements and interactions as accurately as possible. NO TEXT. NO WATERMARK. NO LOGO. NO CAPTIONS
