1.Подготавливаем фотографии. (Раздел ИЗОБРАЖЕНИЯ - Gpt Image 2 - Соотношение сторон 9:16 - Детализация среднее - Качество 2к):
-Сначала сделал карту персонажа, первые три фотки это лицо с разных сторон, фото 4 и 5, это кепка и очки, которые я добавил себе:
Возьми внешность человека из загруженных фото и Create a professional photorealistic character reference sheet (character turnaround sheet) of the SAME person from the uploaded reference photo(s) 1-2. If multiple reference images are provided, carefully analyze ALL of them together to build a complete and accurate understanding of the person's appearance, face, hair, clothing and accessories from different angles, and use this combined information to generate the turnaround consistently.
CRITICAL — preserve identity 100%:
Keep the face absolutely identical to the reference: same facial features, same face shape, same eye shape and color, same nose, same lips, same eyebrows, same skin tone, same age, same makeup (if any).
Keep the hair 100% identical: same color, same length, same hairstyle, same parting, same texture, same volume.
Do NOT beautify, do NOT change proportions, do NOT make the person younger or older, do NOT alter ethnicity. The person must be perfectly recognizable in every shot.
CRITICAL — preserve outfit and accessories EXACTLY as in the reference:
Keep the clothing 100% identical to the reference photo(s): same garment, same color, same cut, same fabric, same details, same prints, same stripes, same logos, same brooches/pins, same neckline. Do not add a jacket, blazer, scarf, hat, sunglasses or any other clothing item that is not visible in the reference.
Keep all accessories 100% identical: same earrings, same necklace, same rings, same watch, same brooch — exactly as shown in the reference.
Do NOT add any new accessories. Do NOT remove existing accessories.
If only the upper body is visible in the reference, logically extend the outfit downward in a neutral, realistic way that matches the style and color palette of the visible clothing (for example simple matching trousers/skirt and plain shoes), keeping it minimal and non-distracting.
Style: hyper-realistic photography, studio quality, 8K, sharp focus, professional color grading, cinematic soft lighting, shot on Hasselblad, 85mm lens, clean neutral light grey studio background, soft even lighting without harsh shadows. Consistent lighting and background across all 6 shots.
COMPOSITION — single reference sheet, overall canvas aspect ratio 16:9 (horizontal), divided into a strict, clean grid with thin neutral dividers between panels:
LEFT HALF of the canvas (occupies 50% of the width, full height) — split vertically into 2 EQUAL TALL VERTICAL PANELS (full-body shots, head to toes, character fills the panel from top to bottom with small headroom):
Left vertical panel — Full body, FRONT view: standing straight, natural relaxed confident pose, arms relaxed at sides, looking at camera.
Right vertical panel — Full body, BACK view: exact same standing pose seen from directly behind, same scale and same vertical alignment as the front view, showing the back of the outfit and hair.
RIGHT HALF of the canvas (occupies the other 50% of the width, full height) — split into a 2x2 GRID of 4 EQUAL PANELS (close-up portraits, head and shoulders, tight framing on the face):
3) Top-left panel — Face CLOSE-UP, FRONT view: head and shoulders, looking straight at camera, neutral calm expression.
4) Top-right panel — Head CLOSE-UP, BACK view: directly from behind, showing the back of the head and hairstyle in full detail.
5) Bottom-left panel — Face CLOSE-UP, THREE-QUARTER view from the RIGHT side (face turned slightly so we see more of the right side of the face).
6) Bottom-right panel — Face CLOSE-UP, THREE-QUARTER view from the LEFT side (face turned slightly so we see more of the left side of the face).
All 6 shots must show the SAME person, SAME outfit, SAME accessories, SAME hairstyle, SAME lighting, SAME background — only the camera angle and framing changes. Consistent character design across all frames, like an official movie character turnaround or a fashion lookbook.
Layout rules: keep the 16:9 master frame; left side = 2 vertical full-body panels of equal width and full height; right side = 2x2 grid of equal close-up panels; thin clean separators between all panels; no text, no labels, no watermarks, no logos, no extra characters, no duplicates of the person inside a single panel.
Ultra-detailed, photorealistic skin texture, realistic fabric texture, realistic hair strands, natural shadows. The final result must look like six real photographs of the same real person taken in one studio session, assembled into one professional 16:9 character reference sheet.
Он должен быть в кепке из фото 4 и в очках из фото 5
-Создаем начальный кадр для видео:
Preserve the person's face and identity exactly as in the reference photo — do not change facial features, skin tone, or identity. Change only the following: place the person into a candid street photography scene, photographed from a very far distance with a long telephoto lens (400-600mm equivalent), creating strong background compression typical of surveillance or paparazzi-style photography. Scene: a busy pedestrian sidewalk somewhere in a Russian city, with many people walking behind and beside the subject, as part of a dense crowd flow. The subject is walking roughly in the middle of this crowd, positioned near the center of the frame but occupying only about 25% of the frame's height — small within the overall composition, emphasizing distance. The subject must be shown fully from head to toe, full-body, unobstructed — leave a clear gap around the subject so no other pedestrians overlap or block any part of their body, even though the crowd is dense elsewhere in the frame. Camera style: shot as if with a hidden or long-range telephoto camera — flattened perspective, shallow depth of field with soft background blur (bokeh) on distant buildings and crowd, slight haze/atmospheric compression typical of long lenses, natural daylight, documentary/candid feel, not posed, subject unaware of the camera, looking forward or slightly to the side naturally. Do not add text, watermarks, or logos. Keep lighting realistic and consistent with an overcast or soft daylight urban environment.
2.Создаем видео. Переходим в Раздел Видео - Выбираем Seedance 2.0 - Режим Pro - 720p (если мажоры, то 1080р) - Соотношение сторон 9:16 - Длительность 13 сек. Загружаем начальное, карту персонажа, видео-референс танца (мой можете взять из тг-канала), после вставляем промпт и вручную удаляем все "@image1" , "@image2" и "@video1" и заново их прописываем (когда будете вводить "@" там можно будет выбрать нужный кадр), так они привяжутся к промпту:
ПРОМПТ ДЛЯ ВИДЕО:
@Image1 as the first frame and full scene reference — a city street sidewalk with pedestrians around, keep the environment, lighting, background and depth of field exactly as in @Image1. The person standing in the center of the frame references @Image2 for exact appearance: same face, same hairstyle, same clothing and accessories — keep identity fully consistent throughout. Scene first, then action: the person in the center starts dancing on the sidewalk, their choreography, movements and rhythm fully reference @Video1 — replicate the dance from @Video1 exactly, fluid, confident and on the beat. Background music and rhythm reference @Video1's audio. Around the dancer the world keeps living naturally: pedestrians walk past behind and cross in front of them in the foreground, going about their day, ignoring the dancer, while the person keeps dancing in the center of the frame. Camera: locked static medium-wide shot on a tripod, no movement, letting the dance carry the frame while pedestrians occasionally pass through the foreground. Cinematic quality, natural lighting matching @Image1, fine film grain. Constraints: maintain consistent facial features, hairstyle, clothing and accessories throughout, no warping or distortion, natural body proportions, smooth fluid motion, stable background, no flickering, no ghosting.
