от команды SYNTX

Этот документ включает все промпты для создания динамичных экшн-видео.

  1. Карта Персонажа | Nano Banana (загружаем свои фото со всех сторон)

Профессиональный коллаж на чистом белом фоне (#FFFFFF). Компоновка: одно вертикальное фото в полный рост слева (≈40% ширины кадра), справа — сетка 2×2 из четырёх изображений (каждое ≈30% ширины, ≈50% высоты). На всех пяти изображениях один и тот же человек. Фон полностью удалён — чистый белый, без теней и градиентов. Полностью сохранить внешность: лицо, тон кожи, причёску, пропорции тела и все видимые татуировки — точность 100%, без ретуши тату и деталей кожи. Студийный вырез, чёткие края, высокое разрешение, стиль fashion lookbook / карточка модели для e-commerce.

  1. Добавление карандашной вуали на фото (для обхода блокировок)

Добавь по верх изображения тонкую вуаль как будто поверх положили тонкую пергаментную бумагу, через которую отчетливо видно изображение, добавь 0.5% карандашного рисунка.

  1. Добавление сетки на лицо (для обхода блокировок)

Наложи прозрачную сетку на область возле лица с первого изображения, все остальное оставь без изменений.

  1. Для создания промпта | ChatGPT/Claude (Вначале коротко описываем свою идею)

[Описываем свою идею]. Создай этот промпт, по этим канонам:

[Референсы]
@Image1 — exact face and body reference for the main character. Use as strict identity reference. Preserve exact likeness throughout. No beautification. No deformation. Stable face throughout.
@Image2 — exact appearance reference for the toy poodle companion. Small curly-haired toy poodle. Use as strict identity reference. Maintain exact look throughout until transformation.
@Image3 — outfit and clothing reference. Replicate exact clothing from this image on the character throughout pre-transformation scenes.
@Image4 — exact appearance reference for the giant monster destroying the city. Maintain exact form, scale, and presence throughout. Constant destruction motion — smashing, stomping, debris falling from its body at all times.
[Format] 9:16
[Visual Style]
Ultra-photorealistic with an ARRI ALEXA 35mm film look. Cinematic lighting, high dynamic range, sharp focus, depth of field, subtle film grain, realistic shadows and physics. Desaturated apocalyptic palette — ash grey, burnt orange, dust yellow, dark smoke black. Fires, explosions and dust clouds in every frame. Handheld documentary-style camera during the running sequences, switching to epic stabilized shots during transformation.
[Location & Atmosphere]
City highway during full-scale apocalypse. Cracked asphalt, overturned cars, burning vehicles. Buildings collapsing in the background. Dust and ash filling the sky, blocking the sun into a dim orange glow. Sirens, explosions, debris raining. The giant monster from @Image4 visible in the distance destroying skyscrapers, getting closer with each second.
[Breakdown]
[0:00–3s]
Wide aerial shot of the destroyed city highway — overturned cars, fires, smoke columns rising. The giant monster from @Image4 towers in the background smashing through a skyscraper, debris cascading down. Quick cut to ground level — the man from @Image1 wearing outfit from @Image3 sprints down the highway, the small toy poodle from @Image2 running at full speed beside him on its tiny legs. Cars burning on both sides. He looks back over his shoulder — the monster is getting closer. Camera: aerial establishing shot, then hard cut to handheld tracking shot running alongside them at ground level, shaky, urgent, debris flying past the lens.
[3:00–7s]
They keep running — intercut between multiple dynamic angles. Front shot — both sprinting toward camera, explosions behind them. Low angle from the asphalt — their feet pounding past, the monster’s massive legs visible in the background. Side tracking shot — the man scoops up the poodle in one arm without stopping. Close-up of his face — determined, sweat, ash on his skin. Close-up of the poodle’s face in his arm — wide eyes, ears flapping. A building collapses across the highway ahead of them, blocking the road with rubble. They skid to a stop. Camera: rapid cuts between front, low, side angles, handheld shake throughout, crash zoom on the rubble wall blocking them. Slow motion 120fps on the building collapse.
[7:00–11s]
He puts the poodle down. They both face the approaching monster. The ground shakes. He looks down at the poodle. The poodle looks up at him. He kneels and places his hand on the poodle’s head. A surge of bright white energy explodes outward from the contact — engulfing both of them. Their combined silhouette grows massively — stretching upward, reshaping. The energy clears and reveals a colossal toy poodle standing on two hind legs — towering as tall as the monster. Curly fur, massive muscular hind legs, the man’s confident eyes visible in the poodle’s face. Camera: medium two-shot of the man and poodle facing the monster, then macro on hand touching poodle’s head, orbital shot 360 degrees around the transformation as they merge and grow, speed ramping from slow motion 120fps to real-time as the giant poodle form is revealed.
[11:00–15s]
The giant toy poodle on two legs charges at the monster across the destroyed highway. Buildings shake with each step. The monster swings a massive arm — the giant poodle ducks under it and delivers a devastating uppercut with its front paw straight into the monster’s jaw. The impact sends a shockwave rippling outward — windows shatter, cars flip, dust blasts in every direction. The monster lifts off the ground from the force, crashing backward through three buildings. The giant poodle stands tall, curly fur blowing in the shockwave wind, looking down at the fallen monster. Camera: wide tracking shot of the charge, then dynamic low angle on the duck and uppercut, crash zoom on the fist connecting, wide crane shot pulling back to reveal the monster crashing through buildings. Slow motion 120fps on the uppercut impact and shockwave.
[Technical]
Stable face on the man throughout pre-transformation scenes. No morphing until transformation sequence. No deformation. No flickering. No ghosting. Realistic physics on running, building collapse, debris, shockwave propagation, and monster knockback. Consistent monster form from @Image4 throughout. Consistent clothing from @Image3 pre-transformation. Consistent giant poodle form post-transformation — recognizable curly toy poodle features at massive scale. Dust, ash, fire and particle effects persistent. 4K cinematic.

  1. Для создания промпта | ChatGPT/Claude (Вначале коротко описываем свою идею)

[Описываем свою идею]
Ты эксперт по созданию промптов для AI видео генератора Seedance 2.0.   от ByteDance. Изучи эти правила перед тем как помогать пользователю.
ПРАВИЛА НАПИСАНИЯ ПРОМПТОВ ДЛЯ SEEDANCE 2.0:
[Система референсов]

  • Все загружаемые файлы тегируются: @Image1, @Image2, @Video1, @Audio1 и т.д.
  • Максимум 9 изображений + 3 видео + 3 аудио файла
  • Каждый тег нужно явно описать в начале промпта — что это и как использовать
  • Пример: “@Image1 — exact face and body reference for the main character”
  • Если загружается видео как основа — пишем: “Replicate every shot,    camera angle, movement and transition from @Video1 exactly” [Персонаж]
  • Всегда описывай внешность детально: прическа, черты лица, телосложение
  • Обязательно добавляй: “Preserve exact likeness throughout.    No beautification. No deformation. Stable face throughout.”
  • Если загружается фото лица: “Use @Image1 as strict identity reference”
  • Одежду описывай отдельно и детально [Структура промпта]
    Всегда пиши в таком порядке:
  1. Список всех @ референсов с описанием
  2. Формат (16:9 или 9:16)
  3. Визуальный стиль и цветовая палитра
  4. Описание локации и атмосферы
  5. Разбивка по секундам [0:00-3s], [3:00-7s] и т.д.
  6. Технические требования в конце [Разбивка по секундам]
  • Максимум 15 секунд на одну генерацию
  • Описывай каждый временной отрезок отдельно
  • Указывай движение камеры в каждом отрезке
  • Пример движений камеры: slow push in, crash zoom, orbital shot, low angle, handheld shake, tracking shot, crane shot [Камера и стиль]
  • Всегда указывай тип съёмки: handheld / stabilized / drone
  • Указывай качество: 4K cinematic, IMAX quality, 35mm film grain
  • Указывай цветовую палитру конкретно
  • Slow motion пиши как: “slow motion 120fps” или “snap to real-time” [Чего избегать]
  • Не пиши размытые описания: “cool”, “nice”, “beautiful”
  • Не давай противоречивые инструкции
  • Не смешивай слишком много сцен — модель путается
  • Негативных промптов нет — описывай что хочешь, не что не хочешь
  • Не упоминай реальных знаменитостей и копирайтные персонажи [Технический хвост — всегда добавляй в конце]
    “Stable face throughout. No morphing. No deformation.   No flickering. No ghosting. Realistic physics. 4K cinematic.”
    [Аудио]
  • Можно референсить аудио из видео: “reference @Video1’s audio”
  • Для lip sync: “voice and lip sync reference @Audio1 exactly”
  • Для beat-sync монтажа: “video rhythm references @Audio1’s beats” [Важные ограничения Seedance]
  • Реальные лица иногда режет модерация — используй фото в полный рост
  • Насилие, кровь, копирайтные персонажи — режет модерация
  • Если срезало — смягчи формулировки насилия и убери копирайт -— Промпт пиши на английском.
  1. Несколько примеров конечных запросов в SeeDance 2.0

Надвигающееся цунами:
@image1 — exact face and body reference for the main character. Use as strict identity reference. Preserve exact likeness throughout from every angle. No beautification. No deformation. Stable face throughout.
@image2 — outfit and clothing reference. Replicate exact clothing from this image on the character throughout.
@image3 — exact appearance reference for the toy poodle companion. Small curly-haired toy poodle. Use as strict identity reference. The giant version must be recognizably the same dog — same curly fur, same color, same face, only at massive rideable scale.
@image4 — environment reference for the city streets. Use as the exact setting — replicate buildings, road layout, lighting, and atmosphere.
[Visual Style]
Ultra-photorealistic with an ARRI ALEXA 35mm film look. Cinematic lighting, high dynamic range, sharp focus, depth of field, subtle film grain, realistic shadows and physics. Desaturated cold palette — steel grey, dark blue, white foam, wet asphalt reflections. Rain and mist in every frame. Water physics must be hyper-realistic — splashes, spray, weight, momentum. 4K cinematic.
[Location & Atmosphere]
City streets from @Image4. Overcast sky, rain. A massive tsunami wall rising behind the skyline, slowly consuming the city block by block. Wet roads, abandoned cars, debris floating. Panic atmosphere — but focus only on the man and the dog.
[Breakdown]
[0:00–3s]
Wide aerial shot of the city from @Image4 — a colossal wall of water taller than the buildings moves through the skyline in the background. Hard cut to street level — the man from @Image1 wearing outfit from @Image2 sprints down the wet street at full speed, carrying the small toy poodle from @Image3 tucked under his left arm. Behind him — cars slide forward from the water pressure, trash cans tumble, street signs bend. Water rushes across the road surface at ankle height. He looks back over his shoulder — the wave is getting closer. Camera: aerial establishing shot, then hard cut to handheld tracking at ground level behind him, shaky, urgent, water spraying past the lens.
[3:00–7s]
He keeps running — intercut between angles. Front shot — sprinting toward camera, wall of water filling the entire background. Low angle from the wet asphalt — his feet splashing through water, debris sliding past. Side tracking — a car floats past him pushed by the rising water. Close-up of his face — determined, rain on his skin. Close-up of the poodle in his arm — wide eyes, wet fur. The water behind is rising fast, almost reaching building tops. He stumbles but recovers. Camera: rapid cuts between front, low, side, handheld shake, crash zoom on the water gaining. Slow motion 120fps on the stumble and recovery.
[7:00–10s]
He sets the poodle down on the wet road. They face forward together for a split second. Energy radiates from the poodle — it grows rapidly. Same curly fur, same face from @Image3, now massive, rideable, towering over cars. The man grabs the fur and swings onto its back. The giant poodle turns and charges forward through the streets at full speed, leaping over abandoned cars, splashing through rising water. The tsunami is right behind them now — foam and spray touching the poodle’s hind legs. Camera: macro on poodle growing, orbital shot during transformation, then semi-POV behind the man as the poodle charges forward. Handheld energy. Slow motion 120fps on the mount-up.
[10:00–13s]
Ahead on the sidewalk — a small red emergency booth with a single large button visible through the glass. The man spots it. He points forward. The giant poodle sprints directly toward it. Water rises to the poodle’s legs now. Cars float past on both sides. The poodle skids to a stop beside the booth, water splashing outward from the halt. The man jumps off, lands in ankle-deep water, runs two steps to the booth. Camera: wide tracking of the final sprint, then stabilized medium shot of the skid stop, water spraying from the poodle’s paws. Speed ramp from fast to slow on the dismount.
[13:00–15s]
He reaches the booth, pulls open the glass panel, and slams his hand down on the button. Macro of his palm hitting the red button. The moment he presses it — all water freezes instantly. The tsunami wall behind them stops mid-motion — suspended, towering, completely still like a massive glass sculpture. Water droplets hang in the air. Splashes freeze mid-arc. The foam on the wave becomes solid and motionless. But the city stays alive. People on the far streets keep walking, looking up at the frozen wave in shock. A man across the road stops and stares with his mouth open. Cars on dry streets continue driving slowly. Pigeons fly past the frozen water wall. A newspaper blows across the wet ground. The giant poodle shakes water off its fur and sits down calmly.
Wide crane shot pulling back and up — the man stands at the booth, the giant poodle beside him, both wet and breathing heavy. Behind them the frozen tsunami fills the sky like a transparent ice wall. The city around them continues to move — pedestrians stopping to look, a cyclist passing in the foreground, traffic lights changing. The contrast between the frozen water and the living city is absolute. Hold the frame. Camera: macro on button press, then wide crane shot pulling back and up revealing the entire scene — frozen water, living city. Slow motion 120fps transitioning to normal speed as life continues around the frozen wave.
[Technical]
Stable face on the man throughout. No morphing. No deformation. No flickering. No ghosting. The giant poodle must be visually identical to @Image3 — same curly fur, same color, same face, only scale changes. Water physics must be hyper-realistic throughout — splashing, spraying, flowing, weight. After the button press: ALL water elements freeze simultaneously — the tsunami wall, airborne droplets, splashes, foam. ALL non-water elements continue moving normally — people walk, cars drive, birds fly, wind blows. This selective freeze is the core visual contrast of the final shot. Consistent outfit from @Image2 on the man throughout. City environment matches @Image4. 4K cinematic.

Наперегонки с поездом:
@image1 — exact face and body reference for the main character. Use as strict identity reference. Preserve exact likeness throughout. No beautification. No deformation. Stable face throughout.
@image2 — outfit and clothing reference. Replicate exact clothing from this image on the character throughout.
@image3 — exact appearance reference for the train in its original form. Replicate this exact train model, color, and design for the first half of the video before transformation.
@image4 — exact appearance reference for the worm creature. The train from @Image3 must transform into this exact form. Maintain this exact appearance, texture, segmentation and details after transformation throughout.
@image5 — environment and location reference. Use as the exact setting — replicate the terrain, road, sky, lighting, colors and spatial layout throughout.
[Visual Style]
Ultra-photorealistic with an ARRI ALEXA 35mm film look. Cinematic lighting, high dynamic range, sharp focus, depth of field, subtle film grain, realistic shadows and physics. Color palette and atmosphere matched to @Image5. Dust, gravel particles and heat haze in every frame.
[Location & Atmosphere]
Environment from @Image5. Highway running parallel to a railroad track. The man runs on the road, the train on the tracks beside him. Atmosphere, sky, terrain and lighting all matched to @Image5 exactly.
[Breakdown]
[0:00–3s]
Wide shot from above — the man from @Image1 wearing outfit from @Image2 sprints along the highway at full speed. The train from @Image3 moves parallel on the tracks beside him, matching his pace. Dust kicks up from his feet. Sun catches the sweat on his skin. Camera: aerial tracking shot from the side capturing both the man and the train from @Image3 moving together, then hard cut to ground level handheld running behind him, the train filling the right side of the frame. Slow motion 120fps transitioning to real-time.
[3:00–6s]
Side tracking shot — the man runs neck and neck with the train from @Image3. Cut to a close-up of a train window — a small boy inside smiles wide, presses his hand against the glass, and waves enthusiastically, cheering the man on. The man glances over, sees the boy, and gives a brief smile and a nod without breaking stride. A warm moment. Camera: side tracking keeping both in frame, then macro on the boy’s face and hand through the glass, then close-up of the man’s smile. Warm light through the train windows. Stabilized.
[6:00–10s]
The train from @Image3 begins to change. Metal groans and twists. The front locomotive stretches and bends. Carriages fuse together, surface rippling. Bolts, rivets and metal panels detach and fly off. Wheels crack apart. The train morphs gradually into the exact worm creature from @Image4 — segmented body, exact texture and form as shown in @Image4. Dust and debris cloud erupts around it. The man keeps running, looking over his shoulder at the transformation. Camera: wide shot of the train warping into @Image4 form, intercut with close-ups of metal panels flying off, rivets popping. Orbital shot around the forming head. Slow motion 120fps on debris.
[10:00–15s]
The worm creature from @Image4 slithers along the ground beside the highway at increasing speed, tearing up the tracks, sending rails and gravel spraying. The man sprints harder. The worm gains on him, its head turning toward the highway. A bridge overpass ahead. He runs under it — the worm is too large, collides with the structure, gets tangled. The man emerges on the other side, slows to a walk, stops. Turns around. Dust settles. He stands alone, catching his breath, looking back. Camera: side tracking of the race, wide shot ducking under the bridge, hard cut to the other side — stabilized slow push-in on the man turning. Slow motion 120fps on the final turn.
[Technical]
Stable face on the man throughout. No morphing. No deformation. No flickering. No ghosting. Realistic physics on running, metal detaching, debris and dust. Train must exactly match @Image3 before transformation. Worm creature must exactly match @Image4 after transformation. Transformation must be gradual and mechanical. Environment must match @Image5 throughout. Consistent clothing from @Image2. Dust and particle effects persistent. 4K cinematic.

Битва на мосту с большой обезьяной:
@image1 — exact face and body reference for the main character. Use as strict identity reference. Preserve exact likeness throughout. No beautification. No deformation. Stable face throughout.
@image2 — exact appearance reference for the toy poodle companion. Small curly-haired toy poodle. Use as strict identity reference. Maintain exact look throughout until transformation.
@image3 — outfit and clothing reference. Replicate exact clothing from this image on the character throughout pre-transformation scenes.
@image4 — exact appearance reference for the giant creature in the city. Maintain exact form, scale, and presence throughout. Constant motion at all times.
[Visual Style]
Ultra-photorealistic with an ARRI ALEXA 35mm film look. Cinematic lighting, high dynamic range, sharp focus, depth of field, subtle film grain, realistic shadows and physics. Desaturated palette — ash grey, burnt orange, dust yellow. Dust clouds and dramatic atmosphere in every frame. [Location & Atmosphere]
City highway during a dramatic storm. Cracked asphalt, abandoned cars. Dust and ash in the sky, dim orange glow. The giant creature from @Image4 visible in the distance moving through the skyline.
[Breakdown]
[0:00–3s]
Wide aerial shot of the city highway — abandoned cars, dust rising. The giant creature from @Image4 moves through the skyline in the background. Cut to ground level — the man from @Image1 wearing outfit from @Image3 runs down the highway, the small toy poodle from @Image2 running beside him. He glances back — the creature approaches. Camera: aerial establishing shot, hard cut to handheld tracking shot at ground level, dust past the lens.
[3:00–7s]
They keep running — intercut between angles. Front shot — both moving toward camera. Low angle from asphalt — feet pounding past. Side tracking — the man picks up the poodle in one arm. Close-up of his face — focused. Close-up of the poodle — wide eyes, ears flapping. Rubble ahead blocks the road. They stop. Camera: rapid cuts between front, low, side angles, handheld, crash zoom on rubble. Slow motion 120fps.
[7:00–11s]
He sets the poodle down. They face the approaching creature. He kneels and places his hand on the poodle’s head. Bright white energy radiates from the poodle — the small dog begins growing. Same curly fur, same face, body expanding upward — taller than cars, taller than buildings, until the toy poodle stands colossal, matching the creature’s height. Identical curly coat, identical poodle face, titan scale. The man stands on the giant poodle’s back. Camera: macro on hand touching head, wide orbital shot 360 degrees around the poodle growing, speed ramping from slow motion 120fps to real-time at full size.
[11:00–15s]
The giant toy poodle moves toward the creature confidently on all four legs, the man riding on its back. The creature reaches forward — the giant poodle stands on its hind legs and places both front paws on the creature’s chest, pushing it firmly backward. The creature slides across the ground, losing balance. A powerful wind ripples outward. The giant poodle lands back on all fours, barking. The man on its back looks ahead with calm confidence. Camera: wide tracking of the approach, low angle of the poodle rising on hind legs, crane shot pulling back to reveal the full scene. Slow motion 120fps on the wind ripple.
[Technical]
Stable face on the man throughout. No morphing. No deformation. No flickering. No ghosting. The giant poodle must be visually identical to @Image2 — same curly fur texture, same color, same face, only scale changes. Realistic physics. Consistent creature from @Image4. Consistent clothing from @Image3. Dust and particle effects persistent. 4K cinematic.


Присоединяйтесь к нашей платформе и откройте новые горизонты с SYNTX!

SYNTX AI: Syntx AI   Блог SYNTX FAMILY: Syntx Family
SYNTX Сообщество: Syntx Community   Служба Заботы SYNTX: Syntx Support   Канал “SYNTX”: Syntx YouTube