Гайды

Промт для видео в нейросети: рыцарь в маковом поле

Максим Солдаткиноснователь GEN202

Обновлено 11 октября 2026

Рыцарь в чёрном доспехе с шипами на наплечниках стоит в маковом поле в утреннем тумане, в опущенной руке шипастый шлем

В этом гайде показываю по шагам, как написать промт для видео в нейросети и собрать с помощью ИИ такое же видео со своим персонажем: сначала лист персонажа, потом реквизит и локация, в конце само видео. Понадобятся фото, референс костюма и нейросети: Nano Banana 2.1 для картинок и Seedance 2.5 для видео. Все шесть промтов готовые: их можно скопировать и подставить своего персонажа.

Идея и подготовка

До первого шага мне нужен мой лист персонажа в обычной одежде.

Лист персонажа в обычной одежде. Я сделал его из селфи. На листе лицо крупно и три вида в полный рост. Рост, вес и телосложение я назвал в промте числами: на селфи фигуры не видно.

Лист персонажа в обычной одежде на белом фоне: крупный план лица и мужчина в чёрной футболке в полный рост спереди, в три четверти и со спины, головы на этих трёх видах закрыты белыми прямоугольниками
Лист персонажа в обычной одежде, Nano Banana Pro. Референс — одно селфи
Промт для листа персонажа по селфи
Создай фотореалистичный лист персонажа по приложенному селфи мужчины. Четыре панели в ряд: 1) крупный план лица анфас, 2) в полный рост анфас, 3) в полный рост в повороте три четверти, 4) в полный рост со спины. Рост 187 см, вес 90 кг, спортивное телосложение. Во всех панелях точно сохрани черты лица, форму и пропорции лица, тон кожи, причёску, бороду и цвет волос; одежда как на фото. Нейтральная естественная поза, мягкое равномерное студийное освещение, чистый белый фон без теней и лишних объектов, единый масштаб и согласованность внешности между всеми видами, без стилизации, без текста и рамок.

Настройки картинок. Все ассеты я делаю в одной модели: формат 16:9, разрешение 2K.

Шаг 1. Лист персонажа

Лист персонажа — четыре панели на одной картинке: рыцарь в полный рост спереди, сбоку и со спины и крупный портрет. С него видеомодель берёт лицо, доспех и плащ.

У каждого референса в промте свой номер и своя роль: лицо, причёску и бороду я беру с листа персонажа в обычной одежде, костюм — с референса доспеха. На один лист я прикладываю один референс костюма: несколько разных модель смешивает в один.

Референс костюма я беру на Pinterest: ищу по-английски, для этого видео — «knight». Выбираю картинку, на которой костюм виден целиком и крупно.

Сначала нужен промт для листа персонажа. Его пишет чат, я использовал Claude Opus 5.5: загружаю два референса — доспех и лист персонажа в обычной одежде — и даю короткий промт:

Промт для создания листа персонажа
Составь промт для листа персонажа: мужчина с картинки 2, лицо и внешность сохрани точно как на референсе. 187 см, 35 лет, 90 кг, спортивное телосложение. Причёску и бороду оставь как на референсе.

Одежда: рыцарский доспех и плащ, как на картинке 1. Материалы и цвет доспеха словами не описывай: их задаёт картинка. Картинки загружены в этом порядке.

Раскладка листа: три вида в полный рост (спереди, сбоку, сзади) и один портрет анфас, на котором хорошо видно лицо.

Материалы и цвет костюма я в этом промте словами не описываю: описание в тексте перебивает референс, и модель рисует по словам, а не по картинке. Вид костюма задаёт референс.

Лист персонажа из четырёх панелей на светло-сером фоне: мужчина в чёрном доспехе с золотым узором и шипами на наплечниках — в полный рост спереди, сбоку и со спины в тёмном плаще, справа крупный портрет
Лист персонажа в доспехах, Nano Banana 2.1. Референсы: доспех и лист персонажа в обычной одежде

Готовый промт листа персонажа:

Готовый промт листа персонажа
SCENE IN BRIEF
Turnaround reference sheet for a single male knight: full-body from three sides (front, profile, back), with one close frontal portrait added. Purpose: lock his identity and costume.

REFERENCES IN USE
image 2 — source of identity (image 2, his character sheet): sets the face, tone of skin, eye color, proportions of the face, hairline, beard and frontal likeness. Match to the reference: 100%.
image 2 — source of identity (image 2, three-quarter view from his character sheet): sets profile anatomy, i.e. how the nose bridge and tip are shaped, how jaw and chin curve, how the lips look side-on, how the ear is shaped. Feeds Panel 2. Hair length and hairstyle are outside its control.
image 1 (image 1) — main source for the whole costume: cut of the armor, engraved pattern, where chainmail sits, rigging of the leather belts, silhouette of the cloak.
image 2 (image 2) — source for the haircut, its shape, length and color: he keeps the cut he already has.

FIXED CHARACTER TRAITS
image 2: a man of 35, 187cm in height, 90kg, athletically built, shoulders broad, muscle mass not exaggerated.

Shape of the face: a long oval, forehead high, cheekbones well defined, no heavy volume in the cheeks; it narrows to a slim chin under a short beard. Nothing artificially contoured.

Skin: fair, undertone neutral-cool; texture and pores natural and in view; a slight natural sheen across forehead and bridge of the nose; finish neither plastic nor airbrushed.

Eyes: almond-shaped, light grey-green, set a little deep beneath a straight line of the brows; lashes dark, short and natural; iris light and clear, its rim darker; the gaze direct and calm.

Eyebrows: straight and natural, medium in thickness, dark brown, not shaped, keeping to the line they naturally grow along.

Nose: bridge straight and medium in width, tip a little rounded, nostrils medium; in profile a straight line without a bump.

Lips: medium-full, a muted natural pink without gloss; the moustache frames them; at rest closed, the expression calm.

Jaw and chin: the jawline is defined beneath a beard that is short, dense and dark brown, fuller at the chin, thinner over the cheeks; the moustache joins the beard; neckline clean; in profile the line from ear to chin runs smoothly.

Distinguishing marks in view: neither moles nor scars. He is identified by the high hairline, a little receded at the temples, and by the beard's outline: on every panel both hold one shape and one position.

Ears: medium-sized, in view, protruding a little from the head; without piercings or earrings.

The face is a 100% match to the reference photos: the front views of image 2 give the frontal proportions; the three-quarter view of image 2 gives the silhouette of the profile and the ear in detail.

HAIR REQUIREMENTS
The hair is short and dark brown, longer on the top, which is brushed upward and back, with short faded sides. Texture natural and a little uneven; no salon styling, no fringe; the forehead stays clear of hair. The color remains the dark brown he originally has on the character sheet reference: not lightened, not darkened as far as black; no added length, waves or center parting. From image 2 take, exactly, the cut, how long it is and the high hairline. The armor reference image supplies no shapes of hair, helmet or crown. The short beard and the moustache stay as dense and as outlined as his character sheet shows.

SOURCE FOR ARMOR AND CLOTHING
The main source is image 1.

BUILD AND HOW HE IS SHOWN
Musculature not exaggerated. Proportions are not a bodybuilder's; bare skin is not emphasized. The armor looks like working plate protection, the torso line straight. Athletic frame, 187cm, 90kg, shoulders broad; each view presents him as modest and heroic.

PANEL ARRANGEMENT
One reference sheet with four panels. Studio background: neutral, seamless, light grey. Lighting is flat and the same on every panel, so the armor shows its true material and color.
Panel 1 — full body seen from the front; he stands neutrally, arms relaxed along the body.
Panel 2 — full body in profile view (he faces screen-left); the silhouette of nose, jaw, lips and ear is locked to image 2; pose and proportions carried over from Panel 1 unchanged.
Panel 3 — full body seen from the back; the cloak shows from behind along its full length.
Panel 4 — close portrait, straight frontal view, from the shoulders up; the whole face visible, expression calm and neutral, eyes into the camera, beard and hairline both clearly seen.

LIGHT
Front-key studio light, even and soft, low in contrast; no strong shadow cast in one direction, no backlit silhouette. The priority is to reproduce materials and colors accurately; neutral daylight (~5600K) is the same on all four panels.

WHAT MUST HOLD
All four panels: one and the same face, armor and cloak, hair of one length and color.
Facial features (shape of the face, the eyes, nose, lips, beard and ears) identical and consistent from panel to panel.
In Panel 2 the profile's silhouette is an exact match to image 2.
The hair and the beard are dark brown, the way his character sheet has them; never lighter, black or longer.
No one added, no further props; the sword keeps to its sheath at the hip.
Muscles are not exaggerated; he does not pose like a bodybuilder.
On Panel 4 the face is visible in full; neither hair nor shadow covers it.

Шаг 2. Шлем и меч

Шлем и меч — отдельный ассет: в видео рыцарь держит шлем в руке и роняет его.

Референсов два: лист персонажа задаёт металл и узор, референс доспеха — форму шлема. Форму предмета, который есть на референсе, я словами не описываю: хватает одной строки о том, с какой картинки его брать. Когда описание и референс расходятся, модель рисует по описанию.

Чёрный шлем с длинными шипами и золотым венцом, рядом меч с рукоятью в тёмной коже и ножны с гравированной оковкой, белый фон
Шлем и меч, Nano Banana 2.1. Референсы: лист персонажа и доспех
Промт шлема и меча
SCENE IN BRIEF
Props alone in the shot: a single arming sword and a single knight's helmet shown together, with no character.

REFERENCES IN USE
image 1 (character sheet, attached) — sets the metal's finish, the engraving style and, on matching parts, the leather and mail tone.
image 2 (armor picture, attached) — sets the helmet, both its shape and every detail of it.

FIXED SWORD TRAITS
An arming sword, refined and well balanced, neither bulky nor oversized.
Blade: straight, two-edged, approx. 75–80cm; a gentle taper ends in a controlled point; one fuller down the center; edges clean, without chips.
Crossguard: straight quillons with a slight taper; a small fleur-de-lis is engraved at the guard's center.
Grip: a tight spiral wrap of brown leather, matched to the leather of the armor's belt.
Pommel: a rounded wheel pommel whose flat face bears a small fleur-de-lis, engraved.
Scabbard: slim, its body of brown leather, with metal fittings at throat and chape.

FIXED HELMET TRAITS
The helmet that image 2 shows on the figure, copied exactly and presented as an object apart.

MATERIAL AND FINISH
The two objects have one and the same finish, exactly. In tone and texture the leather repeats the armor's belt.

FRAMING AND PLACEMENT
Seamless background of pure white; the sword and the helmet are the only props. The sword lies on a diagonal or stands tip down next to the helmet; the helmet sits upright at its side. Each object is visible in full, uncropped; no character, no hands.

LIGHT
Studio light, even and soft, low in contrast, without harsh shadows; neutral daylight (~5600K). Colors render consistently, so the finish of each object reads accurately.

WHAT MUST HOLD
The sword's finish is identical to the helmet's.
The leather's tone is an exact match to the belt of the armor.
The background is pure white and carries no shadows; no objects besides these two, no character and no hands.

Шаг 3. Локация

Локация — маковое поле в утреннем тумане, без людей. Я делаю её по одному промту, без референсов.

Свет я описываю словами: время суток, тон и чего в кадре быть не должно. Те же слова о свете повторяю в промте видео.

Маковое поле в утреннем тумане: красные маки на переднем плане, за ними группа деревьев и полосы леса под пасмурным небом
Локация, Nano Banana 2.1. Без референсов, только промт
Промт локации
SCENE IN BRIEF
A landscape plate left empty: a vast field of red poppies, running from the foreground at the camera to a horizon layered with fog. The light is cool and flat: early dawn, the sun not yet up. No characters, no vehicles.
LAYOUT OF THE PLACE
Camera position: at ground level, within the poppy field, at about waist-to-chest height, looking straight into the field's depth, at the horizon.
Foreground (0–3m): red poppies, dense and fully in bloom, with still-closed buds among them, on thin, wiry stems of green; tall blades of grass; low on the left, a few spikes of purple-blue lupine; morning dew in plain view on petals and along leaf edges; the weight of the stems makes them lean a little.
Midground (3–40m): over gently rolling ground the poppy field goes on; a low, clipped hedgerow crosses the frame as a horizontal line, dividing the field into a near part and a far part; by that hedgerow stands a copse of trees, compact, a dark green-black silhouette whose lower trunks the ground fog has already swallowed.
Background (40m to horizon): going back into the distance, three to four lines of trees and low hills; fog and distance make every layer paler, softer and lower in contrast than the layer nearer the camera. The furthest ridge is hardly more than a pale grey silhouette that dissolves in the mist; land meets sky without a hard edge.
Sky: an overcast dawn sky, pale blue-grey, clouds textured softly and diffusely; a gradient leads down to a band of neutral off-white just over the horizon, where sky and fog merge.
AIR AND SENSE OF DEPTH
A heavy morning fog hugs the ground and lies low over the field, densest at the horizon and thinner toward the camera, which leaves the nearest poppies crisp and detailed in full. The air between the camera and the copse of trees holds a faint haze that can be seen; it softens the copse's edges but does not hide it. No fewer than four distinct bands of depth must be clearly readable: poppies in sharp foreground, the hedgerow/field line half behind a veil, the tree silhouette softened by fog, and the furthest ridge almost dissolved. Layered this way, the picture has true aerial-perspective depth instead of a flat backdrop.
LIGHT
Light of the time before sunrise: diffused, soft, flat; the sun's disc is not visible anywhere in the frame. The overcast sky itself is the source; a directional sun is not. Color temperature: cool-neutral, about 6500-7000K. Shadows range from extremely soft to nearly absent. The poppies' red stays saturated and true, with no shift toward warm. Sky, fog and field carry no cast of orange, gold or amber anywhere. No sunrise halo anywhere in the frame, and no warm rim light on the trees.
PHYSICAL BEHAVIOR
The fog acts as a true volumetric medium: thicker the farther away it is, lower close to the ground, thin by the camera. Gravity holds dew droplets on petals and grass; the only highlights they catch are soft, cool and low in contrast. The poppy stems sway naturally and flexibly, with nothing rigid or artificially stiff.
WHAT MUST HOLD
Nowhere in the image is there a color cast of orange, amber or gold.
The sun's disc is not visible; no sunburst and no lens flare.
The sky keeps a pale, cool grey-blue and is never warm.
The background has to show no fewer than four separate layers of depth that dissolve into the fog step by step, rather than one flat line of trees.
In the foreground the poppies' red remains saturated and true; the cool ambient light does not change it.
No people, animals, structures or roads.

Шаг 4. Промт для видео

Видео собирается из трёх ассетов: лист персонажа, шлем с мечом и локация идут референсами.

В промте ассеты названы по номерам — image 1, image 2, image 3, — поэтому я загружаю их в том же порядке. Промт должен совпадать с ассетами: чего нет на ассете, того нет и в тексте. Действие расписано по секундам и по планам; состояние персонажа стоит там же — через дыхание, взгляд и походку. Если нужен определённый план, я называю его в блоке камеры одним термином, например medium close-up.

В блоке «Добавление картинок» я выбираю «Картинки-референсы» и загружаю ассеты в таком порядке: 1) лист персонажа, 2) шлем и меч, 3) локация. Затем вставляю промт:

Промт видео
SCENE IN BRIEF
At a foggy dawn a knight worn down by battle crosses a dense poppy field on his own, exhausted by the fight. The helmet, and nothing else, falls out of his hand. He is then shown from the waist up, fighting for breath; after that his knees fail and he falls back into the standing flowers, where he rests at last.

REFERENCES IN USE
image 1 (image1, character sheet) — sets his face, hair, proportions of the body and the whole identity of armor/cloak/sword, exactly the way it is shown, the sword in its sheath on his hip belt included. Match to the reference: 100%.
image 2 (image2, helmet + sword) — in this shot it sets the helmet alone: the finish of its material and its engraving.
image 3 (image3) — sets the geography of the poppy field, the misty background in layers and the cool light of an overcast dawn. No color cast of orange or amber in any part.

FIXED CHARACTER TRAITS
image 1: a man, 35yo, 187cm, athletically built; dark hair, short and combed back; a short beard, also dark; plate armor and a cloak. For the whole shot the sword sits in its sheath on the belt at his hip, just as the reference has it: nobody draws it, it does not fall, no hand touches it. The helmet is the one thing he carries; his right hand holds it loosely. The skin is textured realistically and naturally: fine pores in view, a slight natural sheen of oil on forehead and nose, a faintly uneven tone, small natural flaws of texture. Smooth plastic CG skin is ruled out, and so is a finish that looks over-retouched or waxy. Marks of battle: light dust and dirt smudge the lower greaves and the cloak's hem, a faint sweaty sheen on his face, hair a little damp and mussed, and no fresh wounds to see.

OPENING FRAME AND POSITIONS IN SPACE
The first frame the viewer sees already holds the knight's boots and lower legs, which push straight through dense poppies and grass that stand. Nowhere in the shot is there a dirt path, a trail of mud or a lane cleared for walking. He goes right through the field as it grows; as he moves, stems and flowers bend against his boots and against the cloak's hem. No empty landscape comes before him, and his reveal is not delayed. His walk leads forward, away from camera and into the field's depth. The camera follows straight behind him, low, keeping exactly to his pace.

SEQUENCE FORMAT
MULTI-SHOT SEQUENCE, CONTROLLED. Two shots, and a single HARD CUT joins them. Fades, crossfades and dissolves are not used.

LENS
EVERY SHOT USES AN 84° LENS. THIS IS NOT OPEN TO CHANGE.
Diagonal field of view of 84°, the character of a classic wide-angle lens. From start to finish the camera is physically near the subject: his boots are 0.8–1.5m away as the shot opens, and after the rise his torso is 1.5–2m away. The surroundings stay visible out to the edges of the frame; spatial context is deep and readable; straight lines are kept rectilinear, with no fisheye curve.
SHOT A: LENS LOCKED AT 84°. SHOT B: LENS LOCKED AT 84°.

SHOT A — CAMERA WORK (one continuous take)
It starts low, straight behind him, near the height of his heels; at walking pace the camera tracks the boots and the cloak's hem as they push through upright poppies. While the helmet is falling, the camera keeps this low position behind him. Next it cranes up and at the same time arcs from his back round to the three-quarter front; it finishes straight in front of him, level with his chest, and moves backward at the pace of his slowing walk. He comes into view from the belly up, together with the misty poppy field in layers at his back; the sword is still in its sheath at his hip, undisturbed. The camera holds that framing from the front while he stops, breathes, glances to the side and then sinks down when his knees give out.

SHOT A — TIMED ACTION
0:00 to 0:02 — The camera is close on the boots and lower legs, which push through dense poppies that stand upright; cleared ground cannot be seen anywhere. His stride is heavy and tired, the cloak's hem drags through the flowers, the camera follows low at his back. A light, steady breeze crosses the frame: the cloak's edges ripple and drift, and the poppy stems around him sway without pause, all in one direction of the wind.
0:02 to 0:03 — His right hand has loosened; the helmet slips out of it and lands next to his feet in the upright flowers. It hits the ground with a heavy, dull thud, and stems flatten nowhere but at the impact point.
0:03 to 0:05 — The camera cranes and arcs up and forward to a new position at the height of his chest, in three-quarter front view. Now he is seen from the belly up with the foggy field at his back, and the sword is visibly still in the sheath on his hip. The breeze keeps blowing; it lifts his hair and moves the cloak. Emptiness and tiredness show in his eyes; he breathes out visibly, his breathing labored and heavy.
0:05 to 0:06 — Exhausted, he looks away to the side, and then his knees buckle: in the frame his upper body suddenly sinks down when he goes onto his knees; the camera keeps its position.

TRANSITION: a HARD CUT into Shot B.

SHOT B — CAMERA WORK
A different angle: the camera is off to one side, a little raised, looking down across a patch of poppies that are undisturbed and stand at full height; this spot was not flattened beforehand or cleared. From his knees he tips over backward into the flowers. Throughout, the sword stays in its sheath at his hip and gets no attention.

SHOT B — TIMED ACTION
0:00 to 0:01 — The flowers he falls backward into stand at full height until the very moment he touches them. The weight of his body bends and compresses poppies and grass live, at impact and not earlier: stems snap and flatten only in the places his body really lands on, the cloak flares out and settles round him, petals come loose and the breeze carries them off.
0:01 to 0:03 — On his back, motionless, he lies among poppies crushed a moment ago; the breeze keeps moving loose strands of hair and the cloak's edge. He lets out one long breath, and at last his eyes close.
0:03 to 0:04 — One hand comes up and is laid gently upon his own belly; his breathing starts to slow.

PHYSICAL BEHAVIOR
The helmet has the mass of real steel: its fall is a short arc, it hits the ground once, with a dull impact, and any bounce is limited to a small, realistic settle. When he falls backward, the weight shifts as it really would: first the knees buckle from fatigue, then the torso drops, the cloak billows and gravity settles it, and where poppy stems are crushed they compress and only partly spring back. In both shots the poppy field stays fully grown and untouched until the very instant of contact: the ground is not cleared in advance, and no empty patch is staged.

LIGHT
Overcast light of the time before sunrise — diffused, soft and flat — in line with the location lock already established. The sun's disc is not visible; his armor, the cloak and the sky carry no cast of orange or gold anywhere. Color temperature: cool-neutral (~6500–7000K). In light diffused by fog the armor gives soft, dulled reflections, and patches of dust cut the shine here and there. Across both shots the shadows stay between extremely soft and nearly absent.

SOUND
A light wind passes steadily through the stems of the poppies; far off, the faint early-dawn birdsong. His boots press down on the dense growth of grass and flowers. The impact of the helmet is a heavy, dull thud. All the way through, his breathing can be heard, heavy and labored; when he lies down it turns into one long exhale. Dialogue, subtitles and music are absent.

WHAT MUST HOLD
The two shots share one identity of face, hair, armor and the sword in its sheath, an exact match to image 1.
In this sequence the helmet is the only thing that falls. Nobody draws the sword, it does not fall, and in each frame of the two shots it remains in its sheath at his hip.
There is no dirt path, no trail of mud and no cleared lane: in each frame where he walks, dense poppies and grass that stand upright fill the ground plane.
The skin has real, natural texture with imperfections; it is never the smooth plastic skin of CG.
Across the two shots a light wind is always visible: it moves his hair, his cloak and the poppies around him.
Up to the exact moment when he hits them, the place he falls on is full of untouched flowers that stand; no clearing is staged.
Nobody is added, nobody speaks, and no fresh wounds are visible.

Шаг 5. Тестовая генерация

Перед генерацией в хорошем качестве я делаю тест: тот же промт и те же ассеты, но на Seedance 2.0 Mini в 480p. Тест нужен, чтобы убедиться, что выходит примерно то, что я ожидал.

В тесте я смотрю три вещи: персонаж, шлем и поле взяты с ассетов; действие идёт по промту — оба плана и склейка между ними; звук на месте. Если всё так, запускаю этот же промт в хорошем качестве. Если нет, правлю промт и повторяю тест.

Тест: Seedance 2.0 Mini, 480p, 9:16, 15 секунд, со звуком

Результат

Получилось два видео по одним и тем же ассетам: на 15 и на 10 секунд. В обоих два плана и одна склейка между ними, как размечено в промте.

В первом видео шесть секунд идёт первый план: камера у земли за спиной рыцаря, шлем падает в траву, потом рыцарь показан спереди по пояс. После склейки он падает в маки и лежит до конца видео. Во втором видео я добавил в блок камеры одну строку о плане, и после склейки камера стоит сверху и берёт только голову и верх туловища. Персонаж, шлем и поле в обоих взяты с ассетов.

Чтобы сделать по этому гайду своё видео, меняются ассеты — лист персонажа, реквизит и локация — и действие в промте видео. Порядок шагов и блоки промта остаются те же.

Seedance 2.5, 720p, 9:16, 15 секунд, со звуком
Со строкой о плане в блоке камеры: «Overhead medium close-up: only his head and upper torso in frame». Seedance 2.5, 720p, 9:16, 10 секунд