AI 视频生成器

Identity Lock Scene

模型

提示词

Prompt:
SINGLE REFERENCE IMAGE — CHARACTER IDENTITY:
Use Image <<<image_1>>> as the only reference image for the main character.
The reference image is used for:
* facial identity,
* facial proportions and distinctive facial characteristics,
* skin tone,
* apparent age,
* visible hairstyle and hair characteristics,
* eyewear if present,
* the person's natural masculine, feminine, or androgynous presentation.
Do not use the uploaded reference image as an outfit, body, pose, lighting, background, or environment reference.
Character appearance — critical for identity accuracy:
Preserve the exact recognizable facial identity from Image <<<image_1>>> throughout every frame.
Accurately preserve the person's natural:
face shape,
eye shape and spacing,
eyebrows,
nose,
lips,
cheeks,
jawline,
skin tone,
facial proportions,
natural asymmetry,
and distinctive visible facial details.
Keep realistic skin texture and natural human imperfections.
Do not redesign, beautify, idealize, stylize, or replace the face with a generic generated face.
The character must remain clearly identifiable as the same person throughout the entire video with no facial identity drift.
HAIR — FOLLOW THE REFERENCE IMAGE:
Use the person's hairstyle from Image <<<image_1>>>.
Preserve the visible:
hair color,
hairline,
haircut,
hair length,
parting,
texture,
curl pattern or straightness,
volume,
and face-framing hair.
Do not replace the person's hairstyle with a predetermined hairstyle.
If some portions of the hair are outside the reference crop, naturally reconstruct the missing portions based on the visible hairstyle.
The person's hair should look like their own natural hair throughout the video.
NO HAT OR HEAD COVERING:
The character must not wear a hat, baseball cap, helmet, hood, crown, or head covering.
If the uploaded facial reference includes any removable hat or head covering, do not reproduce that item.
Instead, naturally reconstruct the person's hair underneath it based on the visible hair information in the reference.
EYEWEAR:
If the person in Image <<<image_1>>> is wearing normal glasses or sunglasses, keep them on the character throughout the entire video.
Preserve the same general:
frame shape,
frame color,
frame size,
lens appearance,
and position on the face.
The eyewear must remain consistent and should not disappear between frames.
If the reference person has no eyewear, do not add any.
BODY AND OUTFIT:
Generate the body and clothing entirely from this prompt.
Use realistic natural human anatomy and proportions appropriate to the person's apparent age and visible presentation in Image <<<image_1>>>.
Automatically tailor the clothing to the character rather than forcing one fixed male or female body type.
If the reference person visually presents as male:
generate a naturally masculine version of the dark armored outfit.
If the reference person visually presents as female:
generate a naturally feminine version of the same dark armored outfit.
If the person's presentation is visually androgynous or uncertain:
use a refined gender-neutral version of the armor.
In every case, keep realistic human proportions and avoid exaggerated masculine or feminine anatomy.
DARK ARMOR OUTFIT:
The character wears a fully original dark medieval-inspired armored outfit.
The design is elegant, mysterious, practical, realistic, and physically wearable.
It should feel like handcrafted real armor photographed in live action rather than a game character or costume.
Primary materials and colors:
matte charcoal-black metal,
deep gunmetal-grey metal,
blackened steel,
weathered dark leather,
dark woven fabric,
subtle muted silver highlights.
The upper body uses layered fitted armor plates over flexible dark fabric and leather underlayers.
The armor has subtle hand-forged texture:
tiny scratches,
natural wear,
slightly imperfect edges,
realistic metal joints,
and physically accurate reflections.
The armor should naturally conform to the person's body without exaggerated anatomical shaping.
For a female character:
the armor remains elegant and appropriately tailored but practical, protective, and non-sexualized.
For a male character:
the armor uses naturally masculine tailoring without exaggerated bulk or oversized muscles.
Shoulders use compact articulated pauldrons.
Arms use fitted dark protective plates and bracers.
The waist transitions into layered dark leather and flexible armored panels.
The lower body uses dark trousers underneath articulated thigh and shin protection.
Practical dark leather boots are suitable for running across grass.
Rigid metal plates remain physically rigid.
Leather flexes naturally.
Fabric moves naturally in the wind.
Armor joints articulate naturally during walking and running.
No cape.
No hat.
No helmet.
No hood.
No crown.
No visible head accessory.
NO WEAPONS:
The character carries absolutely no weapon of any kind.
No sword.
No sword sheath.
No sword on the back.
No dagger.
No knife.
No spear.
No bow.
No arrows.
No axe.
No shield.
No gun.
No weapon belt.
No weapon holster.
Nothing resembling a weapon is attached to the character.
Both hands remain empty throughout the entire video.
Location — fully original, massive and majestic fairytale palace, photorealistic, light-toned, genuinely huge in scale, not a small manor or house:
A vast open lowland meadow with a winding S-shaped stream cutting through fresh, light, vividly green grass.
In the distance stands an enormous original fairytale palace-complex — not a single small building, but a sprawling, monumental structure that dominates the horizon, clearly scaled many times larger than a house or manor, with a footprint wide enough to suggest an entire royal residence.
The palace is enclosed by a long, low pale-stone perimeter wall with rounded corner turrets marking its boundary, giving it a sense of a fortified royal estate rather than an open building.
Behind this outer wall rises the main palace structure itself:
multiple tall, gracefully tapering towers of varying heights clustered together, crowned by one soaring, elongated central spire with a small gold finial that reaches dramatically higher than everything around it.
Each tower is capped with a smooth cone-shaped roof in soft light teal-blue with delicate thin gold trim.
Walls are pale cream and light warm-grey natural stone with real, photographically accurate texture — visible stone joints, subtle weathering, natural imperfections — grounded and believable rather than smooth, glossy, or plastic-looking.
The overall silhouette should feel vast, layered, and imposing — many rooftops and towers stacked and overlapping in depth, not a single flat façade.
The palace must feel genuinely awe-inspiring in scale even from a great distance, filling a substantial portion of the horizon rather than reading as a tiny distant object.
Windows are tall, rounded-arch, evenly spaced across multiple levels, glowing subtly with warm interior light against the pale stone.
The palace sits far in the distance near the horizon line — noticeably further away than a typical foreground landmark.
Natural atmospheric distance and gentle phone-camera depth rendering soften the distant palace slightly, while its overall silhouette, tower arrangement, layered structure, and dominant central spire remain clearly identifiable.
No drawbridge.
No dark stone.
No heavy moss or ivy overgrowth.
No logos.
No signage.
No text of any kind.
No recognizable real-world landmark.
The palace must be a completely original architectural design.
Sky and light:
Clear evening dusk sky in cool blue tones.
No orange, amber, or golden environmental cast.
Soft fading twilight light.
A few faint early stars beginning to appear.
Large dramatic sculptural clouds in cool blue-grey and pale lavender slowly drifting across the sky.
Ambient light on the grass, skin, hair, armor, and figures is cool and soft, not warm.
The dark armor catches subtle cool blue twilight highlights while remaining mostly matte and realistic.
Camera setup:
Not a selfie.
A friend films the character from a few steps away, holding an iPhone 15 Pro.
Natural handheld micro-shake.
Realistic phone motion blur.
Slight auto-exposure breathing throughout.
Natural phone-camera depth rendering.
The distant palace and background are rendered with gentle atmospheric softness while the foreground character and immediate grass remain relatively sharp.
The footage should look maximally photorealistic yet authentically captured with a real smartphone.
Full realism throughout:
real skin texture,
real hair movement,
real eyewear if present,
real metal,
real leather,
real fabric physics,
real grass,
real water behavior.
No stylization.
No illustrative appearance.
No artificial game-render look.
Scene timing:
0–7s:
Wide shot, full body clearly visible from head to feet.
The character stands alone in the vast meadow at a distance that keeps their entire figure in frame.
No crop at the waist, knees, feet, or head.
The camera faces the character from behind and slightly from the side at an angle where the palace is not yet in frame.
The palace exists just off to the left, beyond the current framing, and is completely invisible to the viewer at this point.
The character gazes toward their left, off toward the same direction where the palace actually is, with a soft, quietly mesmerized expression.
The character does not look at the camera.
Their gaze acts as the visual cue for the direction the scene will later move toward.
Cool blue dusk sky above with drifting clouds.
Wind moves the character's reference-matched hair naturally.
If the person has glasses in Image <<<image_1>>>, the same glasses remain naturally positioned on their face.
Wind also moves the flexible dark fabric and leather sections of the armor.
Rigid armor components move only through their proper joints.
Grass sways naturally.
Within this same window, a small flock of sheep briskly crosses close past the person filming from that same left side.
The camera dips naturally with a subtle handheld jolt at the moment they pass close, then rises back to reframe the character, still keeping their full body visible.
The character turns their head to briefly watch the sheep with a naturally delighted glance.
They then turn their gaze back toward that exact same left-hand direction.
They take a short breath and break into a run toward their left — exactly the same direction they had been looking toward.
All of this happens within seven seconds total.
No lingering static holds.
7–11s:
The moment the character begins running left, the friend filming begins jogging after them in that same leftward direction at an unhurried pace.
The phone remains at the same low, steady handheld height as before.
No deliberate rise.
No zoom.
No artificial push-in.
No major reframing beyond natural handheld drift.
The filmer trails several steps behind the character at a roughly constant distance.
The camera axis remains consistent with the character's running direction.
The armor behaves realistically as the character runs.
Articulated armor plates shift naturally at the joints.
Leather sections flex.
Fabric moves with the body and wind.
The person's reference-matched hair moves naturally while preserving the original hairstyle.
If eyewear is present in the reference image, it remains securely and consistently on the face.
The character's hands remain completely empty.
No sword or other weapon appears.
11–12s:
In one deliberate motion, the person filming lifts the phone upward and slightly outward, continuing to face the same leftward direction the character has been running toward this entire time.
This should look like a natural human gesture of raising the phone higher for a bigger view.
The lift happens only once.
There is no earlier gradual rise.
As the phone reaches the new higher position, the vast landscape opens up ahead along the same axis.
The palace comes into view for the first time exactly where the character's gaze had been pointing from the beginning.
12–15s:
With the phone held higher, the palace becomes visible for the first time, far away on the horizon along the same direction the character has been running.
It is softly hazed by natural atmospheric distance and phone-camera depth while remaining clearly identifiable by its monumental layered silhouette, numerous towers, pale stone walls, light teal-blue roofs, and dominant central spire.
The moment the palace comes into view, a bright magical fireworks display begins above the towers.
Vivid bursts of warm gold, white, and soft blue-violet light.
Visible glowing trails rise upward before every major burst.
Each trail blooms into a full realistic explosion.
Fireworks obey realistic physics, timing, scale, smoke behavior, light falloff, and photographic exposure while remaining bright and visually impressive.
After the single phone lift, the camera remains at the new higher position with only natural handheld drift.
No further rising.
No additional lifts.
No zoom.
No closing the distance to the character or palace.
The character remains a relatively small distant figure moving along the winding stream through the vividly fresh green meadow.
They run lightly and unhurried strictly along the grassy edge.
Their feet never touch or enter the water.
They stay a natural safe distance away from the water's edge as they follow its curve.
No splashing.
No wet boots.
The water surface trembles gently from the breeze, reflecting the cool blue sky and distant fireworks in soft rippling reflections rather than glassy stillness.
In the distance, exactly three sheep stand grazing calmly near the palace view.
All three sheep have correct realistic quadruped anatomy.
Overall mood:
Cinematic, vast open landscape, cool blue dusk light, epic slowly drifting cloudscape, gentle wind animating the character's reference-matched hair, armor underlayers, grass, and water.
The dark armored character contrasts dramatically but naturally against the fresh meadow and enormous light-toned palace.
A bright and magical yet realistic fireworks display appears over the distant palace.
Ultra-photorealistic textures.
Natural film grain.
Authentic phone-camera depth.
Natural handheld imperfections.
One deliberate single phone-lift gesture exactly as the palace view opens along the same axis the character has been looking and running toward.
Otherwise, the sequence remains a continuous steady handheld take at constant trailing distance.
Negative prompt:
wrong face,
generic face,
idealized face,
face not matching reference,
identity drift,
changing identity between frames,
incorrect facial proportions,
copying outfit from reference image,
copying reference background,
wrong hairstyle,
completely redesigned hairstyle,
wrong hair color,
wrong hair texture,
wrong hairline,
hat,
baseball cap,
helmet,
hood,
crown,
head covering,
removing reference eyewear,
glasses disappearing between frames,
changing eyewear unexpectedly,
adding glasses when none exist,
incorrectly tailored armor,
exaggerated gender characteristics,
sexualized armor,
impractical armor,
exaggerated body proportions,
cape,
sword,
sword sheath,
sword on back,
sword in hand,
dagger,
knife,
bow,
arrow,
spear,
axe,
shield,
gun,
weapon,
weapon holster,
weapon belt,
anything held in hands,
bright colorful armor,
glossy armor,
plastic armor,
chrome armor,
futuristic armor,
superhero-like armor,
cartoon armor,
oversized shoulder plates,
glowing armor,
excessive spikes,
armor clipping through body,
metal bending like rubber,
orange sky,
golden-hour environmental cast,
warm amber atmosphere,
palace visible before the camera lift,
character looks one direction and runs another,
palace revealed on opposite side,
camera zooming while character runs,
camera gradually rising while character runs,
multiple camera lifts,
repeated major reframing,
camera closing distance,
character entering water,
feet touching water,
wet boots,
splashing,
flat static sky,
heavy fog obscuring palace,
dark dull grass,
cropped body,
close-up only,
waist-up framing,
feet outside frame,
small castle,
tiny castle,
single small tower,
manor-scale building,
cottage-scale building,
palace reading as a tiny distant dot,
palace too close,
flat palace façade,
cartoon architecture,
plastic architecture,
oversaturated toy colors,
recognizable existing landmark,
logos,
signage,
text,
incorrect sheep anatomy,
more or fewer than exactly three sheep near the palace during final reveal,
stiff running,
warped hands,
extra fingers,
deformed limbs,
jump cuts,
low-frame-rate stutter,
character looking at camera,
character smiling directly at camera,
fireworks barely visible,
fireworks without rising trails,
flat artificial fireworks,
static hair,
static grass,
static water,
overly polished artificial rendering,
illustrative appearance,
long static opening,
harsh midday light.

16850 / 50000

参考图片（可选）

画面比例

时长

分辨率

生成音频2 倍积分Basic+

所需积分50

视频预览

加载中...