提示词库
万相 3.0 提示词，每条附生成的视频。
每条提示词都原样摘自模型厂商的指南或创作者的帖子，旁边是它生成的成品。点开卡片看全文。
万相 3.0 是阿里巴巴的视频模型，在阿里云百炼（Model Studio）上提供 wan3.0-video 和 wan3.0-video-prime 两个版本。单段 2 到 30 秒，每秒 30 帧，分辨率 480P、720P 或 1080P，默认同时生成对白、音乐和音效。除了文字、首帧和首尾帧，它还能用最多 10 张图、5 段视频和 5 段音频作参考，或者一份文档、一个网页，合计最多 20 份，也能改视频和续写视频。下面的提示词都是创作者在 X 上分享的，每条附成片。
来源：Wan 3.0 video generation guide · Wan 3.0 prompt guide
MiniMax H3、万相 3.0、Seedance 2.5、可灵 4.0 的提示词，创作者怎么写，实测 →
更新于 2026-11-03 · 共 60 条，最新的在前
其他模型页：GPT Image 2.6 · Nano Banana 2.1 · Seedance 2.5 · 可灵 4.0 · MiniMax H3
工业厂房中的直播风射击游戏实录
熊猫池塘场景，竹林与樱花树
身穿奶油色西装的女子阅读紧急派对便条
杂技演员穿越火环跃上奔马
头戴毛巾的女子在夜间厨房煎鸡胸肉
工位女孩在14次硬切中换造型
风中霓虹灯下的 Korin 面部特写
火焰剑客与紫电恶魔在竹林暴雨中决斗
夜间街头帮派斗殴，带游戏界面
两名男子在雨夜街道对峙
粉色开衫女子在宴会厅与刺客搏斗
英国游客在孟买街头吃 pav bhaji
红色阿迪达斯外套女子在酒店前看手机
留胡子的男子在黄金时刻的公寓里推广耳机
红色皮卡驶过沙漠峡谷公路
渔夫在木质码头追赶巨蟹
赛博武士在雨夜屋顶与蜘蛛臂机器人对决
女主角被传送进开放世界游戏城市
首尔少女画中的花与纸飞机成真
红衣侠女掠过喀斯特峡谷河面
虚构相机 LENS ONE 置于深色台面
红色真丝吊带裙奢华时装广告片
剑士在漂浮城市上空与巨型骑士交战
紫色超跑在霓虹湿润街道上飙驰
自拍博主试吃爆红餐厅的招牌菜
傍晚村庄水井旁的老人
黑长发女生在日本街头回望镜头
战士与堕落天使在漂浮大教堂对决
黑色冷幽默博主在院子里讲解AI视频识别
沙漠赛车在火山爆发与峡谷崩塌中飞驰
夕阳悬崖上的动漫少女俯瞰海岸
女孩犹豫后起跳完成换装
未来概念赛车黎明沙漠中冲击马赫1
巨型空间站悬于条纹气态巨行星之上
无人机与送餐机器人向储物柜投递餐食
父亲穿过车站追赶登上列车的儿子
殖民车辆冲向坍塌的火星穹顶
健身 Vlog 女生完成一组训练动作
焦点从女子移向雾中旋转木马
昏暗房间里泪流满面哭泣的中年女子
由同心凹槽构成的银色 ECHO 字标
安静卧室里熟睡的猫耳女孩
动漫法师召唤飞船穿越云端传送门
蒙面女侠蛛丝摆荡撞上出租车
带抓拍闪光的超近景时尚人像
夜晚码头闺蜜同游与跳水
咖啡馆柜台前身穿驼色大衣的男人
没有匹配的条目，换个短一点的词试试。
值得照着改的 万相 3.0 提示词
本页每个分类里点赞最多的一条，全文贴出，并说明它为什么好用。复制一条，换掉主体就能用。
带抓拍闪光的超近景时尚人像
提示词把15秒拆成五个时间段，每段写明运镜、眼神和闪光时机。同时列出真实的皮肤瑕疵并禁止塑料感，让人脸保持摄影质感。
Generate this video: Create a 15-second, 9:16 vertical screen, ultra-high-definition 4K, 30fps, cinematic realistic portrait quality fashion portrait short film.
The main subject of the frame is a young Asian woman around 20-25 years old, with a high-end face—cool with a touch of youthful greenness, delicate features but not overly perfect, smooth facial contours, natural nose bridge, full lips, eyes with depth, overall vibe like a car model or a close-up candid shot by a photographer at a fashion event.
No influencer vibe, no exaggerated plastic surgery face, no plastic skin, no excessive skin smoothing, no overly heavy fake lashes, no exaggerated highlights, no anime feel.
The character's skin must be realistic: clearly visible pores, subtle skin texture, light foundation caking, faint acne marks, very shallow pockmarks, under-eye fine lines, natural contours on nose wings and cheeks, lip lines, a few stray hairs and hair edge wisps. Makeup is refined but imperfect high-end subtle look: base slightly adherent but not flawless, clean eye makeup, naturally realistic lashes, lips with a subtle sheen.
Hair is black or dark brown straight long hair, with natural stray strands and a slight messy feel, not too deliberate.
Scene and Atmosphere
Scene is a high-end fashion event venue / auto show activity area / luxury brand launch event close-up portrait atmosphere, but the background must be blurred, not highlighting specific brands or text. Background can include soft exhibition hall lighting, minimal out-of-focus neon reflections, blurred crowds, blurred car body highlight reflections or metallic texture backgrounds, overall atmosphere high-end, restrained, fashionable.
No subtitles, no stickers, no watermarks, no readable text, no exaggerated effects, no illogical flashy lights.
Lighting Requirements
Lighting must conform to real-world logic, like a real photographer shooting close-up at the event site.
Main light is soft but directional facial key light, coming from slightly off to the front upper side, creating natural facial light and shadow layers.
Fill light is very subtle, retaining some shadows to enhance facial dimensionality.
Environment can include soft cool-toned rim light or auto show site reflection light to make the face more three-dimensional.
Throughout, allow 2 to 3 brief flash occurrences; flashes must feel like natural on-site photography flashes, quickly lighting up only in certain moments, causing temporary enhancement of facial highlights, more obvious pupil catchlights, brighter skin reflections, but no constant strobing, no illogical flashing.
When flashes occur, it should feel like a photographer's candid shot: lights up once and done, realistic light ratios, face not overexposed, skin details still visible.
Camera and Movement (Top-Tier Cinematography)
Overall, use ultra-close portrait movement, with the lens always orbiting the face to highlight eyes, nose bridge, lips, cheeks, hair strands, and skin texture.
Cinematography language is high-end, fluid, restrained, with a fashion ad and backstage photography candid feel.
No heavy shaky cam, no disorienting motion, no exaggerated AI camera moves, no sudden pull-backs.
Movement must feel natural, like a high-end photographer using a handheld gimbal or slight handheld shooting:
0-3 seconds
Lens starts with an ultra-close side-front face push-in, focus on eyes and nose bridge area, subject slightly lifts gaze to lens, natural breathing feel, lips subtly relaxed.
Lens does an extremely subtle slow push-in, with very light lateral sliding, showcasing cheeks, under-eye texture, and hair strands.
This segment should make viewers feel the face's high-end realism right from the first glance.
3-6 seconds
Lens slides along the face from eyes to nose tip, then to near the lips, in an extremely steady micro-arc orbit, close-up observation of features.
Subject's gaze subtly shifts, no exaggerated acting—just gently looks toward lens edge, then slowly back to lens.
First flash can occur here, briefly lighting up to enhance skin texture and lip highlights.
6-9 seconds
Lens slowly turns from front-left to front-right, still maintaining face-push distance, emphasizing cheek skin texture, nose wing details, lower eyelid fine lines, lashes, and pupil reflections.
Subject lightly brushes a bit of hair, or a stray strand naturally falls to the side of the face.
Lens has a subtle breathing-like float for real photography feel, not rigid.
9-12 seconds
Lens subtly shifts down to highlight lips, jawline, and cheeks, then slowly back to eyes.
Subject's expression stays cool, restrained, with a touch of high-end aloofness, like a car model caught close-up by a photographer, occasionally a gentle blink, corners of mouth with extremely subtle relaxation.
Second flash occurs here, natural and instantaneous like on-site candid, returning to normal light after.
Post-flash, catchlights in eyes more prominent, giving the frame more fashion editorial vibe.
12-15 seconds
Lens does one of the most beautiful ultra-close light arc closes, subtly orbiting the face and lightly pushing in, finally settling on an ideal composition of eyes + nose bridge + half lips.
Subject gently lifts gaze to straight-on lens in the last second, eyes clean, calm, captivating.
Ending can have a third very subtle flash or environmental highlight sweep, but must feel natural, not overdone.
Finally holds on a high-end face-push frame perfect for a thumbnail cover.
Frame Style
Overall style is:
Real cinematic fashion portrait, close-up commercial photography, auto show model face-close candid, high-end magazine feel, natural skin texture, non-studio fake-face sense.
Include real lens bokeh, natural depth of field, fine skin quality, subtle on-site feel, restrained high-end vibe.
Not influencer selfie, not live stream, not Korean-style skin smoothing, not low-quality AI face.
Key Emphases
Face must be real, not fake
Skin can't be overly perfect, retain subtle imperfections
Features stable, no distortion
Eyes can't wander, lips can't move randomly
Hair strands must be real, with stray hairs
Flashes must follow real candid logic
Lighting must be real, no random strobing
Camera must be high-end, smooth, cinematic, fashion photography feel
Whole piece centers on "high-end face close-up showcase," no plot, no big movements
Negative Restrictions (Suggest Adding Together)
Negative prompts:
Cartoonish feel, CG feel, plastic skin, excessive smoothing, over-beautification, feature distortion, eye misalignment, skin wax figure feel, overexposure, illogical flash strobing, face shake, low clarity, low-quality hair strands, exaggerated filters, influencer face, overly heavy fake lashes, heavy makeup, shaky lens, limb distortion, cluttered background, readable text, subtitles, stickers, watermarks, 夜晚码头闺蜜同游与跳水
先写风格、角色和场景,再用带时间戳的分镜表,每个片段单独指定运镜,如倒退跟拍、甩镜下摇和俯拍。单独的音频段要求现场同期声且无音乐,强化手机随拍的真实感。
Raw vertical phone video (UGC), handheld throughout, filmed by a friend who walks with the group, casual sway, natural night lighting, wide-angle lens, deep focus.
Characters
The hero is a young woman with long dark hair in a deep-purple one-shoulder dress with a small shoulder bag.
Her three friends: one in a dark brown dress, a blonde in a black wrap dress, a blonde in a white mini dress — party dresses and sandals.
Laughing, spontaneous night-out energy.
Setting
A luxury marina at night — a stone quay along a narrow water slip flanked on both sides by large moored luxury yachts, big white hulls rising on either side, masts and rigging above, fenders on the hulls.
Warm orange street lamps, palm trees, distant city lights.
The yachts glow with blue-turquoise underwater LED lights spilling into the water; a yellow mooring line runs to the stone.
Shot list
0.0s–4.0s — Wide tracking shot, eye level, camera moving backward
The camera walks backward ahead of the four women as they stroll toward it down the marina promenade, framed head-to-toe, laughing and chatting in French; warm lamplight and palm trees behind, slight handheld bob.
4.0s–6.0s — Medium shot, eye level, backward track + quick left pan
The woman in the purple dress surges ahead of the group toward the lens, one arm raised, hair flying; the camera, on their left, keeps tracking back and pans to hold her.
6.0s–8.0s — Medium-wide, camera tilts down-left to follow, whip motion
She reaches the quay edge between the moored luxury yachts and launches headfirst over the edge; the handheld camera whips left and tilts down to chase her off the stone.
8.0s–10.0s — High-angle wide, looking down at the water, slight jolt
From up on the quay the camera looks down as she plunges headfirst into the blue-turquoise water in the slip between the big yacht hulls — a big splash, the underwater lights glowing around the impact; the handheld frame jolts slightly with the motion.
10.0s–13.0s — High-angle medium, looking down, small handheld drift
She surfaces in the teal water, gasping and laughing up at the camera, hair slicked back, treading water.
13.0s–16.0s — Handheld tilt up and back down, slight zoom
The camera lifts off the water and tilts up to a person leaning on the deck railing of one of the moored luxury yachts, watching her in the water, then tilts back down and zooms slightly in on her still treading and laughing.
16.0s–20.0s — High-angle medium, tilting to follow her up
She swims to the stone quay wall, reaches up and grabs a thick dark mooring pole, and hauls herself out; the camera tilts upward with her as she rises from the water, legs braced on the stone.
20.0s–23.0s — Low/close medium at the quay edge, static handheld
She climbs onto the stone lip, rolls onto her side, soaked dress clinging, then pushes up to her feet; the camera holds close, catching the water running off her.
23.0s–27.0s — Medium shot, eye level, loose handheld reframe
Back on the promenade, dripping wet, she flips her long wet hair forward and wrings it out, laughing, as her friends and passing pedestrians move behind her.
27.0s–30.0s — Medium shot, eye level, camera settles
Still dripping, she straightens up and strikes a confident, playful pose for the camera — hand on hip, tossing her wet hair back with a big smile — her friends laughing around her; the camera holds on the pose to end.
Audio
Live sync sound — friends laughing and shouting excitedly in French, a scream at the leap, a big water splash, water sloshing against the yacht hulls, wet footsteps on stone, ambient night marina sounds.
No music.
Style
Raw amateur vertical phone footage, natural night lighting, warm street lamps mixing with the turquoise underwater yacht glow, authentic handheld motion. 动作不变，换机位重拍你的视频
上传一段连续镜头作动作参考，Wan 从新角度重拍；某个镜头跑偏就重生成。
STEP 1: Upload your video (must be continuous one shot) to any platform that supports WAN 3.0
STEP 2: Use the prompt below (adjust it to your video):
"Use the original video @ TAG ORIGINAL VIDEO as the exact action reference
0–1s: Camera stays exactly as in the original video as the wolf walks toward the scooter.
1–3s: Hard cut to a tight front close-up as the wolf continues walking toward and reaches the scooter.
3–5s: Hard cut to a high angle above as the wolf as he gets onto the scooter and starts riding.
5–7s: Hard cut to a full shot as the wolf rides the scooter down the street.
7–9s: Hard cut to a side tracking shot as he keeps riding down the street.
9-10s: Hard cut to an over the shoulder shot as he rides forward and notices the banana peel in his path.
10–12s: hard cut to a birds eye view shot as the fallen wolf is rolling over the asphalt.
12-13s : hard cut to an insert shot of his fist as wolf slams it into the asphalt.
Only the camera angle and framing change. Make sure every action, movement, timing and performance exactly identical to the original video @ TAG ORIGINAL VIDEO. The action continues seamlessly across every cut."
NOTE: If some scenes didn't work (action was different than original video) simply regenerate the video, that usually fixes the problem. 咖啡馆柜台前身穿驼色大衣的男人
先用一张参考图生成三视图角色设定板，左侧为正面和背面全身，右侧为一张细致的头像。正面全身的脸被有意虚化，让头像成为唯一的面部参考，同时写清服装、面料和胶片质感，让角色在多个场景中保持一致。
Create a three-panel character reference sheet of the woman in
@image1
in a single wide 16:9 image.
Preserve her exact facial identity, blue eyes, dark espresso-brown hair color, skin tone and natural facial
proportions from
@image1
.
The left half contains two full-body views side by side: front and back, identical in scale, with both figures
completely visible from above the bun to below the shoes. The right half contains one large
head-and-shoulders portrait, facing directly into the camera. All three views share one continuous neutral
mid-grey studio backdrop, matching soft lighting and the same warm film color treatment.
In the full-body front view, replace the facial features with a smooth, softly blurred skin-toned oval. Keep the
hairstyle and head silhouette intact. The large portrait on the right is the only detailed face reference.
Restyle her exceptionally long, thick hair into a high, generously sized, loosely sculpted bun at the crown.
Preserve the fullness of her hair, with lifted roots and soft volume around the head. Two delicate curved
tendrils fall beside her temples toward her jaw, framing the outer edges of her face. Her eyebrows, blue eyes
and cheekbones remain clearly visible. The back view shows the gathered hair and the substantial rounded
shape of the bun. The hairstyle feels elegant and practical, like a beautiful café waitress in a romantic film
from around 2000.
Dress her in a fitted ivory fine-knit top with short sleeves and a modest scoop neckline, tucked into a
high-waisted black tailored pencil skirt ending at the knees. The skirt has a small functional back vent. Over it,
she wears a short deep-burgundy cotton waist apron, extending from the waist to mid-thigh, with two discreet
rectangular front pockets. Narrow apron ties wrap around her waist and form a neat bow at the center of her
back, clearly visible in the rear view. Her shoes are simple black leather ballet flats with gently rounded toes
and low flat heels. Small plain gold stud earrings complete the outfit.
Her figure is tall and slender, with a defined narrow waist, long legs and a graceful neck. The clothing follows
her natural proportions, with realistic knit texture, tailored seams and soft cotton folds around the apron ties.
02
The same outfit, fit, hem lengths and hairstyle appear in all three views.
In the front view she stands upright, square to the camera, shoulders level, weight evenly balanced, arms
relaxed at her sides and hands empty. The back view shows the identical stance turned fully away. In the
portrait her head is upright, chin level and both eyes look directly into the lens. Her lips are gently closed with
a faint, relaxed smile. Preserve the recognizable face from
@image1
, with natural skin texture, visible pores
and fine facial hairs.
Soft frontal studio lighting keeps her features and clothing clearly readable, with gentle shading under the
chin and soft contact shadows beneath the shoes. The grey backdrop remains plain throughout the image.
The composition consists entirely of the three views and the seamless backdrop.
Maintain a glamorous late-1990s to early-2000s 35mm film publicity look: warm natural skin, creamy ivory
fabric, rich burgundy cotton, deep readable blacks, fine organic grain and restrained highlight glow. Her face,
blue eyes, hair strands and garment details remain clearly resolved. Clean presentation with uninterrupted
backdrop between the views.
📌 PROMPT 2 (Guy map):
Create a three-panel character reference sheet of the man in
@image1
in one wide 16:9 image. Preserve his
exact face, eye color, skin tone, facial proportions and hairstyle from the reference, including the natural
swept-back waves and the loose strand near his forehead.
The left half of the image contains two full-body figures side by side: a straight-on front view and a complete
back view, identical in scale, both visible from above the hair to below the shoes. The right half contains one
large head-and-shoulders portrait looking directly into the camera, with his entire hairstyle inside the frame.
All three views share one continuous neutral mid-grey studio backdrop.
On the full-body front figure, replace the facial features with a smooth, softly blurred skin-toned oval,
retaining the head silhouette and hair. The large portrait on the right is the single detailed face reference.
He is tall, approximately 188 cm, with broad shoulders, a lean athletic build, a defined waist and long legs. His
proportions remain natural and identical across all three views.
Dress him in an elegant camel-colored wool-and-cashmere overcoat reaching just below the knees. The coat
has broad notched lapels, softly structured shoulders, a single-breasted three-button front, discreet welt
pockets and a single center-back vent. He wears it fully unbuttoned, its front panels hanging naturally to
reveal the clothing underneath. The fabric is substantial, softly brushed and richly textured, with realistic
weight and long vertical folds.
Under the overcoat, he wears a dark chocolate-brown tailored wool blazer with a two-button front, worn
unbuttoned, over a pale-blue cotton shirt with a pointed collar and the top button undone. The shirt is tucked
neatly into high-waisted charcoal-grey wool trousers with a single pleat on each side, a relaxed straight leg
and a slight break over the shoes. A dark-brown leather belt with a small brushed-metal buckle completes the
waist.
His shoes are polished dark-brown leather penny loafers with matching dark socks. A slim rectangular gold
wristwatch on a dark-brown leather strap sits on his left wrist, partly visible beneath the cuff. The styling has
04
the understated elegance of a wealthy young man in a romantic film set around 2000.
In both full-body views, he stands upright with shoulders level, weight evenly balanced and arms relaxed
slightly away from his sides so the coat silhouette remains readable. His hands are empty. The back view
shows the coat's full length, center seam, vent and natural drape.
In the large portrait, his face is strictly frontal, head upright, chin level and both eyes directed into the lens.
His lips are relaxed and closed. Preserve the exact facial structure, subtle beard shadow and natural skin
texture visible in
@image1
. The portrait includes the camel coat lapels, brown blazer and pale-blue shirt
collar.
Soft, even studio lighting keeps his face and all clothing details clearly readable, with gentle facial shading,
broad highlights on the wool and soft contact shadows beneath both pairs of shoes. The backdrop is plain and
continuous, with clean empty space separating the three views.
Maintain warm late-1990s to early-2000s 35mm film color: natural warm skin, rich camel wool, deep
chocolate tailoring, soft pale-blue cotton and readable charcoal shadows. Fine organic film grain and
restrained highlight glow give the image a tactile cinematic finish while the face, hair and garment
construction remain clearly resolved.
📌 PROMPT 3 (Cake box map):
A photorealistic pastry packaging reference sheet on a uniform medium-grey seamless studio background, horizontal 16:9 composition, divided into two equal side-by-side views. The surface beneath both boxes and the background are the same neutral grey.
On the left, a closed rectangular ivory cardboard pastry box with a fitted hinged lid, viewed from a slightly elevated three-quarter angle. Matte cardboard, subtle paper texture, clean folded edges and a plain unprinted surface.
On the right, exactly the same box at the same scale, with its hinged lid fully raised behind it. A higher three-quarter viewing angle clearly reveals the contents: six individual chocolate mousse pastries arranged in two rows of three, each on a small gold pastry board. Comfortable spacing makes every pastry’s complete silhouette visible.
Each pastry has a playful, stylized penis shape: a smooth rounded elongated form with two rounded lobes at the base, recognisable as a humorous bachelorette-party confection. All six share the same shape and size, with glossy milk-chocolate mirror glaze and a tiny touch of edible gold leaf. Smooth moulded chocolate surfaces and appetising highlights give them the appearance of carefully crafted patisserie desserts.
The box dimensions, ivory colour and construction are identical in both views. Neatly fitted ivory food-safe paper lines the interior. Both boxes are fully visible with comfortable grey space around them.
Soft studio lighting, gentle contact shadows, clear high-resolution detail and natural optical sharpness with restrained sharpening. Subtle fine film grain and softly warm highlights, consistent with late-1990s cinematic photography. The entire surrounding background remains solid neutral medium grey.
📌 PROMPT 4 (Cafe map):
Create a cinematic location reference photograph of an intimate, elegant neighbourhood café in the year 2000, shortly before closing on an autumn evening. One wide horizontal 16:9 image showing a single coherent interior, photographed at eye level from the customer side, looking diagonally toward the service counter. A natural 28mm lens perspective with straight architectural lines and clearly readable room geography.
A long service counter occupies the right half of the composition, running from the foreground toward the back wall. It has a polished dark walnut front with recessed wooden panels and a warm ivory stone countertop. The central serving section is a broad, clear, uninterrupted surface, ready for a pastry box and a conversation between a standing customer and the waitress behind the counter. This section of the counter is bare, with the stone surface fully visible.
Behind the counter is a comfortable working aisle. The rear service wall contains dark wooden shelves, neatly arranged porcelain dishes, a stainless-steel espresso machine and a small period-appropriate cash register at the far end. A wooden doorway in the back wall leads to the kitchen. A small glass pastry display stands at the far end of the counter, leaving the main serving area open.
On the left side of the room, three small dark wooden tables stand beside burgundy leather banquette seating along the wall. Each table has two bentwood chairs and a small glass sugar dispenser. Above the banquette hang framed botanical prints and warm brass wall sconces with ivory shades. The walls are painted creamy butter yellow above dark wooden wainscoting. The floor is laid with small cream and muted terracotta tiles.
At the far left, tall street-facing windows and a glazed wooden entrance door reveal a quiet city street at dusk. Soft, cool evening light enters through the windows, balanced by warm pools of light from the wall sconces and three opal-glass pendant lamps above the counter. The entrance, seating area, service counter and kitchen doorway are all clearly visible within the same spatially consistent room.
The café feels established, lovingly maintained and slightly worn: subtle patina on brass fittings, gentle wear on the leather seats, fine scratches in the wooden furniture and realistic texture in the plaster walls. The tables are neatly reset for the end of the evening. The room is unoccupied so its layout and furnishings are fully visible.
The visual atmosphere is a romantic feature film from 1999–2000, captured on 35mm film: warm amber practical lighting, creamy highlights, rich walnut brown, muted burgundy and natural shadow detail. Fine film grain, subtle halation around glowing lamps and soft highlight roll-off. Maintain enough depth of field to clearly describe the counter, seating and rear service wall. Photorealistic materials, believable proportions and an inviting, lived-in atmosphere. A clean cinematic location photograph filling the entire frame.
📌 PROMPT 5 (Animation prompt):
SCENE CONTEXT A 30-second scene in a neighbourhood café in the year 2000, near closing time. A headstrong waitress packs her handmade pastries. A wealthy customer approaches to pay, interrupts her work with questions, tastes a sample and asks her to pose as his girlfriend at dinner. Play this as a lively romantic dramedy with friction, humour and unexpected attraction. She wants to finish her work and refuses to entertain his casual charm. He wants to keep her attention and finds himself having to earn it. Their reactions evolve naturally as they listen to each other. ACTIVE REFERENCES
@image1
— the waitress. Use the detailed portrait for her exact face and the full-body views for her work outfit and high loose bun.
@image2
— the customer. Preserve his exact face, hair and clothing, including the unbuttoned camel overcoat.
@image3
— the café architecture, counter, furnishings and dusk outside.
@image4
— one ivory pastry box and its six chocolate novelty pastries. Preserve their shapes, chocolate glaze, gold decoration and individual gold boards. The panels show two states of the same box. SPACE AND PROPS She stands behind the counter; he approaches on the customer side. In shared shots he is screen-left, she is screen-right. Keep the camera on the same side of their dialogue axis, maintaining matching eyelines. The box begins open with five pastries inside. She is placing the sixth into the remaining space, holding its gold board. Beside the box is a small ivory plate with an additional matching dessert cut into two portions and a dessert fork. He samples this separate dessert; the six boxed pastries remain whole. He carries a folded banknote in his left hand, ready to pay. His right hand is free. The central serving surface contains the box and tasting plate. Cups stay on the background shelves. CAMERA AND EDITING 16:9, 30 seconds, real-time action. Film coverage with foreground depth, deliberate camera movement, expressive changes in shot size and motivated focus pulls. Hard cuts at the listed boundaries. Stable focal length within each shot. Let performances flow inside the shots, with natural listening and gestures overlapping speech. 0.0–3.0 SECONDS 63° FOV. A short lateral tracking shot moves along the customer side of the counter with the approaching man, revealing the waitress already at work beyond him. The polished counter creates a strong diagonal into the frame. She places the final pastry into the box. He stops opposite her, notices the contents and suppresses a laugh. She remains occupied with packing. HARD CUT — 3.0–5.5 SECONDS 47° FOV. Low oblique view across the open box: the pastries are sharp in the foreground, his face visible beyond them. Pull focus from the chocolate shapes to his amused expression as he looks up at her. MAN: “Is this the new menu?” He is teasing, expecting an easy smile in return. HARD CUT — 5.5–10.5 SECONDS 47° FOV. Side-on medium shot of the waitress, her face and hands visible, his coat softly framing one edge. WOMAN: “Test batch for a bachelorette party. My coworkers approved.” She answers with dry impatience while straightening the paper inside the box. She gives him a pointed look on “approved,” then slides the tasting plate toward him with a brisk, controlled gesture: he can taste it and let her finish. The plate stops within his reach. Her delivery has bite and personality; her attention moves naturally between him and her work. HARD CUT — 10.5–14.5 SECONDS 47° FOV. Counter-height three-quarter medium shot of the man, showing his face and right hand, with the box edge in the foreground. He takes a small forkful from the cut dessert. As he tastes, his teasing manner gives way to spontaneous appreciation. He finishes the bite and lowers the fork onto the plate. She is visible at the edge of frame, briefly watching to see whether he likes it before returning to her work. The banknote remains in his left hand. HARD CUT — 14.5–20.5 SECONDS 47° FOV. A layered two-shot looking diagonally along the counter, both faces readable. The camera makes a gentle lateral move that settles as their eyes meet. MAN: “You make these yourself?” His interest is genuine now. WOMAN: “After my shifts. I'm saving for my own bakery.” She answers firmly, with pride beneath the impatience. For a moment she engages fully because her bakery matters to her. Then she returns her attention to the box. He watches her and considers his next words. HARD CUT — 20.5–24.0 SECONDS 29° FOV. Medium close-up of the man from beside her shoulder. A slow, restrained dolly inward makes the conversation feel more personal. MAN: “Have dinner with me tonight?” He attempts an easy invitation, but there is a real question in his eyes. Allow a short listening beat after the line. HARD CUT — 24.0–30.0 SECONDS 29° FOV. Close-up of the waitress from his eyeline, her face in three-quarter view and his blurred shoulder at the edge. Begin with a very slight forward camera move. She looks up from her work and gives him a brief, sceptical half-smile, amused by his nerve. MAN, naturally offscreen: “With my family. As my girlfriend.” His voice hesitates briefly between the two phrases. Stay on her throughout his clarification so the audience sees it land. Her half-smile disappears; she stops what she is doing and looks directly at him with an immediate, incredulous reaction. She lets out a short disbelieving breath and settles her right palm on the counter. The camera settles with her. Hold the final moment as she silently sizes him up, ready to challenge him. End after the hand movement is complete, on her living reaction. ACTING Give the actors room for spontaneous conversational behaviour. She is spirited, quick and proud, with dry humour and a short tolerance for being patronised. Her annoyance changes with the conversation: she dismisses his teasing, briefly cares about his response to her baking, defends her ambition and finally reacts to the audacity of his request. He starts charming and self-assured, adjusts when she gives him little encouragement, becomes genuinely impressed by the dessert and risks the invitation. His confidence becomes less certain as he waits for her response. Use connected reactions, natural breaths, shifts of attention and expressive vocal rhythm. Each person listens while the other speaks. Movements arise from packing, tasting and trying to hold or end the conversation. Keep their faces recognisable and their physical behaviour grounded. LIGHTING AND COLOUR Preserve the café's amber pendant lamps, brass sconces and cool dusk windows. Soft directional side lighting shapes their faces, with darker shadow cheeks and gentle reflected fill preserving their eyes. Photograph from the shadow side. Loose hair, camel wool and chocolate glaze catch warm highlights. Fixed 4000K white balance. Rich walnut browns, burgundy leather, creamy practical highlights and deep textured shadows. Natural skin colour beneath the warm light. Consistent lighting across all angles. AUDIO — FIXED CHARACTER VOICES WOMAN VOICE: Adult woman in her mid-twenties; medium pitch, warm clear mezzo speaking voice with a lightly husky lower register, neutral General American accent, natural conversational rhythm, crisp intelligible consonants and restrained expressive inflection. MAN VOICE: Adult man in his early thirties; medium-low pitch, warm resonant baritone with a subtle dry rasp, neutral General American accent, measured conversational rhythm, clear articulation and understated playful inflection. Preserve these vocal identities. Her delivery is dry, spirited and impatient, becoming firm when discussing her bakery. His delivery moves from playful confidence to sincere curiosity and slightly hesitant persuasion. Use the written English dialogue exactly. Clear turn-taking and accurate lip synchronisation for visible speakers. His final clarification is heard offscreen while the camera stays on her reaction. Quiet café room tone, distant traffic, paper rustle, a soft plate scrape, the fork touching porcelain and her final palm contact. Keep dialogue clear and foregrounded. FILM LOOK Late-1990s theatrical romantic dramedy captured on visibly textured 35mm film. Organic moving grain is clearly perceptible across midtones, skin and shadows at normal playback size. Warm halation surrounds the glowing lamps; bright reflections bloom gently. Rich film-print contrast, creamy highlight roll-off and natural optical sharpness with restrained edge enhancement. Preserve readable eyes, realistic skin and fabric detail beneath the film texture. Natural motion blur and 24 fps cinematic cadence. CONTINUITY After packing, all six boxed pastries remain whole and in place. The separate tasting dessert retains its missing bite. The box stays open on the same section of counter. After the slide, the tasting plate and fork remain on his side. His banknote stays in his left hand. Maintain exact faces, hair, clothing, screen direction, café geometry and dusk throughout. Reference sheets guide appearance; the output is full-frame cinematic footage. 手绘动画风，对战巨型怪物
两个带标签的参考（主角、怪物），要求保留手绘线条。Wan 的动画有自己的味道。
Use @[character ref] for the hero's exact identity and appearance; @[monster ref] for the monster's anatomy, scale, pale body, black projections and luminous horns. Preserve their angular hand-drawn contours, sparse sketch lines, matte cel-shadows, muted lavender and ash gray, blue-black shadows, peach rim light and fine grain. City and effects share this 2D anime style. References bind appearance, not framing or poses.
Five epic shots: the monster destroys the city as it charges toward the rooftop hero. ONE sudden, thick white line bursts from his index fingertip and instantly punches through it, leaving one enormous hole.
The hero stays at one skyscraper's avenue-facing edge, level with the monster's upper torso. The monster approaches directly IN FRONT OF him along the avenue. All cameras stay on one side of their shared axis: hero and tower screen-left, monster screen-right advancing left. The hero faces right. In his profile close-ups, the monster remains ahead, offscreen right, never behind his head or shoulders. No axis-crossing angles or orbits.
Shot 1. Low lateral wide tracking beside the monster charging right-to-left. Its shoulder tears through a building; the facade folds into the street as its next footfall buckles the roadway, hurling rubble forward. Debris whips past the lens through lavender dust. Widen and tilt upward to reveal the target tower ahead on screen-left, the tiny hero at its edge facing the approaching monster on screen-right. Keep the gap between them readable.
Shot 2. Right-side profile of the hero facing screen-right toward the offscreen monster. Face and right hand share the frame against sky and rooftop. Wind pulls his hair and coat; he stays relaxed. With a knowing smile, he extends his right arm toward the target: index straight, thumb raised, other fingers curled. Cut as the finger aligns with the target; no charging glow, gathering energy or held aiming pause.
Shot 3. Steady rear three-quarter view outside the hero's right shoulder, on the established side. Keep a clear gap between his fingertip screen-left and the monster's upper torso screen-right. It reaches for the rooftop directly ahead. Immediately on the cut, ONE thick, straight, brilliant-white line snaps from the index fingertip through the upper torso, below the neck between the shoulders. Its entire length appears instantly, connecting finger, target and the space beyond. Release and penetration happen together: no projectile travel, growing beam or drilling. One tiny finger recoil coincides with a building-sized circular section being punched through the body. The white line switches off immediately, leaving an enormous EMPTY opening, its broken dark rim framing the skyline. The torso jolts and tips sideways at once, the reaching arm knocked off course. Read the hole during the uninterrupted fall. No charging, stunned pause, slow motion or hit-stop. No sphere, pressure ring or full-screen whiteout obscures the contact. One abrupt discharge, never a sustained beam or repeated pulse.
Shot 4. Cut as that sideways tilt begins, keeping the same viewing direction and matching the ALREADY OPEN hole on the first frame. The monster's arm drops short of the roof; its weight carries it diagonally sideways into the blocks beside the avenue before reaching the tower, still in front of the hero. Track downward with the chest exposed: no edge-on turn or face-down roll. The SAME hole stays fixed below the neck between the shoulders, with the same physical diameter, broken rim and missing volume. Its outline foreshortens naturally as the torso tilts; background buildings slide visibly through the opening. Keep dust below the chest and clear of the aperture. Follow the fall through contact with the ruins and briefly show the fallen torso with its hole still open. The rim does not regrow, the opening does not fill, and the torso never returns to its intact state.
Shot 5. Return to the hero's right-side profile, facing screen-right, face and right hand together against sky and rooftop. The fallen monster remains OFFSCREEN ahead to the right. He bends his right elbow and brings the SAME extended index fingertip just in front of his lips. Only after it arrives, he gently purses his lips and visibly blows across it. A fine white wisp bends away in his breath. He lowers the hand slightly with a satisfied smile. Collapse thunder rolls in from ahead and a gust lifts his coat. Hold his effortless composure.
One instantaneous fingertip discharge, one impact, one permanent through-hole carried through the uninterrupted fall and landing. The hero keeps the same hand and facing. Sound: masonry and heavy footfalls, one dry electrical crack and a deep impact report together at the instant of contact, then the intimate breath against distant collapse. No dialogue or subtitles. 工业厂房中的直播风射击游戏实录
按时间段写出分镜（0到6秒、6到12秒及之后），让模型有明确的逐镜目标，并给出确切的界面数值便于核对。要求画面像可操作的实际游戏而非预告片，有助于镜头和界面保持真实。
Create a 30-second ultra-realistic AAA third-person/first-person hybrid livestream gameplay video, 16:9, 4K, resembling an actual professional competitive game being livestreamed. The game is an original tactical snapping/shooting game with a futuristic precision weapon. Make everything feel like real playable gameplay, not a cinematic trailer.
GAMEPLAY STYLE: Unreal Engine 5 quality, realistic physics, responsive controls, detailed environments, professional streamer HUD, weapon animations, hit feedback, minimap, ammo counter, health/armor, objective tracker, kill feed, crosshair, score, ping/FPS indicators, and livestream overlay.
PLAYER: Professional gamer controlling a futuristic Snapper Precision Gun capable of switching between normal direct targeting, optical lens targeting, and microscopic ultra-precision targeting.
0–6 SEC — DIRECT SNAP
The player moves through an abandoned industrial facility while enemies appear from different directions.
Quick camera movement → player raises the Snapper Gun → direct hip/shoulder aim → rapid precision shots eliminate two enemy targets.
HUD instantly updates the score and kill feed.
ON-SCREEN: ENEMIES: 6 | AMMO: 24/120 | HP: 100 | ARMOR: 75 | SCORE: 1,250
6–12 SEC — OPTICAL LENS MODE
The player spots an enemy far across the facility.
Weapon switches into high-powered optical lens mode. Screen smoothly transitions into a realistic scope view with lens reflections, distance marker, wind indicator, rangefinder and moving reticle.
The player tracks the enemy through the lens → precise shot → enemy drops behind cover.
HUD: OPTICAL MODE | RANGE: 187M | ACCURACY: 94%
12–19 SEC — MICROSCOPIC SNAP MODE
The player discovers a tiny exposed component on an enemy's futuristic shield.
Activate Microscope Snap Mode. The screen dramatically magnifies the target, showing an extremely detailed microscopic targeting interface.
The reticle locks onto the tiny weak point → satisfying precision snap → shield breaks → enemy becomes vulnerable.
HUD: MICRO SNAP LOCKED | WEAK POINT DETECTED | +500
19–25 SEC — RAPID COMBAT
Three enemies rush the player from different angles.
The player instantly switches back to direct mode, slides behind cover, leans out and performs fast precision snaps.
Add realistic muzzle flashes, recoil, shell/effect particles, impact sparks, environmental destruction and smooth weapon animations.
KILL FEED:
PLAYER — PRECISION SNAP — ENEMY
PLAYER — HEADSHOT — ENEMY
25–30 SEC — FINAL TARGET
Only one enemy remains. The player runs upstairs while the enemy escapes through a narrow corridor.
Switch between direct view → optical lens → microscope precision mode in one fluid sequence.
The microscopic reticle locks onto a tiny glowing weak point.
FINAL SNAP.
The screen displays:
TARGET ELIMINATED
MISSION COMPLETE
PRECISION: 97%
TOTAL SCORE: 4,850
End with the professional livestream interface still visible, including LIVE indicator, viewer count, chat panel, FPS, ping, minimap, health, armor, ammo, objective tracker, kill feed and weapon status.
IMPORTANT: Everything must look like a polished real AAA game: consistent speed, realistic movement, accurate weapon animations, responsive HUD, professional UI, natural camera motion, synchronized sound effects, realistic lighting and shadows, no glitches, no floating objects, no random UI elements, no copyrighted game interface, and no text appearing except intentional game HUD/livestream elements.
5 best caption also add now creating these Livestream video games through wan 3.0 on budgetpixel is so easy 这个模型的提示词怎么写
先只写一句
阿里云的指南说，一句话就够万相 3.0 生成一段视频，不需要控制的部分可以不写。创作者写得长得多：这页提示词的中位长度是 3,058 个字符，约 340 个英文词。 来源：Wan 3.0 prompt guide · AgentCraft: MiniMax H3 and Wan 3.0 prompt data
照指南的顺序往上加
需要更多控制时，指南的顺序是：先写整体描述，再写参考素材、镜头、对白、声音和音乐、风格，最后是反向要求。 来源：Wan 3.0 prompt guide
按上传顺序写 Image N
写成 Image N、Video N 或 Audio N，编号必须和上传顺序一致，不然会引用错素材。只有一份时，写参考图就行。同一份素材可以引用多次，比如一次用它的外观，一次用它的声音。这页 12 条参考生视频提示词里有 4 条给素材编了号。 来源：Wan 3.0 prompt guide · AgentCraft: MiniMax H3 and Wan 3.0 prompt data
镜头按 2 到 5 秒一段排
多镜头时，每个镜头写编号和用方括号或圆括号括起的时间段，全篇用同一种写法，每段 2 到 5 秒，首尾相接，不留空也不重叠。别写模型控制不了的高频动作。这页 60 条里有 42 条给镜头排了时间，但只有 2 条用了括号写法。 来源：Wan 3.0 prompt guide · AgentCraft: MiniMax H3 and Wan 3.0 prompt data
台词写成：人物、说、冒号、引号
需要口型对上时，加上 Lip sync。不想要人说话就写明，不然模型会自己决定加不加台词。 来源：Wan 3.0 prompt guide
配乐用一句话定
配乐可以不写、写清楚，或者用 No BGM, generate only ambient and action sounds 关掉。动作、材质和天气写清楚了，环境声会跟着出来。这页有 5 条关掉了配乐。 来源：Wan 3.0 prompt guide · AgentCraft: MiniMax H3 and Wan 3.0 prompt data
反向要求只写真不想要的
只写一定不能出现的东西，没有就空着；别凑数，也别重复正面描述里已经写过的。这页 60 条里有 44 条带不要、避免或没有这类字样。改视频时，一次只改一处最稳。 来源：Wan 3.0 prompt guide · AgentCraft: MiniMax H3 and Wan 3.0 prompt data
关于这个模型的常见问题
万相 3.0 是什么？
阿里巴巴的视频生成模型，在阿里云百炼上提供 wan3.0-video 和 wan3.0-video-prime。一个模型就能做文生视频、图生视频、参考生视频、改视频和续写。 来源：Wan 3.0 video generation guide
万相 3.0 一段视频最长多久？
2 到 30 秒。duration 设为 -1 时由模型决定，默认 5 秒。续写时，原视频加新生成的部分最多 30 秒。 来源：Wan 3.0 video generation guide
万相 3.0 最多能用多少参考素材？
最多 10 张图、5 段视频（合计不超过 15 秒）、5 段音频（合计不超过 15 秒），另可加一份文档或一个网页链接，每次请求合计最多 20 份。 来源：Wan 3.0 video generation guide
万相 3.0 能根据文档或网页生成视频吗？
能。它接受一个文件，比如 PDF、Word、Excel、PowerPoint、纯文本或 Markdown，或者一个网页链接。文件或网页必须能公开访问，需要登录的页面读不到。 来源：Wan 3.0 video generation guide
首尾帧能和参考素材一起用吗？
不能。指南说首尾帧模式和全参考模式互斥。图生视频时，画面已经定了主体和风格，提示词就写动作、情节、对白和运镜。 来源：Wan 3.0 prompt guide
万相 3.0 输出什么分辨率和帧率？
480P、720P 或 1080P，默认 1080P，输出 30 帧的 MP4，画幅 16:9、4:3、1:1、3:4、9:16 或自适应。默认带声音，把 audio 设为 false 可以关掉。 来源：Wan 3.0 video generation guide