AI 提示词库

Portrait of a young woman with windblown hair, soft natural light coming from a window on her left, shallow depth of field f/1.8, 85mm lens, slight rim light, warm color grading, freckles visible, photographed on a Canon R5, magazine-quality skin retouch, ultra-detailed eyes

blurry, oversharpened, plastic skin, distorted face, low resolution

推荐模型： Midjourney v7, Stable Diffusion XL, FLUX.1

指明相机型号、焦段、光圈。摄影元数据训练过的模型对这些参数响应强烈。

A cozy fairytale village built into the side of a mossy hill, hand-painted in the style of a 1960s Polish childrens book, warm autumn palette of ochre and burnt orange, soft watercolor textures with visible brushstrokes, lanterns glowing in the windows, a winding cobblestone path, a black cat watching from a window, no text

photo realistic, harsh lines, digital art, text, signature

推荐模型： Midjourney, NovelAI

引用某地区+某年代的插画传统，能让风格更统一。

Anime style, two highschool girls leaning on a chainlink rooftop fence, late afternoon golden hour, distant city haze, cherry blossom petals drifting through frame, Makoto Shinkai inspired lighting, ultra-detailed background, soft cel-shading, 4k

3d render, photo, distorted hands, extra fingers

推荐模型： NovelAI, AnythingV5, SDXL anime checkpoints

点名导演（新海诚、宫崎骏、汤浅政明）比单写"动漫风格"效果好得多。

Studio product render of a minimalist white ceramic coffee mug, floating against a soft gradient background fading from teal to off-white, dramatic side lighting casting a soft long shadow, subtle reflections, octane render quality, ultra clean, 8k

cluttered background, hand, person, text, logo, watermark

推荐模型： Midjourney, FLUX.1, Imagen 3

"studio product render" + 光线方向是关键。加上 octane/cinema4d 等渲染器名让画面更精致。

Concept art of a tiny noodle stall in a cyberpunk Tokyo alley, neon kanji reflecting on wet asphalt, steam billowing from the pot, a lone customer in a translucent rain jacket, color palette dominated by magenta and cyan with deep shadow areas, painterly digital art, Beeple meets Syd Mead, ultra-detailed

推荐模型： Midjourney, FLUX.1

用 "X meets Y" 组合两个艺术家风格融合。

Overhead flat lay food photography of a rustic ramen bowl with soft-boiled egg, scallions, nori, and chashu, on a dark wooden table with chopsticks, small bowl of chili oil, golden side lighting, professional food magazine style, shallow depth of field on the bowl, 50mm lens, natural shadow

unappetizing, soggy, dim, blurry, distorted text

推荐模型： Midjourney v7, SDXL, FLUX.1

"Flat lay 俯拍" + 黄金光是出片率最稳的美食组合。

Cinematic aerial drone shot, slow flythrough over snow-covered mountain peaks at sunrise, golden light kissing the ridges, soft mist drifting between valleys, dramatic camera push-in, anamorphic lens flare, 24fps, IMAX scale, 5 seconds, smooth motion

推荐模型： Sora, Runway Gen-3 Alpha, Veo

对 Sora/Runway，明确写镜头运动（推进、平移、环绕）。模糊的镜头描述会出抖动画面。

Studio product video, white minimalist sneaker rotating slowly 360 degrees on a glossy black turntable, soft top-down key light with subtle rim, gradient grey backdrop, shallow depth of field, 5 seconds, perfectly smooth rotation, no human in frame

推荐模型： Runway Gen-3 Alpha, Pika 1.5, Kling

告诉模型主体只旋转不移动位置，否则容易漂移。

Handheld vlog shot inside a cozy small cafe, late morning warm light through window, slight camera shake, racking focus from latte art on table to barista in background, ambient bokeh, 5 seconds, natural color grading, real-world physics

推荐模型： Runway Gen-3, Pika, Sora

"racking focus" + "handheld" + "ambient bokeh" 组合出纪录片感。

Anime style 2D animation, a teenage girl in a yellow raincoat walking through a Tokyo street at night, raindrops, puddle reflections of neon signs, gentle camera tracking sideways, Makoto Shinkai inspired lighting, 5 seconds, hand-drawn feel, no morphing characters

推荐模型： Sora, Kling, Hailuo (MiniMax)

加 "no morphing characters" 抑制人脸变形——这是当前视频模型最常见的瑕疵。

Cinematic underwater shot, slow forward dolly toward the ruins of an ancient submerged city, beams of sunlight cutting through clear blue water, schools of fish swimming past, particles drifting, mysterious atmospheric score implied, 6 seconds, IMAX scale, smooth tracking

推荐模型： Sora, Veo, Runway Gen-3

水下场景明确写 particles 与 light beams，能出标志性的水下质感。

Editorial fashion photograph of a young Asian model in oversized streetwear, standing in a brutalist concrete underpass in Tokyo, shot on Hasselblad medium format, soft overcast daylight, slightly desaturated film grain, full body composition, magazine layout sensibility

studio backdrop, smiling, beauty filter, plastic skin, low resolution

推荐模型： Midjourney v7, FLUX.1

点名真实相机（Hasselblad / Leica）比写 "professional" 效果稳定得多。

Wide-angle landscape photograph of black volcanic sand beach with basalt sea stacks rising from crashing North Atlantic waves, low storm clouds, golden hour light breaking through, long exposure smoothing the water, shot on Sony A7R V with 16-35mm lens, 4 stop ND filter, ultra-detailed foreground texture

people, distorted horizon, oversaturated, HDR halos, blurry

推荐模型： Midjourney v7, FLUX.1, Imagen 3

长曝光术语（ND filter / smoothed water）摄影元数据训练过的模型能识别。

3D character design of a friendly round-bellied raccoon mascot for a tech brand, soft pastel color palette, large expressive eyes, holding a tiny coffee cup, isometric pose, soft global illumination, Octane render, no background, neutral white floor with soft shadow

cartoon style, flat shading, sharp angles, distorted proportions

推荐模型： Midjourney, FLUX.1

品牌吉祥物要写 "neutral background"，否则模型会自作主张加奇怪背景。

Isometric illustration of a small tech city — server racks, satellite dishes, a tiny coffee shop, fiber cables glowing between buildings, cyan and magenta accent lighting against a deep navy background, flat-shaded with subtle gradients, similar in spirit to Antonio Uve and Studio MUTI

推荐模型： Midjourney, FLUX.1

等距 + 平面着色 + 具体配色 = SaaS 引导页那种插画风。

Cyberpunk anime portrait of a young woman with neon pink hair, wearing a high-collar tactical jacket with subtle LED trim, rain-slick Tokyo street at night reflecting holographic billboards in the background, dramatic side lighting, cel-shaded with sharp linework, in the style of Ghost in the Shell

photoreal, 3d, extra fingers, blurry, low quality

推荐模型： NovelAI, AnythingV5, Midjourney --niji

点名具体动漫（攻壳机动队/阿基拉）比模糊的 "cyberpunk anime" 出图稳定。

Concept art of a cozy fantasy tavern interior at night, warm fireplace casting orange glow, wooden beams overhead, a bard playing lute in the corner, patrons laughing at long tables, mugs and copper cookware on the walls, dust motes in the firelight, painted in the style of Magali Villeneuve, ultra-detailed

推荐模型： Midjourney, FLUX.1

点名奇幻插画家（Magali Villeneuve、John Howe）能立刻得到辨识度高的风格。

Macro product photograph of a luxury automatic mechanical watch, sapphire crystal showing the movement inside, brushed steel bracelet, dramatic 45-degree key light from the upper left, deep black background with a single soft rim light, shallow depth of field, Hasselblad H6D, 4k, magazine-quality retouch

cluttered, person, watermark, low resolution, soft focus on the dial

推荐模型： Midjourney v7, FLUX.1

产品摄影三要素：明确主光方向、纯黑背景、微距构图。

Black and white documentary portrait of an elderly fisherman with weathered face and kind eyes, harbor in the soft-focus background, shot on Leica M with 50mm Summilux, natural window light from the left, deep contrast, film grain, in the style of Sebastiao Salgado

color, smiling for camera, studio lighting, retouched skin

推荐模型： Midjourney v7, FLUX.1

黑白纪实人像要点：真实摄影师参考 + 颗粒感胶片描述。

Studio 3D product video, a wireless earbud charging case slowly disassembles into its components — case, two earbuds, internal battery — components float apart and rotate, soft top light against a gradient grey background, 5 seconds, smooth choreography, no human in frame, end with all parts settled in space

推荐模型： Runway Gen-3 Alpha, Kling, Sora

告诉模型相机静止，否则它会自作主张推拉，与爆炸动画冲突。

Cinematic shot of a small specialty coffee shop at 7am, golden morning light streaming through tall windows, barista in apron pulling a shot of espresso, steam rising slowly, dust particles visible in the light, shallow depth of field on the cup, gentle camera push-in, 6 seconds, Roger Deakins inspired color grading

推荐模型： Sora, Runway Gen-3, Veo

点名摄影指导（Roger Deakins、Emmanuel Lubezki）能锁定特定光影风格。

Anime style, a young girl with a school bag flies above a Japanese city at dusk, hands outstretched, gentle wind moving her hair and clothes, warm orange sky with scattered clouds, distant Mt. Fuji silhouette, camera tracks alongside her, 6 seconds, hand-drawn cel shading, Hayao Miyazaki inspired sky

推荐模型： Sora, Kling, Hailuo

飞行场景需要 "camera tracks alongside"，否则模型容易出静帧。

Macro video shot of fingers typing on a mechanical keyboard, warm desk lamp light, individual keys depressing in detail, slight camera dolly along the keys, shallow depth of field, soft bokeh of monitor light in background, 5 seconds, hyper-detailed, smooth motion

推荐模型： Runway Gen-3, Pika, Sora

微距 + 单一主体 + 推轨运镜 = 多数模型都能稳定出片。

Close-up of rain droplets sliding down a window pane, blurred city lights in the background, slight bokeh, gentle wide-aperture cinematography, soft natural light, lo-fi music video aesthetic, 6 seconds, no camera movement, gentle natural variation in droplet flow

推荐模型： Pika, Runway Gen-3

相机静止 + 自然物理（droplet flow）比要求运镜的结果干净得多。

iPhone 16 Pro mockup floating in three-quarter perspective, displaying a fintech mobile app dashboard with dark mode, neumorphic cards, soft cyan and purple accent gradients, subtle particle background, premium SaaS aesthetic, ultra-clean typography, 8k product render

cluttered, text artifacts, distorted, low resolution, fake-looking

推荐模型： Midjourney v7, FLUX.1

UI 模型图要明确写手机型号 + 视角 + 渲染品质，模型对"mockup"理解需要具体锚点。

Stylized 3D character of a friendly young fox wearing a tiny scarf, big expressive eyes, soft fur with subsurface scattering, neutral standing pose, three-quarter view, soft three-point lighting, Pixar-inspired character design, white seamless background, Octane render quality

photorealistic, scary, distorted anatomy, multiple characters

推荐模型： Midjourney, FLUX.1

角色设计要明写姿势 + 视角 + 中性背景，否则模型会自作主张加环境。

Documentary photojournalism shot inside a busy modern newsroom, editor in mid-conversation gesturing at a wall of monitors showing live news, shallow depth of field on the editor, ambient bokeh of fluorescent lights, shot on Leica Q3, slight grain, in the style of Magnum photography

staged, beauty filter, stock photo aesthetic, clean and tidy

推荐模型： Midjourney v7, FLUX.1

引用 Magnum / VII Photo 才有真纪实感 —— 笼统 "photojournalism" 容易出 stock 风。

Minimalist logomark design for a sustainability-focused tech startup, geometric leaf motif fused with a circuit pattern, single-color (forest green), clean negative space, vector-style flat design, isolated on white background, multiple variant exploration, professional brand identity

text, photographic, complex shading, multiple colors, ornate

推荐模型： Midjourney, FLUX.1

Logo 探索写 "multiple variant exploration"，能拿到一组变体挑选。

Editorial illustration for a New Yorker-style article on AI in healthcare, a doctor and a robot working together over a glowing diagnostic screen, muted blue and warm beige palette, subtle texture, expressive but restrained, hand-drawn feel with digital polish, in the style of Christoph Niemann

cute mascot, 3d render, photo, comic book, oversaturated

推荐模型： Midjourney, FLUX.1

编辑插画跟一般数字艺术不同。点名刊物（纽约客、NYT）能拿到对的调性。

Abstract 3D data visualization, intertwining ribbons of light flowing through a dark space, representing data streams, soft glow with chromatic aberration at edges, deep navy background with cyan and magenta highlights, organic curves with mathematical precision, suitable as a hero image for a tech company

推荐模型： Midjourney, FLUX.1

抽象数据图最好用具象短语（ribbons of light、data streams），纯抽象描述效果差。

Anime style, vast magical library interior with towering bookshelves reaching impossibly high, floating books and glowing magical orbs drifting through warm beams of light from stained glass windows, a young apprentice in robes reading in a cozy nook, Studio Ghibli inspired warm color palette, painterly background

photoreal, 3d, modern setting, distorted perspective

推荐模型： NovelAI, Midjourney --niji

"Studio Ghibli inspired" 能稳定唤起温暖手绘感（提示词本身不涉版权）。

Architectural photograph of a modern minimalist museum building, clean geometric concrete forms with large reflective glass panels, taken at the blue hour with the interior lights just beginning to glow, dramatic long shadow from late sun, ultra-sharp, shot on Hasselblad H6D, in the style of Iwan Baan

people in frame, motion blur, distorted perspective, vignetting

推荐模型： Midjourney v7, FLUX.1

建筑题材点名真实建筑摄影师（Iwan Baan、Hélène Binet），模型懂这些惯例。

Cinematic POV shot from the passenger seat of a car driving slowly through a neon-lit Tokyo street at night, rain on the windshield blurring the colorful signs, gentle motion blur, anamorphic lens flare from passing lights, soft city ambient lighting, 6 seconds, smooth tracking

推荐模型： Sora, Runway Gen-3, Veo

POV 镜头写明座位（"passenger seat"）锚定取景。

Seamless morphing transition video, a snowy mountain landscape gradually transforms into a tropical beach, sky color shifts from cold blue to warm orange, particles drifting through the transition, 5 seconds, smooth interpolation, dreamlike quality, no harsh cuts

推荐模型： Runway Gen-3, Kling

变形转场最好让天/地有物理映射（地平线对地平线）。

Macro slow-motion video of espresso pouring into a white ceramic cup, golden crema forming on top, steam rising slowly, dramatic side lighting against dark background, shallow depth of field, 1000fps style smoothness, 5 seconds, no camera movement

推荐模型： Runway Gen-3, Sora

液体视频：明写慢动作 + 蒸汽/油脂细节 + 相机不动 = 稳定出诱人成片。

Looping abstract animation of luminous particles drifting upward through deep navy void, with occasional bright flashes, cyan and purple bioluminescent palette, slow ethereal motion, suitable as a background for product launch hero section, 8 seconds, seamless loop

推荐模型： Runway Gen-3, Pika

背景视频写 "seamless loop"，结尾接回开头才能循环播放。