🏛 MusePublic新手教程：正面Prompt编写5大原则与10个优质示例
1. 为什么Prompt写得好，图才真正“有艺术感”
你有没有试过输入“一个穿裙子的女人站在海边”，结果生成的图里人物僵硬、光影平庸、背景空洞，完全不像杂志封面？这不是模型不行，而是——提示词没写对。
MusePublic不是普通文生图工具。它专为艺术感时尚人像打磨过：能捕捉模特微微侧头时下颌线的弧度，能还原柔光打在丝绸裙摆上的细微反光，能讲出一张图里的故事感——比如“雨后咖啡馆窗边低头翻书的她，发梢微湿，指尖停在某一页，窗外梧桐叶影斑驳”。
但这一切的前提是：你得告诉它足够具体、足够有画面感的语言。不是堆砌形容词，而是像导演给摄影师说戏一样，用可感知的细节构建视觉逻辑。
本教程不讲晦涩参数，不列抽象理论。只聚焦一件事：怎么用最自然的方式，写出MusePublic真正“听得懂”的正面Prompt。你会学到5条实操原则（每条都来自上百次生成失败与惊艳的总结），并直接套用10个已验证效果出色的中英混合示例——复制粘贴就能出图，改两个词就能换风格。
2. 写好Prompt的5大核心原则（小白也能立刻上手）
2.1 原则一：从“主体”出发，先定人，再加戏
别一上来就写“梦幻光影+复古胶片+高级灰调”。MusePublic最擅长的是人像，它的底层理解优先级是：人物 > 姿态 > 表情 > 服装 > 环境 > 风格。
正确顺序：
A confident East Asian woman in her late 20s, standing tall with one hand on her hip, slight smile, sharp cheekbones, wearing a tailored ivory blazer and high-waisted black trousers
常见错误：
Dreamy cinematic lighting, vintage film grain, muted color palette, elegant woman
→ 模型不知道“elegant woman”长什么样，更不知道“tall”还是“slim”，结果就是模糊的剪影。
小技巧：写Prompt时，心里默念三遍“她是谁？她在做什么？她此刻在想什么？”——答案就是你的第一句。
2.2 原则二：姿态比五官更重要，用动词代替形容词
MusePublic对动态描述极其敏感。“Standing with weight on left leg, right knee slightly bent” 比 “graceful pose” 有效10倍。因为“graceful”是主观感受，而“weight on left leg”是物理状态，模型能精准建模。
试试对比：
- “beautiful face, elegant posture” → 模型随机组合五官和站姿
- “head tilted 15 degrees to the left, eyes gazing just past the camera, shoulders relaxed but spine straight, left hand resting lightly on a marble windowsill”
关键点：多用身体部位+方向+角度+接触状态（如：fingertips brushing the collarbone,elbow bent at 90 degrees）。这些是模型训练数据里高频出现的真实人体结构描述。
2.3 原则三：光影必须“可定位”，拒绝空泛术语
“Soft lighting” 是无效的。MusePublic需要知道光从哪来、打在哪、产生什么效果。
有效写法：
Late afternoon sun streaming through a tall arched window, casting long diagonal shadows across her face and catching golden highlights on her hair
→ 光源（late afternoon sun）、路径（streaming through… window）、作用（casting shadows, catching highlights）、位置（across her face, on her hair）全部明确。
小提醒：MusePublic对“window light”“candlelight”“overhead studio light”这类真实光源词响应极佳，但对“ethereal glow”“magical aura”等虚化词容易失控。
2.4 原则四：背景不是“陪衬”，而是“叙事伙伴”
很多新手把背景写成“simple white background”或“blurry cityscape”，这等于放弃了一半故事感。MusePublic的强项恰恰在于环境与人物的情绪共振。
高阶写法：
She stands in front of a rain-streaked floor-to-ceiling window, blurred neon signs of a Tokyo street glowing faintly behind her, reflection of city lights shimmering on the wet glass beside her temple
→ 背景交代了地点（Tokyo）、时间（夜晚）、天气（rain）、情绪（疏离又迷幻），还通过“reflection”“shimmering”把人物和环境物理连接。
注意：背景描述长度建议控制在整段Prompt的30%-40%，太长会稀释人物主体性。
2.5 原则五：风格要“可验证”，少用玄学词，多用作品/作者/媒介
“Artistic style”“cinematic vibe” 这类词模型无法映射到具体视觉特征。但“in the style of Annie Leibovitz’s Vogue portraits” 或 “shot on Kodak Portra 400 film, medium format” 就非常明确——因为MusePublic的训练数据里，有大量Leibovitz的构图逻辑和Portra胶片的色彩科学。
推荐组合：
- 摄影师名：Annie Leibovitz, Peter Lindbergh, Ellen von Unwerth
- 电影镜头：Kodak Portra 400, Fujifilm Superia X-TRA 400, expired Polaroid SX-70
- 绘画流派：Renaissance oil painting, Art Deco illustration, Japanese ukiyo-e woodblock
- 杂志质感：Vogue Italia editorial, i-D magazine 90s issue, Dazed & Confused cover
记住：风格词放在Prompt末尾，用逗号隔开，不超过2个，避免冲突。
3. 直接可用！10个已实测的优质Prompt示例
以下所有示例均在MusePublic WebUI中实测生成成功（24G显存，30步，EulerAncestral调度器），无需修改即可一键出图。每个都标注了核心设计逻辑，方便你举一反三。
3.1 经典杂志风：优雅与力量并存
A South Korean model in her early 30s, wearing a sculptural black leather trench coat, standing confidently on a rain-slicked Parisian sidewalk at dusk, dramatic chiaroscuro lighting from a single streetlamp above, sharp focus on her determined gaze and wind-blown hair, Vogue Paris editorial style, shot on Leica M11
亮点：用“sculptural coat”定义服装质感，“rain-slicked sidewalk”锚定环境，“chiaroscuro lighting”精准控制明暗对比，“Vogue Paris editorial”锁定高级时装摄影语感。
3.2 复古胶片感：温柔叙事
A young Black woman with natural afro hair, sitting cross-legged on a sun-dappled wooden floor, holding a vintage paperback, soft diffused light from a large north-facing window, warm Kodak Portra 400 color palette, shallow depth of field, gentle bokeh background of bookshelves
亮点：“sun-dappled wooden floor”带出光线质感，“vintage paperback”暗示时代感，“north-facing window”是专业摄影中柔和光的标志，“Kodak Portra 400”直接调色。
3.3 都市诗意：静谧张力
An East Asian non-binary artist in their late 20s, leaning against a graffiti-covered brick wall in Berlin, wearing oversized denim jacket and chunky boots, looking thoughtfully down at their hands holding a sketchbook, moody overcast light, cinematic shallow focus, film grain texture, inspired by Wong Kar-wai's 'In the Mood for Love'
亮点：用“graffiti-covered brick wall”定义城市肌理，“moody overcast light”替代“dramatic lighting”，“Wong Kar-wai”精准传递色调与情绪节奏。
3.4 极简高级感：留白哲学
A Scandinavian woman with platinum blonde bob cut, seated on a minimalist white concrete bench in a vast empty gallery space, wearing an oversized cream wool sweater, hands resting gently on knees, even ambient light from hidden ceiling fixtures, ultra-clean composition, negative space dominant, aesthetic of Jil Sander SS24 campaign
亮点：“vast empty gallery space”强化空间感，“hidden ceiling fixtures”解释光源，“Jil Sander SS24”绑定极简主义基因。
3.5 自然光影实验：皮肤质感突破
Close-up portrait of an Indian woman with deep brown skin and intricate henna patterns on her hands, facing slightly left, soft directional light from a large softbox creating gentle catchlights in her eyes and subtle texture on her cheekbones, hyperrealistic skin detail, shallow depth of field, Canon EOS R5 portrait mode
亮点：特写+“directional light from softbox”确保皮肤纹理清晰，“hyperrealistic skin detail”是MusePublic对人像细节的强响应指令，“Canon EOS R5”强化真实相机感。
3.6 动态瞬间捕捉：打破静态人像
A Latin American dancer mid-pirouette on a polished wooden studio floor, wearing a crimson silk leotard and tulle skirt frozen in motion, sweat glistening on her temples, dynamic blur on skirt edges, strong backlight from high windows creating rim light around her silhouette, dance photography by Lois Greenfield
亮点：“mid-pirouette”定义动作，“frozen in motion”+“dynamic blur”制造速度感，“rim light”是专业舞蹈摄影关键词，“Lois Greenfield”是动态人像权威。
3.7 文化符号融合：东方意境
A Chinese woman in her 40s with silver-streaked black hair, wearing a modern reinterpretation of hanfu with asymmetrical silk sleeves, standing beneath a blooming wisteria trellis in Kyoto, soft morning mist, delicate petals falling, traditional ink wash painting texture overlay, subtle gold leaf accents
亮点：“modern reinterpretation of hanfu”平衡传统与当代，“wisteria trellis in Kyoto”锁定文化场景，“ink wash painting texture”提供视觉基底，“gold leaf accents”增加精致度。
3.8 未来主义肖像：科技与人文
A Nigerian tech entrepreneur in her 30s, wearing a sleek chrome-finish bodysuit with bioluminescent circuit patterns glowing softly on her arms, standing before a curved holographic interface displaying data streams, cool blue ambient light reflecting on her skin, cyberpunk elegance, shot on Hasselblad X2D
亮点：“bioluminescent circuit patterns”比“glowing lines”更具体，“holographic interface”定义科技语境，“cyberpunk elegance”精准限定风格边界。
3.9 暖调生活切片：烟火气真实感
A Mexican grandmother with warm brown skin and kind eyes, laughing while stirring a large clay pot of mole sauce in a sunlit adobe kitchen, steam rising, copper pots hanging on stucco wall, rich earthy color palette, natural documentary photography style, Leica Q3
亮点：“stirring a large clay pot”定义动作，“steam rising”增加动态，“adobe kitchen”“copper pots”构建可信环境，“documentary photography”强调真实感。
3.10 黑白情绪肖像：回归本质
Extreme close-up of an elderly Japanese man’s face, deep wrinkles like topographic maps, eyes closed, hands folded in his lap, stark black and white contrast, dramatic Rembrandt lighting from upper left, grainy Ilford HP5 Plus film texture, portrait by Dorothea Lange
亮点：“extreme close-up”强制聚焦，“wrinkles like topographic maps”提供具象比喻，“Rembrandt lighting”是经典人像布光，“Ilford HP5 Plus”定义粗粒胶片质感。
4. 避坑指南：3个高频失败原因与解法
4.1 问题：生成图人物比例失调（头太大/腿太短/手变形）
→根源：Prompt中缺少明确的姿态约束，或混入矛盾指令（如同时写“standing tall”和“crouching”）
→解法：在Prompt开头固定一句基础姿态，例如：
Full-body portrait, standing upright with balanced weight distribution, natural human proportions
4.2 问题：光影混乱，人物像被“打碎”
→根源：用了多个冲突光源（如“sunlight + neon lights + candlelight”），或未指定主光源方向
→解法：只保留1个主光源+1个辅助光，用方位词锁定：
Key light: soft window light from upper left; Fill light: subtle bounce light from white wall on right
4.3 问题：风格跑偏，变成油画/插画/3D渲染
→根源：风格词过多或自相矛盾（如“photorealistic + watercolor painting”）
→解法：严格遵循“1个摄影师/1个胶片/1个媒介”原则，且放在Prompt末尾。若需微调，用程度副词：
slightly softened focus, very subtle film grain, minimal color grading
5. 总结：让每一次输入，都成为一次精准的艺术对话
写Prompt不是填空，而是与MusePublic进行一场关于视觉语言的深度协作。它早已学会识别“an ivory blazer”和“a cream jacket”的微妙差异，能理解“gazing just past the camera”里那种若有所思的距离感，也记得“Kodak Portra 400”在阴影处特有的暖棕过渡。
所以，放下“越多越好”的执念。记住这5个锚点：
- 主体先行：先说清“谁”，再谈“美”；
- 姿态即语言：用动词写身体，比用形容词写感觉更有力；
- 光影要落点：光从哪来、打在哪、留下什么痕迹；
- 背景会说话：让它参与叙事，而非充当模糊背景板；
- 风格可验证：用真实存在的作品、媒介、作者建立共识。
现在，打开你的MusePublic WebUI，选一个示例，点击「 开始创作」。当第一张图在屏幕上缓缓浮现——那个眼神、那道光、那份恰到好处的留白——你会明白：技术从未如此贴近直觉，而艺术，始终始于一句准确的描述。
获取更多AI镜像
想探索更多AI镜像和应用场景？访问 CSDN星图镜像广场，提供丰富的预置镜像，覆盖大模型推理、图像生成、视频生成、模型微调等多个领域，支持一键部署。
