AI Prompt Engineering · 课时

风格与艺术媒介规范

Q: 「风格与艺术媒介规范」这节课中我会学到什么？

照片级写实、油画、水彩、3D 渲染——风格词汇。 你通过在浏览器中直接运行的动手代码来练习 AI Prompt Engineering，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

照片级写实、油画、水彩、3D 渲染——风格词汇。

风格与艺术媒介规范 是 CoddyKit 上的免费 AI Prompt Engineering 课时。这是第 2 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 AI Prompt Engineering 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 AI Prompt Engineering 课程共包含 4 节课。

风格词汇为何重要

图像模型基于数十亿张带有说明文字的图像进行训练。它们“了解”特定风格、艺术流派和艺术家的视觉语言。使用精确的风格词汇，就像在说模型的母语——它可以直接激活模型学到的视觉模式，而不是依赖含糊的描述。

写实照片风格

写实照片提示词会指示模型生成与真实照片难以区分的图像。以下是关键词汇及其应用方式：

photorealistic_prompts = [
    # Basic photorealism
    'photorealistic portrait of a woman, DSLR photograph, '
    'f/1.8 aperture, 85mm lens, bokeh background, natural light',

    # Hyperrealism (even more detailed than photo)
    'hyperrealistic painting of a weathered old door, '
    'every crack and grain of wood visible, photorealistic quality',

    # Studio photography
    'product photography, white background, studio lighting, '
    'sharp focus, commercial quality, shot on Canon EOS R5',

    # Environmental photography
    'documentary photograph, photojournalism style, '
    'natural unposed lighting, 35mm film grain, raw authentic moment',
]

for prompt in photorealistic_prompts[:2]:
    print(prompt[:80], '...')

# Key terms for photorealism
photo_keywords = [
    'DSLR', 'shot on [camera model]', 'f/[aperture]', '[mm] lens',
    'bokeh', 'depth of field', 'film grain', 'natural light'
]

油画风格

油画涵盖了数百年间形成的多种独特传统。指定油画内部的具体风格——巴洛克、印象派或荷兰黄金时代大师风格——会激活截然不同的视觉词汇。

oil_painting_styles = {
    'classical_baroque': (
        'oil painting in the style of Rembrandt, '
        'chiaroscuro lighting, rich earth tones, '
        'detailed fabric and texture, dramatic shadows, '
        'museum quality, Dutch Golden Age'
    ),
    'impressionist': (
        'impressionist oil painting inspired by Monet, '
        'visible brushstrokes, soft edges, '
        'light and atmosphere over detail, '
        'en plein air quality, dappled light'
    ),
    'post_impressionist': (
        'oil painting in the bold style of Van Gogh, '
        'swirling expressive brushwork, '
        'complementary color contrasts, '
        'thick impasto texture, emotional intensity'
    ),
    'contemporary_realism': (
        'contemporary oil painting, hyperrealistic, '
        'smooth blending, photorealistic quality, '
        'intense focus on textures, classical technique'
    )
}

for name, prompt in oil_painting_styles.items():
    print(f'{name}:')
    print(' ', prompt[:70], '...')
    print()

水彩风格词汇

水彩具有鲜明的视觉特征：柔和的边缘、透明的晕染、可见的纸张纹理，以及色彩的自然渗化。以下词汇可以激活这些特质：

watercolor_prompts = [
    # Classic watercolor
    'delicate watercolor illustration of cherry blossoms, '
    'soft washes of pink and white, paper texture visible, '
    'loose brushwork, gentle bleeding edges, '
    'traditional Japanese botanical art style',

    # Urban sketching style
    'watercolor urban sketch of a Paris cafe, '
    'loose gestural lines with ink, '
    'selective color washes over pen sketch, '
    'travel journal illustration style',

    # Children\'s book style
    'charming watercolor illustration for a children\'s book, '
    'soft pastel colors, cute rounded shapes, '
    'whimsical details, storybook warmth',
]

# Key watercolor vocabulary
watercolor_keywords = [
    'watercolor wash', 'soft edges', 'translucent', 'paper texture',
    'bleeding colors', 'wet-on-wet technique', 'loose brushwork',
    'granulation', 'wet-on-dry', 'color bloom'
]

print('Watercolor example:', watercolor_prompts[0][:100], '...')

三维渲染风格

三维渲染涵盖多种独具特色的视觉效果。指定渲染引擎和软件名称，可以激活特定的视觉质量和审美风格。

rendering_styles = {
    'octane_render': (
        '3D render, Octane renderer, photorealistic, '
        'subsurface scattering on skin, ray-traced reflections, '
        'global illumination, cinematic depth of field, '
        'highly detailed environment'
    ),
    'blender_stylized': (
        'Blender 3D render, stylized low-poly art, '
        'smooth shading, vibrant colors, '
        'isometric perspective, indie game aesthetic'
    ),
    'unreal_engine': (
        'Unreal Engine 5 render, photoreal environment, '
        'Lumen global illumination, Nanite geometry, '
        'cinematic quality, game engine screenshot'
    ),
    'clay_render': (
        '3D clay render, matte white material, '
        'soft studio lighting, clean shadows, '
        'product design visualization, no textures'
    ),
    'vfx_compositing': (
        'VFX quality 3D render, photorealistic compositing, '
        'motion blur, film grain, color grading, '
        'blockbuster movie quality'
    )
}

print('Octane render example:',  rendering_styles['octane_render'][:80], '...')

漫画与插画风格

线条表现力强的插画风格需要描述线条质感、明暗处理技法和色彩处理方式的词汇。漫画风格是最受要求的插画风格之一。

illustration_styles = {
    'comic_book': (
        'comic book style, bold black outlines, '
        'flat cel-shaded colors, halftone dot pattern, '
        'dynamic action pose, 1960s Marvel Comics aesthetic'
    ),
    'manga': (
        'manga illustration style, clean line art, '
        'screen tones for shading, large expressive eyes, '
        'black and white with selective screentone, '
        'Shonen Jump aesthetic'
    ),
    'graphic_novel': (
        'graphic novel illustration, Moebius-inspired, '
        'detailed line work, flat color fills, '
        'SF aesthetic, intricate background architecture'
    ),
    'vector_flat': (
        'flat vector illustration, minimal outlines, '
        'geometric shapes, limited color palette, '
        'modern app icon style, clean and scalable'
    ),
    'concept_art': (
        'concept art for video game, detailed character design, '
        'front view turnaround sheet, '
        'multiple color variants, reference quality'
    )
}

for style, prompt in list(illustration_styles.items())[:3]:
    print(f'{style}: {prompt[:60]}...')

艺术家风格参考

提及特定艺术家会激活模型所学到的该艺术家的视觉特征。请使用“以……为风格”或“受……启发”，确保模型将其理解为风格参考，而不是要求生成该艺术家的实际作品。

artist_style_prompts = [
    # Classic masters
    'oil painting inspired by Vermeer, domestic interior scene, '
    'soft diffused window light from the left, pearl earring quality',

    # Impressionists
    'painting in the style of Monet, water lilies and reflections, '
    'dappled light, loose impressionist brushwork',

    # Modern
    'illustration in the style of Alphonse Mucha, '
    'art nouveau ornamental border, flowing hair, '
    'decorative circular frame, spring theme',

    # Contemporary illustrators
    'digital illustration in the style of James Gurney, '
    'painterly quality, dinosaur, Dinotopia aesthetic',

    # Graphic designers
    'poster design inspired by Saul Bass, '
    'geometric minimalism, bold primary colors, '
    'mid-century modern style'
]

# Best practice: 'in the style of' activates learned style patterns
# Not appropriate: 'a painting by [living artist]' (may raise IP concerns)
print('Style reference example:', artist_style_prompts[2][:80], '...')

摄影类型风格

每种摄影类型都有独特的视觉词汇。指定摄影类型可以立即确定预期的视觉语言、构图惯例和后期处理审美。

photography_genres = {
    'street_photography': (
        'street photography, candid unposed moment, '
        'grain, black and white, Henri Cartier-Bresson style, '
        'decisive moment, natural light'
    ),
    'fashion_editorial': (
        'high fashion editorial photography, '
        'Vogue magazine quality, strong directional lighting, '
        'model on textured white backdrop, '
        'Avedon-inspired minimalism'
    ),
    'macro_nature': (
        'macro photography of dewdrops on a spider web, '
        'extreme shallow depth of field, '
        'lens bokeh background, natural dawn light, '
        '1:1 macro ratio'
    ),
    'astro_photography': (
        'astrophotography, milky way over mountain landscape, '
        'long exposure, star trails, '
        'foreground silhouette, deep space colors'
    ),
    'documentary': (
        'documentary photography, photojournalism, '
        'natural ambient light, authentic emotion, '
        'Magnum Photos aesthetic, raw unedited'
    )
}

for genre, prompt in list(photography_genres.items())[:3]:
    print(f'{genre}: {prompt[:60]}...')

风格混搭：创意组合

一些最有趣的图像提示词，会将通常不会共存的风格结合起来——结果是一种独特的混合美学。关键在于连贯的组合：共享视觉 DNA 的风格，比彼此冲突的风格更容易融合。

# Coherent style combinations
coherent_combos = [
    # Shared aesthetic DNA
    'art nouveau meets watercolor illustration, '
    'organic flowing forms, soft pastel palette',

    'vintage film photography meets dark academia aesthetic, '
    'sepia tones, dust and grain, leather-bound books',

    'Studio Ghibli inspired oil painting, '
    'lush pastoral landscape, impressionist brushwork, '
    'gentle nostalgic atmosphere',

    'cyberpunk neon aesthetic meets traditional Japanese woodblock print, '
    'bold outlines, flat colors, ukiyo-e composition'
]

# Incoherent combinations (avoid)
incoherent = [
    'minimalist AND maximalist AND baroque AND minimalist',  # contradictory
    'photorealistic AND cartoon AND oil painting AND pixel art',  # too many
]

print('Best combo example:', coherent_combos[3][:80], '...')
print('Warning: avoid contradictory combinations like:', incoherent[0][:50])

特定模型的风格技巧

不同的图像模型对风格词汇的响应各不相同。DALL-E 3、Midjourney 和 Stable Diffusion 各有自己的优势、特性以及偏好的提示词写法。

model_style_notes = {
    'DALL-E 3 (OpenAI)': [
        'Responds well to natural language descriptions',
        'Artist name references work well',
        'Avoid very long prompt strings (>300 words loses coherence)',
        'Instruction-following is strong ("Do not include text")',
        'Good for: illustrations, portraits, abstract concepts'
    ],
    'Midjourney': [
        'Aesthetic quality is very high with minimal prompting',
        'Suffix parameters: --ar 16:9 for aspect ratio, --v 6 for version',
        'Style references: --sref for image style transfer',
        'Photography terms work especially well',
        'Good for: artistic renders, character design, landscapes'
    ],
    'Stable Diffusion': [
        'Negative prompts are very important and effective',
        'LoRA models enable fine-grained style control',
        'CFG scale controls prompt adherence (7-12 recommended)',
        'Model checkpoints determine base style (e.g., SDXL, Juggernaut)',
        'Good for: customization, fine-tuning, commercial use'
    ]
}

for model, tips in model_style_notes.items():
    print(f'{model}:')
    for tip in tips[:2]:
        print(f'  - {tip}')

风格表：快速参考

风格参考表会将项目或品牌认可的词汇规范化，确保团队成员和工具生成的图像保持一致。

# Project style sheet — game UI concept art
GAME_STYLE_SHEET = {
    'base_style': 'dark fantasy concept art, game UI illustration quality',
    'medium': 'digital painting with painterly brushwork',
    'lighting': 'dramatic rim lighting, deep shadows, magical glow effects',
    'color_palette': 'deep purples, midnight blues, gold accents, shadow blacks',
    'composition': 'centered composition for icons, rule of thirds for scenes',
    'quality': 'highly detailed, professional game art, AAA quality',
    'forbidden': [
        'photorealistic', 'photograph', 'anime', 'cartoon',
        'watercolor', 'sketch'
    ]
}

def build_game_prompt(subject, composition_override=None):
    s = GAME_STYLE_SHEET
    return ', '.join([
        subject,
        s['base_style'],
        s['medium'],
        s['lighting'],
        s['color_palette'],
        composition_override or s['composition'],
        s['quality']
    ])

print(build_game_prompt('a fire mage casting a spell, flames erupting from hands')[:120], '...')

快速检查

您想生成一幅看起来像经典荷兰黄金时代油画的图像，并带有戏剧性的阴影和烛光。应该使用哪些风格术语？

风格与媒介总结

指定风格和媒介，可以精确控制生成图像的视觉语言：

媒介：艺术形式决定纹理、线条质感和渲染美学
风格词汇：艺术运动术语和艺术家参考会激活模型学习到的视觉模式
摄影类型：街拍、时尚、微距、纪实摄影各自具有独特的惯例
三维渲染：引擎名称（Octane、Blender、Unreal）会触发对画面质量等级的预期
风格混搭：结合具有共同视觉 DNA 的风格，形成连贯的混合美学
风格表：规范项目词汇，确保整个团队保持一致

常见问题解答

「风格与艺术媒介规范」课时是免费的吗？

是的 — 「风格与艺术媒介规范」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 AI Prompt Engineering 课程的其余内容，请升级到 CoddyKit PRO。 AI Prompt Engineering 课程共包含 4 节课。

「风格与艺术媒介规范」这节课中我会学到什么？

照片级写实、油画、水彩、3D 渲染——风格词汇。你通过在浏览器中直接运行的动手代码来练习 AI Prompt Engineering，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 AI Prompt Engineering 需要有经验吗？

无需任何先前经验。CoddyKit 上的 AI Prompt Engineering 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 2 节课，共 4 节。

「风格与艺术媒介规范」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 AI Prompt Engineering 课中编写并运行代码吗？

能。每节 AI Prompt Engineering 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 AI Prompt Engineering