YOHO AI

博客

  • 高端利姆卡饮料广告海报设计

    高端利姆卡饮料广告海报设计

    高端利姆卡饮料广告海报设计

    高端利姆卡饮料广告海报设计


    🤖 ChatGPT

    🇺🇸 English Prompt

    Create a premium, hyper-realistic Instagram SMM poster for Limca that feels like an official Coca-Cola India × Ogilvy campaign. Avoid generic beverage ads, floating bottles, stock fitness imagery, excessive splashes, AI artifacts, clutter, or cartoonish elements.

Concept: Capture the authentic moment of refreshment after effort. A young Indian adult has just finished running, cycling, playing cricket, basketball, or walking through intense summer heat. They twist open an ice-cold Limca bottle, with condensation, carbonation, and subtle mist emphasizing the instant feeling of relief and revitalization.

Setting: A warm Indian urban environment such as a running track, cricket ground, basketball court, rooftop, or city street during golden summer light. Contrast the hot surroundings with the icy freshness of the drink.

Composition (4:5 | 1080×1350):
• 40% human action
• 35% Limca product hero
• 15% environment
• 10% typography & CTA

Show the Limca bottle with crystal-clear branding, realistic condensation, chilled glass, micro bubbles, and premium beverage photography lighting. Include subtle lime accents, clean editorial motion graphics, and minimal freshness overlays.

Color palette: Lime Green, Lemon Yellow, Crisp White with Cool Mint and Soft Silver accents.

Headline: **FEEL THE ZING** (large bold editorial sans-serif)
CTA: **Refresh Now**

Lighting: Luxury commercial beverage lighting with warm sunlight, bright backlight, clean rim light, cold specular highlights, and natural shadows.

Style: Behance-featured quality, Coca-Cola India commercial, premium editorial advertising, hyper-realistic photography, authentic Indian lifestyle, energetic, refreshing, modern, minimal, ultra-detailed.

    🇨🇳 中文提示词

    为利姆卡（Limca）创作一张高端、超现实的 Instagram SMM 海报，感觉就像是可口可乐印度公司与奥美（Ogilvy）的官方宣传活动。避免通用的饮料广告、漂浮的瓶子、库存健身图像、过多的水花、AI 伪影、杂乱或卡通元素。

概念：捕捉努力后的真实清爽瞬间。一名年轻的印度成年人刚刚完成跑步、骑行、打板球、打篮球或在酷热的夏日行走。他们拧开一瓶冰凉的利姆卡，瓶身上的冷凝水、碳酸气泡和细微的水雾强调了瞬间的解脱和恢复活力的感觉。

环境：温暖的印度城市环境，如跑道、板球场、篮球场、屋顶或夏日金光下的城市街道。将炎热的环境与饮料的冰爽清新形成对比。

构图 (4:5 | 1080×1350)：
• 40% 人物动作
• 35% 利姆卡产品主角
• 15% 环境
• 10% 排版与行动呼吁 (CTA)

展示带有清晰品牌标识、真实冷凝水、冷冻玻璃、微气泡和高端饮料摄影灯光的利姆卡瓶子。包含细微的青柠装饰、简洁的编辑级动态图形和极简的清新覆盖层。

调色板：青柠绿、柠檬黄、纯净白，搭配酷凉薄荷和柔和银色色调。

标题：**FEEL THE ZING** (大号粗体编辑类无衬线字体)
行动呼吁 (CTA)：**Refresh Now**

灯光：奢华商业饮料灯光，带有温暖阳光、明亮逆光、简洁的轮廓光、寒冷的镜面高光和自然阴影。

风格：Behance 推荐质量、可口可乐印度广告、高端编辑广告、超现实摄影、真实印度生活方式、充满活力、清新、现代、极简、超精细。
  • 明亮清透仪式感网球公开赛海报

    明亮清透仪式感网球公开赛海报

    明亮清透仪式感网球公开赛海报

    明亮清透仪式感网球公开赛海报


    🤖 ChatGPT

    🇺🇸 English Prompt

    Generate a bright, clear, and ceremonial visual poster around any subject: The composition establishes a primary order through a strict central axis, with the main form or symbolic structure of the subject standing at the distant vanishing point, elevated like a monumental landmark while maintaining clear details; the foreground spreads out a wide subject field that converges toward the center, with sidelines, stepped tiers, or structures on both sides forming a strong perspective corridor, and a small-scale viewer avatar or subject carrier placed below the center as a scale anchor. The upper portion leaves an open, high-brightness light field with a strong sense of ventilation, allowing light circular symbols derived from the subject to float along both sides of the central axis from large to small, scattered front to back, creating a rhythm and depth like a ball rain, which must not dissipate into ordinary decoration. If text appears, place it before the main visual at the top, using a spacious uppercase feel, clean geometric strokes, large letter spacing, and two font sizes; the information color should be clear and restrained, sharing the central axis with the main form; the bottom features a small circular seal and a short sentence as a conclusion, without overshadowing the main visual. Colors are extracted from the subject's own materials, scene mood, and cultural semantics, retaining large areas of high-brightness clear base colors, wide warm or warm-cold balanced ground color blocks, a small amount of high-saturation glowing accent colors, clear dark information text, and clean shadow role relationships; the overall feel is bright and light, with sufficient airiness, clear color gradients, and textures consisting only of fine print grain and soft sunlight, avoiding gray, dull, turbid, or crowded collage feelings.

——————
Event Theme: Tennis Open Finals Day + Championship Trophy + Spectator Stands + Brand Seal
Publication Use: Event schedule announcement cover + Stadium guidance screen main image
Warm Colors: Clay orange + Milky white sunlight + Dark green fencing + Soft warm light
Visual Focus: Foreground trophy close-up + Side spectator stands corridor + Sky tennis ball arcs + Vertical 9:16

    🇨🇳 中文提示词

    围绕任意主题对象生成一张明亮、清透、具有仪式感的视觉海报：画面以严格中轴建立第一眼秩序，主题的主形体或象征结构立在远处消失点，像纪念性地标一样被拉高并保持细节清晰；前景铺开一条宽阔的主题场域向中心收束，边线、阶梯式层级或两侧结构形成强透视夹道，中心下方放置一个小尺度观者化身或主题载体作为尺度锚点。上方留出通风感很强的开放高明度光场，让主题派生的轻质圆形符号沿中轴两侧由大到小悬浮、前后错落，形成像球雨一样的节奏和纵深，不能散成普通装饰。文字若出现，放在上方主视觉之前，使用疏朗大写感、干净几何笔画、较大字距和两级字号，信息色清晰克制，与主形体共享中轴；底部设置小型圆章和短句作为收束，不抢主视觉。色彩从主题自身的材质、场景情绪和文化语义中提取，保留大面积高明度清透底色、宽阔温暖或冷暖平衡的地面色块、少量高饱和发光强调色、清晰深色信息文字与干净阴影的角色关系；整体明亮轻快、空气感充足、色阶清晰，纹理只做细腻印刷颗粒和柔和日光，避免灰旧、浑浊或拥挤的拼贴感。


——————
活动主题: 网球公开赛决赛日+冠军奖杯+观众看台+品牌圆章
发布用途: 赛事日程公告封面+场馆导视屏主图
暖系色彩: 红土橙+奶白日光+深绿围挡+柔和暖亮
画面焦点: 前景奖杯近景+两侧看台夹道+天空网球弧线+竖版 9:16
  • 黄尾袋鼠梅洛红酒写实瓶身手绘探险海报

    黄尾袋鼠梅洛红酒写实瓶身手绘探险海报

    黄尾袋鼠梅洛红酒写实瓶身手绘探险海报

    黄尾袋鼠梅洛红酒写实瓶身手绘探险海报


    🤖 ChatGPT

    🇨🇳 中文提示词

    Create an ultra-premium advertising poster for Yellow Tail Merlot red wine, built around a strong contrast between one highly realistic wine bottle and a surrounding world of lively hand-drawn exploration-route line art. The image should feel clean, adventurous, graphic, playful, premium, and highly brand-specific, like a collectible illustrated wine poster on textured paper.

Main composition:
Place one realistic bottle of Yellow Tail Merlot exactly at the center as the absolute hero object. The bottle must be highly realistic and unmistakably Yellow Tail: deep merlot-red wine glow, crisp label, iconic yellow kangaroo mark clearly visible, elegant foil capsule, glossy glass reflections, realistic shadow, and premium wine-bottle rendering. The bottle must be the only fully realistic object in the image.

Background:
Use a very clean warm ivory or pale cream textured paper background with generous negative space. The paper should feel premium and lightly tactile, like an artist’s poster sheet. Keep the background open and uncluttered.

Core visual system:
Around the bottle, create a hand-drawn black-ink exploration route that wraps and travels around the bottle like a map-based adventure path. The line art should not behave like a decorative frame or random doodles. It must feel like a drawn route map unfolding around the bottle, with winding paths, directional arrows, signposts, handwritten labels, route marks, tiny icons, and small illustrated discoveries placed along the journey.

Exploration-route storytelling:
Make the bottle feel like the key landmark in the middle of an Australian wine adventure map. Along the route, integrate small hand-drawn symbols and discoveries related to Yellow Tail and Australian Merlot identity:
leaping kangaroo icon,
vineyard hills,
grape clusters,
eucalyptus sprigs,
sun symbols,
wine barrels,
harvest crate,
country road marks,
southern stars,
tiny birds,
wine glass icon,
open plains,
small cellar sketch,
route arrows,
distance marks,
handwritten place notes,
little exploration badges,
“stop here” or “discover this” style sign graphics.
The drawings should feel lively, curious, slightly playful, and full of discovery.

Layout behavior:
The route should move around the bottle in a dynamic way, as if circling and guiding the eye across the page. Some parts can curve behind the bottle, some can run beside it, some can create mini stops around top left, right side, lower side, and bottom areas. The composition should feel like a hand-drawn travel/adventure map centered on the bottle. Do not make the line art symmetrical or evenly distributed. Let it feel exploratory and organic.

Style:
The line art should be crisp black ink with a premium hand-drawn identity feel, combining route-map graphics, handwritten signage, and light illustrative storytelling. Add a subtle Art Nouveau influence only in the flowing vine curves and rhythmic line movement, but keep the overall feeling modern, clean, and brand-driven.

Color system:
warm ivory paper background,
deep merlot red bottle,
black ink route illustration,
small Yellow Tail signature yellow accents,
tiny touches of vineyard green only if necessary.
Keep the palette highly restrained so the bottle remains the most saturated and realistic focal point.

Typography:
Integrate typography into the route-map world rather than using a standard ad layout. Use small elegant hand-drawn or poster-style phrases woven into the exploration route, such as:
“YELLOW TAIL”
“MERLOT”
“LEAP INTO RED”
“EXPLORE THE TASTE”
“AUSTRALIA IN A BOTTLE”
These text fragments should feel like route labels, signposts, or little discovery notes around the bottle.

Optional subtle Chinese text can be integrated in a similarly restrained way:
“黄尾袋鼠 梅洛干红”
“澳洲风味探索”
“沿着果香一路前行”
Keep Chinese minimal, refined, and secondary.

Mood:
clean, graphic, adventurous, Australian, lively, route-based, exploratory, premium, collectible, high-contrast, hand-drawn, fresh, intelligent, brand-driven.

Rendering:
hyper-detailed realistic wine bottle, crisp black ink route drawing, premium textured paper background, elegant negative space, subtle paper shadows, luxury campaign poster quality, 8k, visually unforgettable.

Negative prompt:
random doodles, decorative border frame, symmetrical line-art frame, photorealistic landscape background, crowded illustration, too many colors, weak bottle focus, generic wine ad, childish cartoon style, low detail, blurry label, cluttered text, flat composition
  • 卧室镜前自拍的东亚年轻美少女

    卧室镜前自拍的东亚年轻美少女

    卧室镜前自拍的东亚年轻美少女

    卧室镜前自拍的东亚年轻美少女


    🤖 Grok

    🇺🇸 English Prompt

    Photorealistic mirror selfie of a beautiful young East Asian woman with long black hair and bangs, wearing a small gold hair clip. She has fair skin and subtle makeup with eyeliner. She is taking a selfie in a bedroom while looking directly at the camera.

She is wearing a tight blue short-sleeve crop top and dark blue star-patterned short shorts with white stars and text on the back. The outfit is very form-fitting. She is barefoot.

Pose: standing in front of a mirror, one leg lifted and bent with foot resting on a pink surface, body slightly turned, holding a light-colored smartphone with both hands to take the selfie. Confident and playful pose.

Setting: modern bedroom with wooden flooring, large white fluffy rug, open white closet with clothes inside, scattered clothes on a pink surface, pink vanity chair, and a white humidifier. Natural indoor lighting, realistic details, sharp focus, photorealistic, 8k resolution --ar 2:3 --stylize 180

    🇨🇳 中文提示词

    一张照片般真实的一位美丽的年轻东亚女性的镜前自拍，她留着黑色长发和刘海，戴着一个小金发夹。她皮肤白皙，画着带眼线的淡妆。她在卧室里自拍，同时直视镜头。

她穿着一件紧身的蓝色短袖露脐上衣和深蓝色星形图案的短裤，背面有白色星星和文字。这身打扮非常贴身。她赤着脚。

姿势：站在镜子前，一条腿抬起并弯曲，脚落在粉红色表面上，身体微侧，双手握着一部浅色智能手机自拍。自信且俏皮的姿势。

场景：带有木地板的现代卧室，巨大的白色蓬松地毯，放着衣服的敞开式白色衣橱，粉红色表面上散落的衣服，粉红色梳妆椅，以及一台白色加湿器。自然室内光线，真实细节，清晰聚焦，照片级真实，8k分辨率 --ar 2:3 --stylize 180
  • 雀巢牛奶纯白圣洁奢华海报

    雀巢牛奶纯白圣洁奢华海报

    雀巢牛奶纯白圣洁奢华海报

    雀巢牛奶纯白圣洁奢华海报


    🤖 ChatGPT

    🇺🇸 English Prompt

    Create an ultra-premium **pure white sacred luxury poster** for **Nestle Milk**, centered on the theme of **a mature East Asian girl ascending in flight and seamlessly fused with flowing milk**, as if she is being born from light, purity, and liquid softness itself. The entire image must feel holy, serene, weightless, and breathtakingly refined, like a fusion of high-end beauty photography, sacred sculptural art, and high-speed milk liquid capture. The woman must not appear to be wearing a costume made of milk; instead, her body and the milk must exist as one continuous living form. Her upper body, face, shoulders, and arms emerge gracefully from luminous milk, while the surrounding milk expands into vast, feather-light wing-like structures, flowing veils, liquid ribbons, and translucent trailing forms. The final image should express purity, nourishment, softness, grace, and spiritual elevation.

Main composition:
Use a clean, centered hero composition with one single female figure floating or ascending in the center of the frame. Her pose should feel balletic, weightless, and prayer-like in elegance, with the body rising upward in a soft vertical-diagonal motion. Her arms should open or lift with graceful extension, and the milk should spread behind and around her like sacred flowing wings made of liquid light. The silhouette must feel iconic, noble, and sculptural, with abundant negative space and perfect visual calm.

Character direction:
The female figure must be a **mature Japanese/Korean East Asian beauty**, elegant, refined, naturally beautiful, with soft bone structure, graceful neck and shoulders, and a calm, luminous, intelligent expression. She must not look like a fantasy angel or influencer model. Her beauty should feel elevated, poetic, and premium, with a serene inner stillness. Her movement should suggest flight, surrender, softness, and ascension.

Pure-white milk fusion concept:
The milk must behave like a sacred, silky, living material suspended between liquid, silk, mist, and light. The body must transition seamlessly into the milk:
milk blooming from the back into wing-like arcs,
milk wrapping the waist and lower torso into flowing sculptural motion,
milk extending from the arms into soft drag trails,
thin translucent milk membranes glowing like liquid silk,
small suspended droplets orbiting the figure like sacred particles.
The transformation must be completely natural, with no hard seams, no dress-like costume effect, and no pasted splash appearance. The girl should feel as though she is formed from milk and light in the same instant.

Liquid realism:
The milk must be hyper-realistic, luxurious, and radiant:
dense creamy white milk,
silky viscosity,
soft luminous translucency in thin areas,
smooth tension on the liquid edges,
velvety highlights,
fine suspended droplets,
mist-like liquid trails,
clean high-speed splash precision.
The milk should feel pure, nourishing, soft, and luxurious, never watery, never chalky, never plastic, never harsh.

Pure-white sacred styling:
Push the whole image toward an almost transcendent all-white visual world. The milk, light, background, and atmosphere should remain within a refined spectrum of white, pearl, ivory-white, silver-white, and barely-there cool gray. The scene should feel flooded with soft purity and silence, as if the figure is floating inside a luminous sacred space. The emotional effect should be clean, healing, angelic in mood but not literal in styling.

Motion and trailing effect:
The liquid drag-tail effect must be elegant and weightless. The milk should sweep outward and upward like soft wings caught in slow-motion lift, with long flowing arcs, gentle liquid ribbons, and airy suspended forms. The figure should feel frozen at the most beautiful microsecond of ascent, with motion that is fluid, graceful, and spiritually calm rather than explosive.

Brand presence:
Include a subtle and premium **Nestle Milk** brand presence. A clean product pack, bottle, or restrained official packaging element may appear near the lower composition with elegant balance. The official logo should be tasteful, small, and integrated naturally into the layout. Brand communication should feel premium, pure, trustworthy, and elevated, never retail-heavy.

Lighting:
Use extremely soft **pure white beauty lighting** with divine clarity:
soft frontal glow on the face and torso,
gentle halo-like rim lighting around the milk wings,
clean highlights on liquid edges,
subtle transmitted glow through thinner milk membranes,
delicate tonal layering with almost no harsh shadow,
micro-contrast only where needed to reveal liquid texture.
The lighting must feel heavenly, creamy, pure, and luxurious, as if the figure is illuminated from both outside and within.

Background:
Keep the background ultra-minimal and nearly all-white, with a soft gradient of pearl white, ivory white, mist white, and faint silver-gray. The space should feel limitless, silent, and luminous. No props, no scenery, no decorative objects, no clutter. Only sacred premium negative space.

Typography:
Use only minimal and elegant typography. Include a refined **Nestle Milk** title, small official logo, and at most one short poetic line. Typography must feel airy, gentle, and sacred-luxury in tone, with perfect spacing and art-directed placement. No bold heavy fonts, no retail poster feeling, no cluttered copy.

Color palette:
pure milk white,
pearl white,
ivory white,
soft silver-white,
mist gray,
barely-there cool light tones.

Mood:
pure, sacred, serene, uplifting, weightless, luminous, maternal, elegant, nourishing, spiritual, premium, poetic, unforgettable.

Rendering:
hyper-realistic high-speed milk liquid photography, sacred luxury beauty-commercial style, ultra-detailed milk splash simulation, elegant flying female silhouette, pure white minimal background, refined beauty lighting, 8k, world-class advertising campaign quality.

Negative prompt:
watery milk, gray dirty whites, messy splash chaos, fake wings, costume-like milk dress, cartoon liquid, plastic skin, influencer face, hard body-liquid seam, cluttered background, oversized packaging, cheap supermarket poster, heavy typography, fantasy angel costume, low-detail droplets, harsh lighting

    🇨🇳 中文提示词

    为雀巢牛奶创作一张超高级的纯白圣洁奢华海报，主题聚焦于一位成熟的东亚女孩在飞行中上升，并与流动的牛奶无缝融合，仿佛她是从光、纯净和液体本身的柔软中诞生的。整个画面必须感受到神圣、宁静、失重且令人叹为观止的精致，如同高端美容摄影、圣洁雕塑艺术与高速牛奶液体抓拍的融合。女子绝不能看起来像是穿着一件由牛奶制成的服装；相反，她的身体和牛奶必须作为一个连续的生命形式存在。她的上半身、脸部、肩膀和双臂优雅地从发光的牛奶中显现，而周围的牛奶扩展成巨大的、轻如羽毛的翅膀状结构、流动的面纱、液体丝带和半透明的拖尾形状。最终画面应表达出纯净、滋养、柔软、优雅和精神升华。

主要构图：
使用干净、居中的主体构图，单个女性形象在画框中心漂浮或上升。她的姿势应具有芭蕾般的感官、失重感和祈祷般的优雅，身体以柔和的垂直对角线动作向上升起。她的双臂应优雅地伸展打开或举起，牛奶应在她身后和周围像由液体光芒构成的神圣流动翅膀一样展开。轮廓必须具有标志性、高贵感和雕塑感，拥有丰富的负空间和完美的视觉平静。

角色指导：
女性形象必须是一位成熟的日韩东亚美女，优雅、精致、天生丽质，拥有柔和的骨骼结构、优雅的脖子和肩膀，以及冷静、明亮、聪慧的表情。她绝不能看起来像幻想中的天使或网红模特。她的美感应是升华的、诗意的、高级的，带有宁静的内在寂静。她的动作应暗示飞行、臣服、柔软和上升。

纯白牛奶融合概念：
牛奶的表现必须像一种悬浮在液体、丝绸、薄雾和光线之间的神圣、丝滑、有生命的物质。身体必须无缝过渡到牛奶中：牛奶从背部绽放成翅膀般的弧线，牛奶包裹腰部和下半身形成流动的雕塑动作，牛奶从手臂延伸成柔软的拖痕，薄薄的半透明牛奶膜像液体丝绸一样发光，微小的悬浮液滴像神圣粒子一样绕着人物运行。这种转变必须完全自然，没有生硬的接缝，没有像衣服一样的服装效果，也没有粘贴上去的喷溅外观。女孩应该感觉像是那一瞬间由牛奶和光线形成的。

液体写实度：
牛奶必须是超写实的、奢华的、光彩夺目的：浓郁的奶油白牛奶，丝滑的黏度，薄处有柔和的发光半透明感，液体边缘的平滑张力，天鹅绒般的亮部，精细的悬浮液滴，雾状的液体痕迹，干净的高速喷溅精度。牛奶应该让人感到纯净、滋养、柔软和奢华，绝不能像水、粉末、塑料或刺眼。

纯白神圣风格：
将整个画面推向一个几乎超然的全白视觉世界。牛奶、光线、背景和氛围应保持在精致的白色、珍珠色、象牙白、银白和几乎感觉不到的冷灰色光谱内。场景应该充满柔和的纯净和寂静，仿佛人物漂浮在一个发光的神圣空间里。情感效果应该是干净、疗愈、带有天使般的情绪，但在造型上不是写实的。

动态与拖尾效果：
液体拖尾效果必须优雅且轻盈。牛奶应像捕捉到慢动作升力中的柔软翅膀一样向外向上扫过，带有长长的流动弧线、温柔的液体丝带和轻盈的悬浮形态。人物应感觉像是被冻结在上升过程中最美丽的微秒，动作流畅、优雅且精神平静，而不是爆发性的。

品牌呈现：
包含一个微妙且高端的雀巢牛奶品牌呈现。一个干净的产品包装、瓶子或克制的官方包装元素可以以优雅的平衡出现在构图下方。官方 Logo 应雅致、小巧，并自然地融入布局中。品牌传播应让人感到高级、纯粹、值得信赖和升华，绝不是沉重的零售感。

灯光：
使用极其柔和的、具有神圣清晰度的纯白美妆灯光：面部和躯干有柔和的正面光辉，牛奶翅膀周围有柔和的哈洛式轮廓光，液体边缘有干净的亮部，透过较薄的牛奶膜有微妙的透射光，细腻的色调层次几乎没有生硬的阴影，仅在需要揭示液体纹理的地方使用微对比。灯光必须感觉像天堂般、奶油色、纯净且奢华，仿佛人物从外部和内部同时被照亮。

背景：
背景保持极简且几乎全白，带有珍珠白、象牙白、雾白和微弱银灰色的柔和渐变。空间应感觉无边无际、寂静且明亮。没有道具，没有布景，没有装饰物，没有杂乱。只有神圣的高级负空间。

排版：
仅使用极简且优雅的排版。包括一个精致的雀巢牛奶标题、小巧的官方 Logo，以及最多一行简短的诗意文案。排版必须感觉轻盈、温柔，在色调上具有神圣奢侈感，间距完美且经过艺术指导。没有粗重的字体，没有零售海报的感觉，没有杂乱的文案。

配色方案：
纯牛奶白、珍珠白、象牙白、柔和银白、雾灰、几乎感觉不到的冷亮色调。

氛围：
纯净、神圣、宁静、令人振奋、失重、明亮、母性、优雅、滋养、精神、高级、诗意、难忘。

渲染：
超写实高速牛奶液体摄影，神圣奢华美容商业风格，超细节牛奶喷溅模拟，优雅飞行的女性轮廓，纯白极简背景，精致的美妆灯光，8k，世界级广告大片质量。

负面提示词：
水状牛奶，灰色肮脏的白色，混乱的喷溅混乱，假翅膀，像服装一样的牛奶裙，卡通液体，塑料皮肤，网红脸，生硬的身体-液体接缝，杂乱的背景，过大的包装，廉价的超市海报，沉重的排版，幻想天使服装，低细节液滴，刺眼的灯光
  • 高精度AI角色固定参考图表生成

    高精度AI角色固定参考图表生成

    高精度AI角色固定参考图表生成

    高精度AI角色固定参考图表生成


    🤖 ChatGPT

    🇨🇳 中文提示词

    You will create a “High-Precision Reference Chart” based on the attached character image that can be used for AI image generation or character consistency.  

The purpose is not to create a character profile, but to produce visual material so that the same character can be stably reproduced in the future.  
Name, personality, hobbies, story setting, and long profiles are not needed.  
Text in the image is limited to short labels for each panel and the minimum necessary characteristic memos.  

────────────────────  
Most Important Rule  
────────────────────  

Treat the character in the attached image as the sole standard, ensuring that they appear as the same character in all panels.  

Particularly, do not change the following:  

・Face contour  
・Eye shape, impression around the eyes, atmosphere of the gaze  
・Eyebrow shape and thickness  
・Nose bridge, nostrils, profile nose shape  
・Lip shape, thickness, tone  
・Cheeks, chin, face lines  
・Skin color, skin texture, natural unevenness  
・Hairstyle, bangs, side hair, back hair, hair color, hair flow  
・Sense of age  
・Body type, skeleton, length of limbs, full-body balance  
・Clothing impression  
・Overall atmosphere, sense of air, photo texture  

Prohibit turning the character into someone else, changing the face, over-beautifying, changing the body type, changing the hairstyle, or changing the hair color.  
In each angle, each expression, and each part, prioritize ensuring the consistency of the face, hair, and body type above all else.  

────────────────────  
Art Style / Texture  
────────────────────  

If the attached image is realistic, create it as a natural realistic photo.  
If the attached image is an illustration, animation, manga, 3D, or stylized, maintain its art style, lines, coloring, shading, texture, and degree of stylization.  

Do not arbitrarily turn it into realism.  
Do not arbitrarily turn it into animation.  
Do not arbitrarily turn it into 3D.  
Please keep the expression form of the original image unchanged.  

In the case of realism, please adopt a realistic texture like a smartphone photo or a natural studio reference photo.  
Skin should not be over-processed; retain natural pores, fine hairs, skin unevenness, and a sense of complexion.  
Prohibit doll-like skin, plastic skin, CG feel, and over-retouching.  

────────────────────  
Image Format  
────────────────────  

A horizontally long single reference chart.  
The aspect ratio is 16:9 to 16:10.  
The background is white, light gray, or pale off-white.  
Use moderate white space, organized with light-colored lines and boxes into a reference-use layout.  
Do not adopt a magazine cover, resume, poster, or social media post style.  

The overall composition should be like a neat and easy-to-read character production reference.  
Each panel is aligned and arranged, with short and easy-to-read labels.  
Do not overlay text on the face or important parts.  

────────────────────  
Clothing Rules  
────────────────────  

If the clothing in the original image is clearly visible, maintain the impression of that clothing.  
In all panels such as front, side, back, diagonal, expression differences, and part close-ups, maintain the same clothing, the same hairstyle, and the same body type.  

If the clothing in the original image is unclear, cropped, out of frame, or blurry, supplement with natural, low-key clothing that fits the character's atmosphere.  

Examples of supplemental clothing:  
・Simple white or light gray top  
・Natural casual bottoms  
・Plain-colored clothing that is not overly conspicuous  
・Neat clothing for reference use with easy-to-identify shapes  

Clothing to avoid:  
・Underwear  
・Swimwear  
・Excessive exposure  
・Clothing that looks sexy  
・Overly flashy clothing  
・Different clothing in each panel  
・Large accessories not present in the original image  

────────────────────  
Included Panel Composition  
────────────────────  

Please organize and arrange the following content within one chart.  

【1. Full Body View】  
・Front full body  
・Side full body  
・Back full body  
・Front-diagonal full body  
・Back-diagonal full body  

Purpose:  
To make the body type, sense of height, length of limbs, shoulder width, posture, clothing silhouette, and full-body balance clearly visible.  

Posture:  
Natural standing.  
Avoid extreme poses.  
Do not over-exaggerate like a model shoot.  
Make both hands and both feet naturally visible.  

【2. Face Close-up】  
・Front face close-up  
・Profile  
・45-degree angle  
・Slightly from above  
・Slightly from below  

Purpose:  
To confirm the face's three-dimensionality, eye/nose/mouth positions, contour, chin, profile, and the way hair falls.  

【3. Expression Differences】  
While maintaining the same face shape, add the following expressions.  

・Normal  
・Natural smile  
・Grin  
・Serious  
・Surprised  
・Shy  
・Thinking  
・Troubled expression  
・Slightly sad expression  
・Calm expression  

Even when changing expressions, please do not make it look like someone else.  
Please do not change the eye size, contour, nose, lips, or hairstyle.  

【4. Face Part Magnification】  
Please display the following parts magnified in small boxes respectively.  

・Eyes  
・Eyebrows  
・Nose  
・Mouth  
・Lips  
・Ears  
・Contour  
・Cheeks  
 skins texture  
・Bangs  
・Side hair  
・Back hair  
・Hair flow  
・Hands  

Purpose:  
To fix details that should be reproduced in future image generation.  
Please depict each part naturally and realistically, avoiding distortions or glitches.  

【5. Hair Reference】  
To make the hair structure clear, please include the following.  

・Bangs seen from the front  
・Side hair seen from the side  
・Hair length and overall feel seen from the back  
・Flow of hair tips  
・Sense of hair bundles  
・The way hair color appears  
・Texture when light hits it  

Please unify the shape of the bangs, parting, side hair, and back hair in all panels.  

【6. Other Angles】  
Please include the following angles.  

・Diagonal left  
・Diagonal right  
・Profile left  
・Profile right  
・Back of the head  
・From above  
・From below  

Purpose:  
To fix the face and hairstyle three-dimensionally.  
Even if the angle changes, please maintain the same character, the same hairstyle, and the same atmosphere.  

【7. Characteristic Memo Column】  
Create a short characteristic memo column on the left side or at the bottom of the chart.  

Included items:  
・Age impression  
・Face  
・Hair  
・Eyes  
・Eyebrows  
・Nose  
・Mouth  
・Skin  
・Body type  
・Sense of height  
・Clothing  
・Atmosphere  
・Characteristics to be fixed  

Please do not use long text, but short bullet points.  
Only write visually observable characteristics or characteristics strongly inferred from the original image.  
Do not include fictional profiles.  

【8. Color Palette】  
Display the colors needed for character reproduction in 5 to 7 small color samples.  

Included colors:  
・Hair color  
・Eye color  
・Skin color  
・Lip color  
・Clothing color  
・Shadow color  
・Overall tone color  

Please attach a short label to each color.  

【9. Consistency Memo】  
Briefly summarize the points to be observed when regenerating in the future.  

Example:  
・Fix face contour  
・Fix the impression around the eyes  
・Fix bangs and hair color  
・Fix skin texture  
・Fix body type and full-body balance  
・Unify clothing  
・Prioritize ensuring it looks like the same character above all else  

────────────────────  
Labels and Language  
────────────────────  

Please use Chinese uniformly for the labels in the image.  
Do not use English labels.  

Label examples:  
・Full body Front (全身 正面)  
・Full body Side (全身 侧面)  
・Full body Back (全身 后方)  
・Face Front (脸部 正面)  
・Profile (侧脸)  
・45-degree angle (斜45度)  
・Normal (通常)  
・Smile (笑容)  
・Serious (认真)  
・Surprised (惊讶)  
・Shy (害羞)  
・Thinking (思考中)  
・Troubled expression (困扰表情)  
・Eyes (眼睛)  
・Eyebrows (眉毛)  
・Nose (鼻子)  
・Mouth (嘴巴)  
・Ears (耳朵)  
・Skin (皮肤)  
・Bangs (前发)  
・Side hair (侧发)  
・Back hair (后发)  
・Hair flow (发丝流动)  
・Hands (手)  
・Characteristic memo (特征备忘)  
・Color samples (色样)  
・Fixed points (固定要点)  

Please display the text short, easy to read, and correctly.  
Garbled text, broken characters, text with unclear meaning, and text that is too small to read are prohibited.  

────────────────────  
Layout Guidelines  
────────────────────  

Left side:  
Character characteristic memo, color samples, fixed points.  

Center:  
Large arrangement of full-body views and face close-ups.  

Right side:  
Organized arrangement of expression differences, other angles, and face part magnifications.  

Bottom:  
Horizontal arrangement of detailed parts such as eyes, nose, mouth, hair, skin, and hands.  

Overall:  
Organize each part with boxes to make it easy to read as reference material.  
Moderate white space, high density but not cluttered.  
Unify the character size in each panel for easy comparison.  

────────────────────  
Shooting / Depiction Atmosphere  
────────────────────  

Natural light, or soft indoor light.  
Shadows should not be too strong, ensuring brightness where the face and hairstyle are easy to identify.  
The background is a solid color to avoid distracting the gaze from the character.  
Postures are natural, prioritizing true front, side, back, and diagonal views to facilitate use as character reference material.  

Flashy productions, cinematic lighting, strong blurring, heavy makeup, and fashion magazine-style poses are prohibited.  
Please make it like a neat and objective reference photo.  

────────────────────  
Excluded Content  
────────────────────  

Please do not include the following.  

・Name  
・Age numbers  
・Personality descriptions  
・Hobbies  
・Story settings  
・Profile text  
・Long descriptions  
・Unnecessary backgrounds  
・Unnecessary small items  
・Food  
・Random objects  
・Large accessories not in the original image  
・Logos  
・Watermarks  
・Signatures  
・Flashy decorations  
・Social media post-style designs  

────────────────────  

Prohibited Matters  
────────────────────  

Do not turn into someone else.  
Do not change the face in each panel.  
Do not change the hairstyle.  
Do not change the hair color.  
Do not change the body type.  
Do not let the clothing be inconsistent.  
Do not change the sense of age.  
Do not make them younger than the original image.  
Do not over-beautify.  
Do not over-simplify.  
Do not give a sexy impression.  
Do not use underwear or swimwear.  
Do not let hands or fingers glitch.  
Do not distort the profile.  
Do not unnaturally enlarge the eyes.  
Do not break the positions of face parts.  
Do not break the text.  
Do not break the layout.  
Do not let the background be cluttered.  

────────────────────  
Negative Prompt  
────────────────────  

Someone else, other characters, face change, eye change, nose change, mouth change, contour change, hairstyle change, hair color change, bangs change, body type change, age sense change, clothing inconsistency, over-beautification, over-youthfulness, looking small, excessive exposure, underwear, swimwear, sexy clothing, gravure style, fashion magazine style, flashy poses, extra small items, food, random objects, cluttered background, low resolution, garbled text, unreadable text, English labels, long profiles, logos, watermarks, signatures, hand glitches, finger glitches, extra fingers, missing fingers, distorted face, distorted profile, overlapping faces, body distortion, unnatural joints, art style change, arbitrary realism, arbitrary animation, arbitrary 3D, CGI feel, plastic skin, over-retouching, layout destruction  

────────────────────  
Final Output  
────────────────────  

Create a horizontally long single reference chart based on the character in the attached image, with character consistency as the top priority.  
Organize full body, face close-ups, profile, diagonal, expression differences, face parts, hair details, other angles, color samples, and fixed points using light lines and boxes.  
The composition should be convenient for use as reference material, neat and easy to read, facilitating the reproduction of character characteristics in the future.
  • Y2K复古街头风时尚设计稿

    Y2K复古街头风时尚设计稿

    Y2K复古街头风时尚设计稿

    Y2K复古街头风时尚设计稿


    🤖 ChatGPT

    🇺🇸 English Prompt

    [
  Reference Image
] Fashion blueprint sheet of a trendy young woman in Y2K grunge streetwear,
full-body front view with detailed outfit annotations and styling callout labels. Burgundy newsboy cap,
long wavy brown hair,
soft glam makeup,
layered necklaces,
burgundy corset tube top,
loosely worn striped necktie,
oversized black leather biker jacket draped off the shoulders,
light-wash distressed baggy jeans,
layered burgundy belt with silver chain charms,
black shoulder bag decorated with lace ribbons and a plush bunny keychain,
stacked rings,
black-and-white canvas sneakers,
confident fashion pose. Surrounded by fashion infographic elements,
fabric texture notes,
accessory breakdowns,
tailoring details,
pose analysis,
modern Y2K aesthetic,
editorial studio lighting,
ultra detailed,
professional fashion concept sheet,
8k.

    🇨🇳 中文提示词

    [
  参考图片
] 一位穿着 Y2K 颓废摇滚街头风的时髦年轻女性的时尚蓝图，
全身正面视图，
带有详细的服装注释和造型标注。酒红色报童帽，
长波浪棕色头发，
柔和的华丽妆容，
多层项链，
酒红色束身抹胸上衣，
松散佩戴的条纹领带，
披在肩膀上的超大黑色皮质机车夹克，
披在肩上的黑色皮质机车夹克，
浅色磨损宽松牛仔裤，
带有银链装饰的多层酒红色皮带，
装饰有蕾丝丝带和毛绒兔子钥匙扣的黑色单肩包，
叠戴戒指，
黑白帆布鞋，
自信的时尚姿势。周围环绕着时尚信息图表元素、面料纹理说明、配件分解图、剪裁细节、姿势分析、现代 Y2K 美学、社论演播室灯光、超详细、专业时尚概念图、8k。

  • OpenAI当年出价5亿没买下，如今它靠游戏录像训练机器人、估值23亿

    两年前，OpenAI 揣着 5 亿美元想买下 Medal——一个让玩家上传和分享游戏录像的平台，创始人 Pim de Witte 拒绝了。当时看有点莽。但到了 2026 年，de Witte 手里的 AI 公司 General Intuition 完成 3.2 亿美元 A 轮融资，估值冲到 23 亿美元，Khosla Ventures 领投，贝佐斯、前谷歌 CEO 施密特，以及 MIT 和 DeepMind 的研究人员都跟了进来。算上去年 10 月启动时的 1.34 亿，累计融资已经超过 4.54 亿。市场用真金白银证明了一件事：那堆游戏录像，比大多数人以为的值钱得多。

    用游戏数据训练机器人世界模型概念图
    用游戏数据训练机器人世界模型概念图（本文配图由 AI 生成）

    游戏画面里藏着什么

    General Intuition 的赌注听上去有点反直觉：用电子游戏里的人类操作记录，去教现实世界的机器人。它的训练数据来自 Medal 积累的数十亿小时游戏画面。但真正关键的不是画面本身，而是内嵌在每一帧里的”动作标记”——玩家在哪一刻按下了哪个键、做了什么决策。多数同行只靠视频去猜动作，而 de Witte 认为，拥有真实的人类操作数据，才是模型理解”自我”和”环境”分野、进而掌握因果关系的内核。

    de Witte 把公司的逻辑讲得很直白：”模型本身的泛化能力，就是产品。”他不想做机器人整机厂商，而是要成为别的机器人公司搭建产品时脚下的那块地基。

    8 分钟，让四足机器人学会走路

    传统机器人训练有两条老路：在真实环境里采数据，贵且慢；在仿真器里练，速度快却躲不开”仿真到现实的鸿沟”——你在 Unity 里训好的模型，放到真实地板和墙壁前常常像个路痴。General Intuition 把游戏数据当成连接两者的桥：先在海量游戏交互里学”物体怎么动、智能体怎么走、动作怎么改变状态”，再拿极少量的真实机器人数据做下游微调。他们演示过一台四足机器人，只经过 8 分钟真实世界微调，仅靠前置摄像头，就在有行人和动态障碍的办公室里零样本导航成功。

    • 领投方 Khosla 看中的是独有数据壁垒，以及”AI 直觉涌现”的愿景
    • de Witte 划下红线：不做任何用于伤害人类的自主武器
    • 目标是对标 NLP 的 GPT-3 之后——通用基座加轻量适配，取代每个任务单独训练

    有意思的是，当中国的人形机器人公司在拼命扩产量、把模仿学习的工业栈跑起来时，General Intuition 想走的是另一条路：卖”世界模型”这个软件层，而不是卖铁。这条路线能不能扛住工厂车间和客厅里那些充满摩擦、接触密集的真实物理，还得等它在同行评审的环境里被反复验证。但至少现在，顶级硅谷资本愿意为”通用世界模型”这个赌注下重注。

  • Grok 4.5发布：马斯克不拼参数了，改拼性价比

    淡出主流模型竞技场两个月之后，马斯克的 Grok 又坐回了牌桌。北京时间 7 月 9 日，SpaceXAI 正式发布旗舰模型 Grok 4.5。和过去每家都在喊”我们最强”的画风不同，这次马斯克显得很务实：他在 X 上连发几十条动态，反复强调一句话——能力和 Claude 的 Opus 4.7 差不多，但快得多，也便宜得多。

    Grok 4.5 概念配图
    Grok 4.5 概念配图（本文配图由 AI 生成）

    一张”够用就好”的牌

    Grok 4.5 是个 1.5T 参数的 MoE 模型，上下文拉到 50 万 token，主打编程、智能体和知识工作，还能看图输入。在考察编程的 SWE Bench Pro 上它拿到 64.7%，刚好压过 Opus 4.7 的 64.3%；Terminal Bench 2.1 冲到 83.3%，DeepSWE 1.0 的 62% 也超过了 Opus 4.8。推理速度能跑到 80 TPS，已经挤进第一梯队。

    有网友引用测试说，Grok 4.5 只花了大约十七分之一的成本，就做到了接近 Claude Fable 的性能。马斯克没有顺杆爬，反而老老实实承认差距：”公平地说，Fable 确实比 Grok 4.5 好得多，但大多数任务并不需要 Fable 级别的能力。”

    真正让对手冒汗的是价格

    但 SpaceXAI 真正想卖的不是分数，而是账本。它的 API 定价是每百万输入 token 2 美元、输出 6 美元。作为对照，Opus 4.7 的输入要 5 美元、输出 25 美元。完成同一个 SWE 任务，Grok 4.5 吐出的 token 量大约只有 Opus 4.8 Max 的四分之一。

    把模型调用价格推向台前，是这一轮竞争里最现实的变化。当企业把 Agent 跑在生产环境、每天调用成千上万次，token 账单就变成了真金白银。谁能在不掉链子的前提下把单位成本压下来，谁就握住了开发者的切换按钮。

    • 和 Cursor 联合训练，面向真实编码场景打磨
    • 下周推送百万 token 上下文版本，月底还有 2T 参数版本预告
    • 已在 SpaceXAI 控制台、Grok Build 和 Cursor 三端上线

    说到底，马斯克这次没再玩”参数军备竞赛”那一套。他赌的是另一件事：在绝大多数真实工作里，用户要的不是天花板级别的智能，而是一份”能力够用、速度快、账单轻”的交付。特斯拉和 SpaceX 的硬核工程师觉得 Grok 4.5 真好用，这比任何榜单都更能说明问题。

  • Claude Cookbooks：Anthropic 官方出品的 Claude 使用范例集，47.8K stars 让 AI 开发少走弯路

    Claude Cookbooks：Anthropic 官方出品的 Claude 使用范例集，47.8K stars 让 AI 开发少走弯路

    Claude Cookbooks

    Claude Cookbooks 是 Anthropic 官方维护的一组 Claude 使用范例/配方集（Jupyter Notebooks），用可直接复制、可运行的代码片段教会开发者如何用好 Claude API。它覆盖了文本分类、RAG、摘要、工具调用、多模态视觉、子代理、自动评估、JSON 模式、提示缓存等核心场景。截至 2026 年 7 月，仓库已收获 47.8K+ Stars、5.6K+ Forks，是学习和参考 Claude 官方最佳实践的必读项目。

    一、项目简介

    Claude Cookbooks 的定位非常清晰：它不是框架，也不是 SDK，而是一本「官方菜谱」。每个 Notebook 都围绕一个真实开发问题展开，例如「怎么用 Claude 做 RAG」「怎么让 Claude 调用外部工具」「怎么处理 PDF 与图片」「怎么自动评估生成结果」。你可以直接复制代码到自己的项目里，也可以把它当成学习 Claude API 的教程。

    二、安装要求和过程

    环境要求：

    • Python 3.10+（示例代码以 Python 为主）；
    • 一个 Claude API Key（可在 Anthropic 官网 免费注册）；
    • 仓库使用 uv 管理依赖，建议安装 uv（pip install uv 或 curl -LsSf https://astral.sh/uv/install.sh | sh）；
    • 运行 Jupyter Notebook 需要浏览器或 VS Code Jupyter 插件。

    快速安装：

    # 1. 克隆仓库
git clone https://github.com/anthropics/claude-cookbooks.git
cd claude-cookbooks

# 2. 用 uv 安装依赖
uv sync

# 3. 设置环境变量
export ANTHROPIC_API_KEY="sk-ant-api03-..."

# 4. 启动 Jupyter 浏览示例
jupyter notebook

    如果你不想安装 uv，也可以直接用 pip：pip install anthropic jupyter，然后逐本运行 Notebook。

    三、核心功能

    • 覆盖 Claude 全能力的实战 Notebook：从基础的文本分类、摘要、RAG，到高级的工具调用（Tool Use）、JSON 模式、子代理（Sub-agents）、自动评估（Evals）和提示缓存，几乎囊括了 Claude 的所有核心能力。
    • 官方出品、持续更新：Anthropic 工程团队直接维护，Notebook 会跟随 Claude 3.5 / 3.7 / 4 系列模型和新 API 特性同步更新，避免你学到过时的写法。
    • 可复制、可扩展的代码片段：每个配方都提供最小可运行示例，代码结构清晰，方便你替换成自己的数据、提示词或业务逻辑。
    • 多模态与文档理解：包含视觉理解、图表解析、PDF 内容提取、表单识别等示例，适合需要处理非结构化数据的项目。
    • 评估与可观测性：提供自动评估 Cookbook 和 Agent 搜索基准复现（DeepSearchQA / BrowseComp），帮助你建立「改提示 → 跑评估 → 看指标」的迭代闭环。

    四、典型使用场景

    1. 快速上手 Claude API：如果你是第一次用 Claude API，直接打开 getting_started 或基础能力 Notebook，五分钟就能把第一条请求跑起来。
    2. 构建 RAG 与文档问答应用：参考 RAG 配方，把自有文档切分、嵌入、检索后交给 Claude 生成答案；也可以看多模态示例，直接让 Claude 读取 PDF 或图片并回答。
    3. 开发工具型 Agent：学习如何让 Claude 调用计算器、数据库、搜索引擎等外部工具，并进一步组合成多 Agent 协作系统（Coordinator + Sub-agents）。

    五、推荐理由

    Claude Cookbooks 给我的最大价值是「少走弯路」。网上关于 Claude 的教程很多，但官方示例在 Prompt 写法、参数选择、错误处理和成本控制上往往最贴近生产实践。特别是工具调用、JSON 模式和提示缓存这几个容易踩坑的场景，官方代码把最佳实践写得非常清楚。而且它完全开源（MIT），Notebook 形式读起来没有框架源码那么重，适合一边看一边改。

    六、下载地址