Back to Details

anti-render

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Anti-Render 智能图像转换

Anti-Render Intelligent Image Transformation

任务目标

Task Objectives

  • 本 Skill 用于:智能识别用户上传的图像所属领域，并根据"理想承诺 vs 残酷现实"的核心理念，生成相应风格的图像
  • 能力包含:领域识别（建筑/人物/其他）、图像状态识别、智能意图判断、三种模式图像生成（理想渲染/真实面貌/对比图）
  • 触发条件:用户上传任何领域的图像并要求"anti-render处理"、"做对比"、"理想化处理"或"现实化处理"
  • This Skill is used to: intelligently identify the field of the image uploaded by the user, and generate images in corresponding styles based on the core concept of "Ideal Promise vs. Harsh Reality"
  • Capabilities include: domain recognition (architecture/person/others), image status recognition, intelligent intention judgment, three-mode image generation (ideal rendering/true appearance/comparison image)
  • Trigger conditions: The user uploads an image from any field and requests "anti-render processing", "make a comparison", "idealization processing" or "realistic processing"

前置准备

Preparations

  • 无需特殊依赖，利用智能体已有的图像识别与生成能力
  • 需要用户提供待处理的图像
  • 根据识别的领域，动态读取对应的风格特征库，从 JSON 中提取灵感与风格要点（而不是逐句追随或照搬文案）：
    • 建筑领域：anti-render.architecture.json
    • 人物领域：anti-render.person.json
    • 其他领域：使用通用核心原则
  • No special dependencies required, utilizing the existing image recognition and generation capabilities of the agent
  • The user needs to provide the image to be processed
  • According to the recognized domain, dynamically read the corresponding style feature library, and extract inspiration and style key points from JSON (instead of following or copying the text verbatim):
    • Architecture domain: anti-render.architecture.json
    • Person domain: anti-render.person.json
    • Other domains: Use general core principles

工作步骤（意图识别）

Work Steps (Intention Recognition)

用户上传图片后，识别其当前状态和期望转换方向：
破烂/不理想状态：
  • 建筑领域：墙体开裂、涂料剥落、金属锈蚀
  • 人物领域：皮肤问题、光线恶劣、妆容糟糕
  • 其他领域：明显的质量问题、使用痕迹、维护不良
  • 输出目标：返回一张理想化的渲染图，达到该领域的宣传级别
普通/正常状态：
  • 建筑领域：建筑外观正常、无明显破损、日常使用痕迹
  • 人物领域：正常的相机直出、无严重皮肤问题
  • 其他领域：无严重质量问题、正常的使用状态
  • 输出目标：返回一张相应的现实主义"真实面貌"图片
对比模式：
  • 用户明确使用"对比"关键词
  • 输出目标：返回一张对比图，中间以显眼但不抢视觉重心的分隔线分割
  • 排列方式：根据原图宽高比确定是上下排列还是左右排列，以便返回一张新的宽高比更均衡的图片
    • 横向图片（宽 > 高）：采用左右排列，分割线为垂直线
    • 纵向图片（高 > 宽）：采用上下排列，分割线为水平线
  • 默认对比：极端理想 vs 普通现实（除非用户明确要求"破败对比"）
中间派：
  • 图片状态介于破烂和普通之间，无法明确判断用户意图
  • 行动：需要主动询问用户期望的转换方向
After the user uploads an image, identify its current status and desired transformation direction:
Damaged/Unideal Status:
  • Architecture domain: cracked walls, peeling paint, rusted metal
  • Person domain: skin problems, poor lighting, bad makeup
  • Other domains: obvious quality issues, usage traces, poor maintenance
  • Output Target: Return an idealized rendering that reaches the promotional level of the field
Normal/Ordinary Status:
  • Architecture domain: normal building appearance, no obvious damage, daily usage traces
  • Person domain: normal camera output, no serious skin problems
  • Other domains: no serious quality issues, normal usage status
  • Output Target: Return a corresponding realistic "true appearance" image
Comparison Mode:
  • The user explicitly uses the keyword "comparison"
  • Output Target: Return a comparison image, split by a prominent but non-distracting dividing line in the middle
  • Arrangement method: Determine whether to arrange vertically or horizontally according to the aspect ratio of the original image, so as to return a new image with a more balanced aspect ratio
    • Horizontal image (width > height): Arrange left and right, with a vertical dividing line
    • Vertical image (height > width): Arrange top and bottom, with a horizontal dividing line
  • Default Comparison: Extreme Ideal vs. Ordinary Reality (unless the user explicitly requests "dilapidation comparison")
Neutral Case:
  • The image status is between damaged and ordinary, and the user's intention cannot be clearly judged
  • Action: Need to actively ask the user for the desired transformation direction

核心理念

Core Concept

Anti-Render 的核心理念是**"理想承诺 vs 残酷现实"的视觉对比叙事**。通过并置(juxtaposition)手法，揭示任何领域中"承诺与交付之间巨大落差"的普遍困境：
  • 广告图 vs 实物图:电商产品的精修宣传图 vs 买家秀真实照片
  • 概念设计 vs 量产版本:汽车/手机发布会的概念图 vs 最终上市的妥协版
  • 包装设计 vs 货架实物:设计稿中的精美包装 vs 超市货架上的实际效果
  • 菜单照片 vs 实际上桌:餐厅菜单上的精致摆盘 vs 服务员端上来的真实样子
  • 食品广告 vs 开箱实物:汉堡广告的诱人特写 vs 拆开包装纸的塌陷汉堡
  • 烘焙教程 vs 翻车现场:美食博主的完美成品 vs 普通人的第一次尝试
  • 官方宣传照 vs 游客实拍:旅游局的空旷美景 vs 人山人海的真实场景
  • 酒店官网图 vs 入住实拍:五星级酒店的样板间 vs 实际分配的房间
  • 景区效果图 vs 建成实景:规划中的主题公园 vs 开业后的廉价感
  • 城市规划图 vs 实际建成:政府展示的未来城市愿景 vs 十年后的实际面貌
  • 公共空间设计 vs 使用现状:设计师笔下的活力广场 vs 无人问津的空旷水泥地
  • 交通规划图 vs 拥堵现实:理想化的道路流量模拟 vs 早高峰的停车场
  • 游戏宣传片 vs 实际画面:E3 展会的精修演示 vs 玩家电脑上的实际运行效果
  • 电影海报 vs 剧照截图:精修的角色海报 vs 电影中的实际镜头
  • 游戏 UI 概念图 vs 最终界面:设计稿中的炫酷界面 vs 上线后的简化版本
  • 健身广告 vs 真实训练:健身房宣传的完美身材与环境 vs 汗流浃背的真实训练场景
  • 减肥前后对比 vs 真实过程:社交媒体的完美转变 vs 中间反复的挣扎过程
  • 瑜伽教程图 vs 初学者实拍:导师的完美体式 vs 普通人的僵硬模仿
  • 联合办公宣传 vs 实际使用:Wework 式的理想办公空间 vs 拥挤嘈杂的真实环境
  • 家居样板间 vs 入住后:宜家展厅的完美收纳 vs 生活三个月后的杂乱
  • 智能家居演示 vs 日常使用:科技展会的流畅操作 vs 家中频繁断连的现实
  • 音乐节官宣 vs 现场实况:主办方的炫酷舞美图 vs 泥泞拥挤的观众视角
  • 婚礼效果图 vs 仪式现场:婚庆公司的梦幻布置 vs 预算有限的实际效果
  • 展会效果图 vs 撤展后:展览开幕的精致布展 vs 最后一天的破败景象
  • 课程宣传图 vs 实际课堂:在线教育的精美课件 vs 卡顿的直播画面
  • 校园宣传片 vs 学生日常:招生简章的理想校园生活 vs 图书馆抢座的真实
  • 培训机构承诺 vs 就业现实:宣传中的高薪就业 vs 实际的求职困境
  • 环保宣传 vs 污染现状:政府报告中的绿水青山 vs 实际的工业污染
  • 季节宣传照 vs 气候现实:旅游手册的四季如画 vs 全球变暖的异常天气
  • 野生动物纪录片 vs 栖息地现状:BBC 级别的壮美自然 vs 人类活动破坏后的荒凉
  • Instagram 生活 vs 真实日常:精心策划的生活方式照片 vs 镜头外的混乱房间
  • 网红打卡点 vs 拍摄角度外:完美构图的网红墙 vs 转身就是垃圾堆的真相
  • 约会软件照片 vs 见面真人:精修的个人资料照 vs 第一次约会的真实样貌
  • 发布会渲染图 vs 量产机:苹果发布会的完美工业设计 vs 实际的天线带/刘海
  • VR 体验宣传 vs 实际佩戴:广告中的沉浸式未来 vs 戴上头显后的眩晕与笨重
  • 智能汽车演示 vs 路测实况:自动驾驶的理想演示 vs 频繁接管的测试现实
The core concept of Anti-Render is visual contrast narrative of "Ideal Promise vs. Harsh Reality". Through the technique of juxtaposition, it reveals the common dilemma of "huge gap between promise and delivery" in any field:
  • Advertising Image vs. Actual Product: Polished promotional images of e-commerce products vs. real photos from buyers
  • Concept Design vs. Mass-Produced Version: Concept images of cars/phones at press conferences vs. the compromised final version on the market
  • Packaging Design vs. Shelf Product: Exquisite packaging in design drafts vs. actual effects on supermarket shelves
  • Menu Photo vs. Served Dish: Exquisite plating on restaurant menus vs. the real dish served by waiters
  • Food Advertisement vs. Unboxed Product: Tempting close-up of a hamburger in ads vs. the collapsed hamburger after unwrapping
  • Baking Tutorial vs. Failed Attempt: Perfect finished product from a food blogger vs. a beginner's first try
  • Official Promotional Photo vs. Tourist's Shot: Empty beautiful scenery from tourism bureaus vs. the real crowded scene
  • Hotel Official Website Image vs. In-Room Shot: Showroom of a five-star hotel vs. the actual room assigned
  • Scenic Area Rendering vs. Completed Scene: Planned theme park vs. the cheap feeling after opening
  • Urban Planning Map vs. Actual Construction: Future city vision displayed by the government vs. the actual appearance after ten years
  • Public Space Design vs. Current Usage: Vibrant square in the designer's vision vs. the empty cement space that no one cares about
  • Traffic Planning Map vs. Congested Reality: Idealized road flow simulation vs. the parking lot-like morning rush hour
  • Game Promotional Video vs. Actual Gameplay: Polished demo at E3 vs. actual running effect on players' computers
  • Movie Poster vs. Screenshot: Polished character poster vs. actual shot in the movie
  • Game UI Concept vs. Final Interface: Cool interface in design drafts vs. the simplified version after launch
  • Fitness Advertisement vs. Real Training: Perfect figure and environment promoted by gyms vs. the real sweaty training scene
  • Weight Loss Before-After vs. Real Process: Perfect transformation on social media vs. the repeated struggles in the middle
  • Yoga Tutorial Image vs. Beginner's Shot: Perfect pose from the instructor vs. the stiff imitation of an ordinary person
  • Co-working Space Promotion vs. Actual Usage: Ideal WeWork-style office space vs. the crowded and noisy real environment
  • Home Showroom vs. Post-Move-In: Perfect storage in IKEA showrooms vs. the mess after three months of living
  • Smart Home Demo vs. Daily Usage: Smooth operation at tech exhibitions vs. the frequent disconnection at home
  • Music Festival Official Announcement vs. Live Scene: Cool stage design from the organizer vs. the muddy and crowded audience view
  • Wedding Rendering vs. Ceremony Scene: Dreamy arrangement from wedding companies vs. the actual effect with limited budget
  • Exhibition Rendering vs. Post-Exhibition Scene: Exquisite booth setup at the exhibition opening vs. the dilapidated scene on the last day
  • Course Promotional Image vs. Actual Classroom: Exquisite courseware from online education vs. the lagging live stream
  • Campus Promotional Video vs. Student's Daily Life: Ideal campus life in admission brochures vs. the real scene of fighting for seats in the library
  • Training Institution's Promise vs. Employment Reality: High-paying employment promoted vs. the actual job-hunting dilemma
  • Environmental Protection Promotion vs. Pollution Status: Green mountains and clear waters in government reports vs. actual industrial pollution
  • Seasonal Promotional Photo vs. Climate Reality: Picturesque four seasons in travel brochures vs. abnormal weather due to global warming
  • Wildlife Documentary vs. Habitat Status: BBC-level magnificent nature vs. the desolation after human activity destruction
  • Instagram Life vs. Real Daily: Carefully curated lifestyle photos vs. the messy room outside the camera
  • Internet Celebrity Check-in Spot vs. Beyond the Shooting Angle: Perfectly composed internet celebrity wall vs. the truth of a garbage dump just around the corner
  • Dating App Photo vs. In-Person Meeting: Polished profile photos vs. the real appearance at the first date
  • Press Conference Rendering vs. Mass-Produced Device: Perfect industrial design at Apple's press conference vs. the actual antenna band/notch
  • VR Experience Promotion vs. Actual Wearing: Immersive future in ads vs. the dizziness and bulkiness after putting on the headset
  • Smart Car Demo vs. Road Test Reality: Ideal autonomous driving demo vs. the frequent takeover during testing

领域识别与资源映射

Domain Recognition and Resource Mapping

1. 建筑领域

1. Architecture Domain

识别特征：建筑外观、城市景观、室内空间、建筑效果图 资源文件：anti-render.architecture.json 核心表达：建筑师理想化渲染图到残酷现实的视觉坠落
Recognition Features: Building appearance, urban landscape, interior space, architectural rendering Resource File: anti-render.architecture.json Core Expression: Visual fall from the architect's idealized rendering to the harsh reality

2. 人物领域

2. Person Domain

识别特征：人像写真、Cosplay摄影、活动拍摄、自拍 资源文件：anti-render.person.json 核心表达：后期精修的完美人像到原片真实的视觉落差
Recognition Features: Portrait photography, Cosplay photography, event shooting, selfie Resource File: anti-render.person.json Core Expression: Visual gap between the perfectly retouched portrait and the original real photo

3. 其他领域

3. Other Domains

识别特征：产品、食物、旅游、游戏、健身、办公、活动、教育、自然、社交媒体、科技等 资源文件：使用通用核心原则 应用原则：灵活应用到具体场景，基于5个核心对比场景进行创作
Recognition Features: Products, food, travel, games, fitness, office, events, education, nature, social media, technology, etc. Resource File: Use general core principles Application Principle: Flexibly apply to specific scenarios, create based on 5 core comparison scenarios

核心要求

Core Requirements

严格保留原图结构、元素与构图：
  • 所有模式下必须严格保留原图的完整结构与所有元素
  • 理想渲染/真实面貌模式：输出图保持原始尺寸（W×H）
  • 对比图模式：左右或上下两部分各自保持原图尺寸与构图（每一部分为 W×H），仅改变材质、光照、色彩和氛围；最终画布为 2W×H 或 W×2H
  • 分割线的存在不应破坏或遮挡原图的核心结构
  • 不得改变原图的构图、透视关系、关键形态、环境布局等
  • 禁止增加标题、边框等任何新元素：输出图像中不得添加任何原图没有的装饰性元素，包括但不限于标题、文字标签、边框、装饰线条、水印等
对比图排列规则：
  • 根据原图宽高比智能决定排列方式
  • 通过拼接得到一张新的宽高比更均衡的图片
  • 分割线位置：正中，2-5像素宽，纯白色或极浅灰色
  • 边缘处理：锐利清晰，不做羽化，强调两个世界的断裂感
  • 分割线本身不属于新增元素，而是构图的一部分
对比图内容规则：
  • 默认对比：极端理想 vs 普通现实
    • 左侧/上方：极端理想化渲染（该领域的完美呈现）
    • 右侧/下方：普通现实主义面貌（真实的日常状态，但不破败）
  • 特殊对比：如果用户明确要求"破败对比"或"极端对比"，则使用理想 vs 破败的对比
    • 右侧/下方：残酷现实（严重的质量问题、衰败、破败状态）
领域专属资源引用规则：
  • 建筑领域：必须读取 anti-render.architecture.json，提取色彩、光照、材质、氛围、镜头感等风格要点，用于指导生成；不要照搬 JSON 的原句或把其中的文字当作要生成的画面元素
  • 人物领域：必须读取 anti-render.person.json，提取肤质处理、布光方式、调色倾向、背景质感等风格要点，用于指导生成；不要照搬 JSON 的原句或把其中的文字当作要生成的画面元素
  • 其他领域：使用通用核心原则，基于5个核心对比场景进行创作
主动询问机制：
  • 如果用户给的图片过于中间派，无法识别用户意图时，需要主动询问
  • 询问内容明确列出三种转换方向供用户选择
Strictly retain the original image structure, elements and composition:
  • The complete structure and all elements of the original image must be strictly retained in all modes
  • Ideal Rendering/True Appearance mode: The output image maintains the original size (W×H)
  • Comparison Image mode: The left/right or top/bottom parts each maintain the original size and composition of the image (each part is W×H), only changing materials, lighting, colors and atmosphere; the final canvas is 2W×H or W×2H
  • The dividing line should not damage or block the core structure of the original image
  • Do not change the composition, perspective, key forms, environmental layout, etc. of the original image
  • Prohibit adding any new elements such as titles, borders: No decorative elements that are not in the original image should be added to the output image, including but not limited to titles, text labels, borders, decorative lines, watermarks, etc.
Comparison Image Arrangement Rules:
  • Intelligently determine the arrangement method according to the aspect ratio of the original image
  • Obtain a new image with a more balanced aspect ratio through splicing
  • Dividing line position: centered, 2-5 pixels wide, pure white or very light gray
  • Edge processing: sharp and clear, no feathering, emphasizing the sense of fracture between the two worlds
  • The dividing line itself is not a new element, but part of the composition
Comparison Image Content Rules:
  • Default Comparison: Extreme Ideal vs. Ordinary Reality
    • Left/Top: Extreme idealized rendering (perfect presentation of the field)
    • Right/Bottom: Ordinary realistic appearance (real daily state, but not dilapidated)
  • Special Comparison: If the user explicitly requests "dilapidation comparison" or "extreme comparison", use Ideal vs. Dilapidated comparison
    • Right/Bottom: Harsh reality (serious quality problems, decay, dilapidated state)
Domain-Specific Resource Reference Rules:
  • Architecture Domain: Must read anti-render.architecture.json, extract key points of color, lighting, material, atmosphere, lens feel, etc. to guide generation; do not copy the original sentences of JSON or treat the text as elements to be generated in the image
  • Person Domain: Must read anti-render.person.json, extract key points of skin texture processing, lighting method, color grading tendency, background texture, etc. to guide generation; do not copy the original sentences of JSON or treat the text as elements to be generated in the image
  • Other Domains: Use general core principles, create based on 5 core comparison scenarios
Active Inquiry Mechanism:
  • If the user's image is too neutral and the user's intention cannot be identified, need to actively ask the user
  • The inquiry content clearly lists three transformation directions for the user to choose