Image Prompts for App Web Design

2D flat vector fashion illustration, {argument name="subject" default="unveiled Middle Eastern woman"}, full body, wearing {argument name="trousers" default="wide-leg trousers in mustard yellow with geometric print"}, {argument name="blouse" default="embroidered cream blouse with floral motifs"}, minimal shadows, clean bold outlines, soft beige background, editorial fashion pose, 3:4

请将我上传的每一张旅行照片分别制作成一张独立的高级编辑海报，不多图拼接，每张照片单独输出。整体采用3:4竖版上下对比构图，上下两个区域约为1:1，中间以笔直、清楚、干净的水平边界分隔。不要羽化、渗透或叠层，不要把原图缩成角标、圆角卡片或悬浮缩略图。上半部分必须保留高保真的原始照片。准确保持主体身份与数量、建筑轮廓、屋顶与门窗比例、人物姿态、车辆、桥梁、船只、道路方向、前后遮挡、观察角度、真实材质、自然光影和原有色彩氛围。仅进行轻微、克制的旅行杂志、艺术出版物或摄影展级调色，并加入极薄胶片颗粒。不得把照片画成插画，不得移动、替换、增删或扭曲主体，不得改变建筑和街道透视。下半部分在温暖米白或浅象牙色纸面上，将同一场景重新组织为一幅具有大量留白的“人工像素消隐景观”。这不是照片马赛克滤镜，也不是把整张照片降低分辨率。先从照片中选择一个最有识别度的视觉锚点，例如一座桥、一段彩色立面、一座穹顶、一辆摩托车、一辆自行车、一艘船、一个门楼或一组明确屋顶，再只保留能够说明地点和空间关系的最少元素。采用缩小提炼的二次构图：主体通常只占转绘区宽度约30%–50%、高度约30%–50%，集中在中下部、偏左或偏右，并依据原照片的方向自然安排。四周保留至少约50%–65%的完整米白纸面。不要把原始街景完整铺满下半区域。背景只保留两三层稀疏像素暗示。删除重复窗格、瓦片细纹、招牌小字、游客杂影、密集管线、枝叶与摄影噪声，但不得删掉决定主体身份的屋檐、桥拱、穹顶、车辆轮廓、人物姿态或道路方向。使用边缘清楚、正方或近似正方的离散像素块重新搭建形体：- 以中等尺寸像素块建立主体的大轮廓、体量、明暗和主要色面。- 在屋檐、塔尖、穹顶边缘、门窗比例、桥拱、人物姿态、车轮、船头等识别关键处使用更小、更密的像素块。- 在建筑外缘、植被、路面、水面、倒影和远景处逐渐减少像素密度，使方块断开、错位、散落并消失到纸面。- 主体内部相对完整，外围形成清楚的像素消隐带。- 消隐必须来自独立方块数量与密度的下降，不得使用柔焦、羽化、烟雾、透明渐变或像素拉伸。- 少量离散方块可以沿屋檐、道路透视、水流、地形或运动方向离开主体，但不能成为随机数字故障。配色完全从当前照片提取并压缩为约6–12个主色阶。保留主体最具识别度的1–3个颜色，再用相邻深浅色建立体积。环境色降低饱和度，并向自然灰、石灰色、灰蓝、灰绿、木褐或照片自身的建筑色统一。可以将照片中真实存在的一处小面积高饱和色作为视觉跳点，例如尾灯、门窗或服装颜色。米白纸面必须是转绘区的最大色…

Elegant abstract illustration of {argument name="reference image" default="[الصورة المرفقة]"} constructed from a single flowing ribbon-like form, continuous curves suggesting movement and volume, subtle dimensional shading, restrained {argument name="color palette" default="[أختيار اللون هنا]"} palette, graceful negative space, premium brand-art aesthetic, sophisticated simplicity, centered composition.” {argument name="aspect ratio" default="9:16"}

Goal: Create a realistic vertical smartphone screenshot of a Telegram chat showing a bot successfully generating an image. Canvas: Tall mobile screenshot, 9:16 aspect ratio, about 576×1024 px, dark mode. Use a black Telegram doodle wallpaper in the background with faint line-art icons. The screen should look like an iPhone with the status bar at the top. Layout: At the very top, show the iOS status bar with the time “6:50” on the left, a blue rounded “TELEGRAM” callout centered, and cellular/Wi‑Fi/battery icons on the right. Below it, show the Telegram chat header with a circular back button on the left, a rounded dark title pill in the center reading “{argument name="bot name" default="Arcturus Alpha"}” with small “bot” underneath, and a small circular avatar on the right showing a glowing bull emblem. Visible chat elements: Include exactly 3 message areas: 1 large upper outgoing-style dark purple message bubble, 1 image message card with an image plus caption, and 1 partially visible lower text message behind the input bar. Text content: The upper message bubble should read exactly: “four, coming up — same minimal test as last time so we're changing exactly one variable:]] Remark: Attempt four — same test, now with the missing await in place...” with timestamp “6:48 PM” at the lower right of the bubble. The image caption should read exactly: “{argument name="caption text" default="gpt-image-2 smoke test — the constellation ox, back in business."}” with timestamp “6:49 PM” at the lower right. The lower partially obscured message should begin exactly: “[[Private: 🤍 Arcturus> 🎉 It worked! That missing await was the culprit — image generated, file” and continue faintly behind the input area, mostly hidden. Generated image inside chat: In the image card, create a square cosmic artwork of a {argument name="constellation animal" default="ox"} formed from glowing golden constellation lines and star nodes. The animal should be a side-view bull/ox with horns, head lowered slightly, one front leg raised, and a body outlined by interconnected stars. Use a deep blue-black galaxy background with scattered stars and a bright golden starburst near the animal’s shoulder/chest. Bottom controls: At the bottom, show the Telegram message composer overlay. Include exactly 5 visible controls: a blue rounded “Menu” button with a hamburger icon, a circular paperclip attachment button, a rounded text field placeholder “Message”, a small circular icon inside/near the field, and a circular microphone button on the right. Also include a small floating down-arrow scroll button near the lower right above the composer. Visual style: Photorealistic app screenshot, crisp white UI text, translucent dark bubbles, rounded corners, natural Telegram dark theme spacing. Keep the screenshot slightly cropped at the bottom as if captured mid-chat. No watermarks, no extra messages, no additional images.

🙃🙂😅 请将我上传的每一张照片分别制作成一张独立的高级设计海报，不多图拼接，每张照片单独输出。整体采用{argument name="构图比例" default="3:4竖版构图"}，上下两个区域高度严格1:1，各占画面50%。 上半部分保留原始照片，保持主体身份、结构、姿态、真实质感、自然光影和原有色彩氛围，仅进行轻微高级调色，使其具有艺术杂志、独立出版物与展览图像质感。为适配画幅，可自然扩展环境背景，但不得拉伸、扭曲或改变主体。 下半部分提取照片中最具识别性的**主体、轮廓、结构、姿态与叙事关系**，先理解原画面真正值得保留的情绪、动作、关系或视觉隐喻，再将其重新组织成一枚具有原创概念的**伪3D圆角图标视觉**。不要把照片简单缩进图标，而是以最少但最关键的元素重建原图的“灵魂”。 图标外轮廓必须始终保持为**标准1:1正方形圆角矩形**，四边等长、圆角统一、正面平视、比例稳定。圆角方形只是一个固定的视觉舞台，所有创意、空间和叙事都发生在这个边界内部。 图标内部采用**正面平视的伪3D / 2.5D空间语言**。主体为绝对视觉核心，通过真实的前后遮挡、体积厚度、高低层次、柔和切面、局部浮雕和空间距离形成丰富立体感。 配色**完全从上方原始照片中推导**。先识别照片中的主色、亮色以及辅助色，再在保持原有色相记忆的基础上进行重新调制：适度**提高明度、提升纯净度、去除灰脏杂色**。最终颜色可以比原照片更鲜活，但仍应让人自然感受到它来源于原图。 质感采用**细腻、哑光、柔和的现代伪3D雕塑感**。形体有真实体积和柔和层次，但不过分光滑、反光或CG化。下半部分外围保持大面积干净留白，只展示这一枚完整的圆角方形图标。 😆😆😂

Goal: Create a polished desktop browser screenshot mockup of an AI image prompt gallery web app called {argument name="site title" default="Images AI Prompt Gallery | Inspia"}, shown floating over a blurred warm brown/orange background. Canvas: 1200×831 landscape image. Center a large rounded-corner browser window with a black top bar, subtle drop shadow, and macOS-style red, yellow, green window buttons at the top left. The browser occupies most of the frame, with soft orange blurred photographic shapes visible behind it. Layout: The web app is split into 2 main panels. Left panel is a large image preview area on an off-white to pale beige background, containing one square artwork with rounded corners. Right panel is a white details sidebar with user information, prompt text, a thumbnail grid, and a bottom action button. Top browser bar details: Show exactly 3 colored window dots on the far left, a small back chevron, the title text “Images AI Prompt Gallery | Inspia”, and exactly 2 icons on the far right: a plus sign and a hamburger menu. Left image preview: Display one large square horror portrait, nearly black-and-white, centered in the left panel. The portrait is an extreme close-up split-face image of a young woman: the left half is a human female face with pale skin, straight black bangs, a single intense dark eye with a tiny red reflection, dark under-eye makeup, and black lipstick; the right half transforms into a monstrous demon jaw with torn glossy black skin, cracked veiny texture with subtle red fissures, many sharp jagged teeth, and a gaping black mouth. Use dramatic high contrast, cinematic lighting, realistic texture, and a horror character concept art feel. Add one small white square close button with an “×” near the top-right of the left panel. Right sidebar details: At the top show a circular profile avatar, the creator name {argument name="creator name" default="Shahid Wani"} with a small external-link arrow, the handle “@meng_dog695”, and a small pill badge labeled {argument name="model badge" default="GPT Image"}. Under it show exactly 2 action controls: a heart icon and “Copy prompt”. Add the category line “Images · Horror Portrait · Character Concept Art”. Then show a prompt excerpt beginning with: {argument name="prompt excerpt" default="Extreme close-up split-face horror portrait, young woman's face on left half pale and human with sharp dark eyes and blunt bangs, right half transforming into monstrous demon jaw with rows of jagged fangs, torn skin revealing black void mouth, black and whit..."} followed by a “show more” label with a down chevron. Thumbnail section: Add the heading “Continue by camera language”. Below it create exactly 8 rounded image thumbnails arranged as a compact 3-column grid: row 1 has 3 thumbnails, row 2 has 3 thumbnails, row 3 has 2 thumbnails. The thumbnails should depict, in order: 1) two side-by-side female portrait close-ups, 2) an extreme close-up brown eye with tiny cartoon stickers on the eyebrow, 3) a seated man in cap beside a small cartoon doll, 4) a group portrait of multiple women in cream outfits, 5) a tall fashion photo of a person in black with vertical text, 6) a street-fashion woman with jeans and sunglasses beside a small doll, 7) a natural headshot of a curly-haired woman in a green top, 8) a woman posing with oversized cute plush characters. Bottom control: At the bottom of the sidebar place one wide black rounded pill button centered horizontally with white text {argument name="button text" default="Use as prompt"}. Visual style: Clean modern SaaS interface, crisp UI typography, realistic browser chrome, soft shadows, rounded corners, minimal iconography, strong contrast between the monochrome horror artwork and the bright white interface. Keep all UI text legible and do not add extra panels, thumbnails, buttons, or watermarks.

为这个 {argument name="location" default="客厅"} 提供 {argument name="number of ideas" default="5"} 个布局方案。为每个方案输出一张具有不同变化的图像。

{argument name="背景" default="Glass studio"}, {argument name="主体" default="holographic canvas"}, {argument name="灯光" default="mixed tungsten and cyan light"}.

Goal: Create a cinematic horizontal website/social banner about AI image generation and accurate text rendering, with a moody studio atmosphere and clear Portuguese typography. Canvas: Wide 2:1 banner, approximately 1200×600 px. Dark, soft-focus photographic background with strong depth of field. Leave the left half for text and the right half for the visual metaphor. Background and subject: A dim artist studio with cool gray-blue lighting, a blurred window on the far left, concrete wall, soft shadows on the floor, and an easel on the right holding a blank white canvas. In the upper center-right, show a delicate mechanical drawing arm or robotic brush holder reaching toward the canvas, holding a thin brush that has a single dark blue paint droplet at its tip. The canvas should be bright white and mostly empty, slightly angled, with a subtle shadow. Use realistic lighting, shallow depth of field, and a premium editorial tech aesthetic. Layout: Place all main text on the left side over a dark translucent gradient overlay for legibility. Use exactly 3 small rounded pill tags above the headline, aligned horizontally: “IA”, “MUSE IMAGE”, and “NOVIDADE”. Below them, place a large bold headline in two lines: “A IA aprendeu a pensar” on the first line and “antes de desenhar” on the second line. Highlight only the word “pensar” with a vivid purple rectangular label behind the word, white text on purple. Under the headline, add one smaller italic subtitle: “O texto deixou de ser o calcanhar de Aquiles. Falta todo o resto.” Add a small website URL in the bottom-right corner: “www.macareno.net”. Text content: Use {argument name="headline text" default="A IA aprendeu a pensar antes de desenhar"}, with the word {argument name="highlighted word" default="pensar"} highlighted in purple. Use {argument name="subtitle text" default="O texto deixou de ser o calcanhar de Aquiles. Falta todo o resto."}. Use {argument name="tag labels" default="IA, MUSE IMAGE, NOVIDADE"}. Use {argument name="website URL" default="www.macareno.net"}. Visual style: Minimal, polished, cinematic, realistic, high contrast between white typography and dark background, with a modern tech-news landing-page feel. Typography should be clean sans serif; headline extra bold, subtitle light italic, tags uppercase with letter spacing. Keep text sharp and perfectly spelled. Constraints: Include exactly 3 pill tags, exactly 1 highlighted word, exactly 1 subtitle line, exactly 1 URL, exactly 1 easel canvas, exactly 1 robotic drawing arm, exactly 1 brush, and exactly 1 visible paint droplet. Do not add extra logos, people, icons, or additional text.

Create a cheerful square 1:1 kawaii tech illustration showing an AI model switch-over machine. The scene is a rounded beige control panel on a warm cream background. On the upper left of the panel, place a teal circuit-board chip labeled {argument name="left chip label" default="Codex"}, with a glowing cyan brain-shaped circuit icon, four visible corner screws, and a teal cable plugged into its bottom. On the upper right, place an orange/gold circuit-board chip labeled {argument name="right chip label" default="GLM"}, with a glowing yellow brain-shaped circuit icon, four visible corner screws, and a yellow cable plugged into its bottom. Between the chips, draw one orange arrow pointing from the left chip to the right chip to indicate switching. The right chip should emit warm glow rays and celebratory orange marks. At the bottom center, show a physical lever console with the yellow cable routed into the lever slot; the lever is tilted toward the right and has an orange ball handle held by a cute small white robot on the right side. The robot has a glossy white rounded body, a black face screen with turquoise smiling eyes and mouth, teal ear-like side disks, a tiny antenna with a teal ball, dark blue joints, and a teal circular chest emblem with a simple smiling bot icon. Include exactly six round console buttons: two teal buttons on the left row, two gray buttons below them, one yellow button near the robot, and one orange button below it, plus one gray button further right. Add a potted green plant on the lower left, a teal mug with the same white smiling bot icon on the lower right, and a small teal star charm near the mug. Decorate the background with confetti strokes and sparkles in orange and teal. Use soft rounded shapes, clean outlines, gentle shadows, warm highlights, vector-like digital art, friendly startup mascot style, no photorealism, no extra text beyond the two chip labels.

2x2 grid, 16:9, do this for 4 tourist cities: input = [{argument name="location" default="city"} + {argument name="infrastructure type" default="infrastructure system"} + environment + {argument name="style" default="optional style"}] run infrastructural_typography_extraction: text_content := parse word_or_place word_identity := infer whether text is a city, brand, concept, landmark, district, event, or fictional place font_personality := infer heavy, elegant, brutalist, futuristic, coastal, industrial, organic, luxury, playful, or monumental letter style infrastructure_dna := infer visual grammar of infrastructure_system network_behavior := infer flow, looping, branching, stacking, crossing, tunneling, ramping, glowing, routing, or circulation logic letter_mapping := map infrastructure onto each glyph’s stems, bowls, counters, diagonals, curves, crossbars, and terminals material_system := infer realistic construction materials light_system := infer city lights, traffic trails, signage glow, window grids, reflections, or ambient illumination environment_logic := infer surrounding urban/landscape context scale_anchors := infer vehicles, people, trees, buildings, waterways, or transit elements camera := infer cinematic aerial, street-level, drone, tilt-shift, or architectural visualization view render: a readable monumental word built from the inferred infrastructure system. constraints: - no fixed city - no fixed road system - no copied composition - no random transport spaghetti - infrastructure must physically construct the typography - word remains readable at thumbnail scale

Create a photorealistic architectural interior rendering of a modern Japanese-inspired living room and small work area, viewed from a fixed wide-angle camera at standing eye level from the front-left corner looking diagonally into the room. The space has warm walnut wood flooring, smooth off-white plaster walls, a muted beige ceiling, and a dark charcoal stone-textured accent wall on the right. Along the left wall place exactly 6 tall narrow vertical windows with black frames, bringing in soft overcast daylight and faint suburban houses and greenery outside. In the foreground place a low, long L-shaped sofa in {argument name="sofa fabric color" default="warm taupe woven fabric"} with wooden legs and exactly 3 visible cushions: one large back cushion section, one chaise seat section, and two loose square throw pillows near the corner. In front of it place a rectangular mid-century wooden coffee table with rounded corners, tapered legs, a lower woven shelf, exactly 2 stacked books, and 1 small potted green plant. On the right accent wall mount 1 large thin flat-screen TV with a dark screen and slim gray side frame. Toward the back center, create a compact raised work nook partially hidden by a white structural wall, with exactly 1 wooden desk, 1 black ergonomic mesh office chair, 1 computer monitor, 1 small desk lamp, and warm wood flooring continuing through the area. At the rear-right, add a decorative partition of exactly 10 vertical wooden slats in front of a large black-framed window, with a small built-in shelf holding exactly 3 tiny decor items or plants. Use black ceiling track lighting with exactly 5 visible spotlights distributed along two tracks, turned on subtly but dominated by natural daylight. Overall style: {argument name="interior style" default="minimal warm modern Japanese with mid-century wood furniture"}, realistic materials, clean uncluttered staging, balanced neutral tones, soft shadows, high-end residential design visualization, no people, no text, no logos, no exaggerated decoration.

Goal: Create a dark-mode desktop web app screenshot for {argument name="app name" default="HeliosGen"}, an AI image/video generation dashboard, shown in a wide 16:9 browser-like canvas. Canvas: 1200×675 style screenshot, very dark navy/black interface with subtle dotted grid texture across the main workspace. Use a clean modern SaaS UI aesthetic, thin dividers, small muted gray text, and teal accents. Layout: A narrow left sidebar occupies about 10% of the width, with the main workspace filling the rest. At the top of the main workspace, show two small tabs: “Generated” selected and “Uploaded” inactive. Beneath the tabs, display exactly 5 generated image thumbnails in one horizontal row, aligned along the top, each separated by thin borders. Sidebar details: Top-left logo text reads “HeliosGen” with a small teal emblem. Include exactly 6 primary navigation items with icons: 1) Image selected, 2) Video, 3) Workflow, 4) Assets, 5) Chat, 6) Settings. Below that, show a small “Folders” section with exactly 1 folder row labeled “All works,” then an empty-state area reading “No chats yet.” At the bottom-left, add two tiny stats/icons resembling “1.8k” and “220.” Thumbnail row: Show exactly 5 thumbnails from left to right: 1) an ornate amber luxury perfume bottle on a dark rock in a golden desert at sunset, 2) a rounded amber perfume bottle with ornate gold cap on a stone slab in desert dunes at sunset, 3) a square amber perfume bottle with a crescent-shaped cap on a black rock in desert dunes at sunset, 4) a single red apple centered on an off-white studio background, 5) a smaller red apple sitting on a white table in a minimal white studio scene. The first three thumbnails should feel cinematic and warm; the last two should be minimal product-photo tests. Bottom prompt panel: Near the bottom center of the workspace, include a floating generation control panel with rounded corners. On the left of the panel, show a small square preview tile labeled “IMAGE 1:1.” In the main text area, include the prompt: “{argument name="generation prompt" default="Luxury Saudi perfume bottle on black stone, desert dunes in the background at sunset, premium commercial product"}”. Along the bottom of the panel, show compact controls reading approximately “Nano Banana Pro,” “2K,” “1:1,” “1/4,” “Multi,” and “JSON (4.14k),” plus a teal send/generate button at the right. Visual style: Crisp UI screenshot, high contrast, minimal glow, fine dotted background, realistic generated-image thumbnails, no browser chrome, no extra popups, no people, no watermark.

Goal: Create a clean screenshot-like composition showing an AI image editing variation of a simple bear character, with the smaller original/reference result above and a larger edited result below. Canvas: Wide horizontal canvas, approximately 1200×960 px, very light gray-white background with lots of empty space. Minimal interface feel, like a chat/image generation workspace. Layout: Place exactly 2 bear illustration cards on the canvas. Card 1 is small and positioned in the upper-right quadrant, with a rounded white image area. Card 2 is much larger, positioned lower-left/center, also inside a rounded white image area. Between them, slightly below the small top bear, place exactly 1 pale blue rounded chat bubble containing Japanese instruction text. On the large bottom image card, include exactly 2 small translucent gray UI controls: one rounded pill button at the bottom-left labeled 「編集」 and one circular icon button at the bottom-right with a simple upload/share arrow icon. Text content: The single chat bubble must read exactly: {argument name="edit instruction text" default="頬の横から首にかけてのあたりを、もっと細くしてください。"} Subject details: The subject is exactly 2 versions of the same cute, hand-drawn, upright bear-like character. The character has off-white fur, thick black sketchy outlines, rounded ears with small inner-ear marks, wide circular eyes with black pupils, a large black oval nose, a tiny straight horizontal mouth, pink circular blush marks on both cheeks, short arms hanging down, stubby legs, and small toe lines. The top bear is smaller and slightly slimmer with a more vertical body. The bottom bear is larger and centered in its card, with a rounded head and body, narrowed neck/cheek area, simple symmetrical pose, and a slightly uncertain blank expression. Visual style: Minimal doodle mascot illustration, rough black ink outline, subtle paper or canvas texture, soft off-white fill, pale pink cheek circles, no shadows except very faint card/UI softness. The overall image should look like a casual GPT image editing interface screenshot, not a polished poster. Constraints: Use exactly 2 bear illustrations, exactly 1 Japanese chat bubble, and exactly 2 UI overlay controls. Keep the background mostly empty, preserve the asymmetrical placement, avoid adding extra characters, extra text, watermarks, or decorative elements.

Create a realistic architectural visualization comparing two 3D-generated ottoman stools in an interior showroom scene. Canvas: wide landscape 16:9 render, camera at low eye level looking slightly downward. Scene: two small Barcelona-style footstools placed side by side on a dark charcoal tiled floor, in front of a tall glossy stone wall with dramatic brown, cream, rust, and gold onyx/marble veining; a narrow strip of green-tinted glass or window appears at the far right edge, with pale exterior floor tiles visible beyond. Subjects: exactly 2 footstools: left stool is the HD version, slightly smaller and lower, with a white tufted square cushion divided into a 4-by-4 grid, rounded soft edges, light gray leather straps underneath, and reflective polished metal X-shaped legs; right stool is the Smart Mesh version, slightly larger and cleaner, with a thicker smooth white cushion, parallel white underside straps, and gray polished X-shaped metal legs. Add thin black 3D viewport guide lines or wireframe axis lines crossing through and around both stools, including vertical and horizontal alignment lines, as if shown in a 3D modeling comparison viewport. Materials: white leather cushions with subtle seams and soft highlights, chrome legs with reflections, dark matte floor tiles with faint grout lines, glossy veined stone wall. Lighting: soft indoor studio lighting, moderate shadows, realistic but slightly CGI-rendered look. Composition: left stool around one-third of the frame, right stool around two-thirds, both facing forward at a slight angle, enough floor visible in front, marble wall filling the background. No people, no labels, no UI panels, no watermark.

Create a charming minimalist 3D miniature {argument name="subject" default="coffee shop"} scene inspired by a cozy handmade paper-and-clay aesthetic. Place a tiny, beautifully designed coffee store in the lower-center of a large textured off-white paper background, leaving plenty of clean empty space around it. The coffee shop should look like a handcrafted miniature café: a small cream-colored storefront with a warm wooden counter, tiny espresso machine, coffee cups, pastries, coffee beans, little menu board, hanging sign, potted plants, stools, and subtle café decorations. Add a few tiny customers enjoying coffee and a barista behind the counter, keeping everything cute, delicate, and uncluttered. Use soft ivory, beige, warm brown, muted sage green, and subtle terracotta accents, with realistic soft shadows and gentle natural lighting. Add tiny hand-drawn black doodles around the café—small hearts, steam lines, stars, coffee beans, arrows, and simple decorative marks—to create a playful contemporary scrapbook feeling. The overall composition should feel premium, cozy, whimsical, handcrafted, and Instagram/Pinterest aesthetic, with a tactile textured-paper background, miniature diorama appearance, soft 3D details, lots of negative space, and a peaceful cozy coffee-shop mood. Vertical 4:5 composition, highly detailed, clean, elegant, photorealistic miniature craft style.

Goal: Create a desktop screenshot-like collage showing 10 AI-generated recruitment poster variations for a Chinese tech company, all with nearly identical purple-and-white branding and a cute rabbit mascot, arranged above a prompt input bar. Canvas: Wide 16:9 white interface screenshot, about 1064×922 px. Leave generous white margins. The image area occupies the upper 80%, and a rounded prompt composer bar sits centered along the bottom. Main layout: Arrange exactly 10 square poster thumbnails in a loose grid with slight spacing, as if displayed in an image-generation results gallery. Count and placement of thumbnails: 3 posters across the top row; 4 posters across the middle row; 3 posters across the bottom row. Each poster is a clean Chinese recruitment flyer with the same core visual identity but small layout variations. Poster design shared by all 10 thumbnails: Use a light lavender-to-white background, deep purple and bright blue typography, rounded cards, small icons, thin dividers, soft shadows, and a polished SaaS/AI startup aesthetic. Every poster should feature the logo text {argument name="company name" default="兔子 AI"} near the top, a large bold Chinese headline {argument name="headline text" default="兔子 AI 继续招聘 全栈开发实习生"}, and multiple information blocks resembling a job description. Use readable-looking Chinese microtext but it may be partially too small to read at thumbnail size. Include iconography such as gift boxes, graduation caps, location pins, calendars, code brackets, laptop icons, check marks, and circular badges. Mascot subject: Include a cute white 3D rabbit mascot in most posters, with pink inner ears, glossy black eyes, a friendly smile, and a purple hoodie or shirt branded with the company name. Vary the mascot pose across the 10 posters: sitting with a laptop, waving, giving a thumbs-up, peeking from the bottom edge, wearing round glasses, and standing beside cards. The rabbit should feel like a friendly AI assistant mascot. Poster content blocks: Each thumbnail should show several compact recruitment sections, including exactly these recurring block types across the designs: recruitment target, requirements, job location, internship participation/work content, work arrangement, and benefits. Some posters may have a purple CTA strip or footer, small QR-code-like square, or decorative code panel, but keep all poster variants visually consistent. Bottom prompt composer UI: At the bottom center, draw a large rounded white input box with soft shadow, resembling an AI image generation chat bar. Inside it show the gray placeholder text {argument name="input placeholder" default="描述你想要创建的图片"}. Along the lower edge of the bar include small UI icons and controls: image icon, grid icon, book icon, @ symbol, a dropdown labeled image, a model label {argument name="model label" default="gpt2"} with tiny green signal bars, a dropdown reading auto / 1K / auto, a count selector showing {argument name="generation count" default="10个"}, and a paper-plane send button on the far right. Add a three-dot menu at the top-right of the composer. Visual style: Clean modern Chinese AI-product interface screenshot, soft pastel lavender, white, cobalt blue, and purple. The 10 posters should look like outputs from the same prompt: highly consistent layout, typography, colors, and mascot styling, with only minor differences in composition. Use crisp vector-like UI, subtle gradients, rounded rectangles, and high-resolution details. Constraints: The collage must contain exactly 10 poster thumbnails and one bottom prompt composer bar. Do not add people, photos, watermarks, browser chrome, or extra panels. Keep the overall impression that the AI generated many nearly identical recruitment poster options.

Goal: Create a realistic smartphone screenshot of a Xiaohongshu-style post featuring a simple two-panel hand-drawn comic about a tired young man trying outdoor activities, with engagement metrics circled in red. Canvas: Vertical mobile screenshot, 9:16 aspect ratio, white app interface, iPhone-like status bar at the top showing 08:34, signal bars, “4G”, and a battery icon. Top app header: Show a back arrow on the left, a small circular avatar, the username {argument name="account name" default="水枪装尿呲谁谁叫"}, a rounded pink-outline follow button labeled “关注”, and a share arrow icon on the right. Main post image: A large white comic card with a thick black rectangular border, divided into exactly 2 stacked panels by one horizontal black divider. Add a small gray rounded page indicator in the top-right corner of the comic card reading “2/11”. Use loose black ink line art with muted watercolor fills, casual sketch style, lots of white space. Comic panel 1: At the top left, handwritten red-and-black Chinese text reads {argument name="first panel text" default="逛洗浴、露营,"}. In the center, draw a slim tired young man with short messy black hair, pale face, droopy eyes, dark green jacket, gray shirt, and dark pants, hands in pockets. Around him place exactly 3 simple activity icons: a steaming bath tub on the left, a hanging towel at lower left, and a camping tent on the right. Comic panel 2: At the top left, handwritten red-and-black Chinese text reads {argument name="second panel text" default="钓鱼、爬山……"}. Draw the same tired young man centered slightly right, holding a fishing rod in one hand with the line and hook hanging down, the other hand in his pocket. Around him place exactly 3 simple outdoor icons: mountains with a tiny cloud at upper left, the fishing rod and hook extending from his hand toward the left, and a backpack on the right. Below the comic: Add a carousel dot indicator with exactly 4 dots, the second dot red and the other 3 gray. Then show the bold caption {argument name="caption text" default="这日子真是越过越有盼头"}. Below it, show exactly 5 hashtag phrases in blue-gray Chinese: “#漫画”, “#当漫画照进生活”, “#原创漫画”, “#突然的绷不住”, “#每天一个治愈小漫画”. Add a rounded search suggestion bar with a magnifying glass icon and Chinese text beginning with “猜你想搜｜日子过得越来越有盼头了”. Bottom interaction area: Add a pale comment input pill on the left with a pencil icon and placeholder “说点什么...”. On the right, show exactly 3 interaction icons with counts: a heart icon with {argument name="like count" default="1703"}, a star icon with “408”, and a comment bubble icon with “1003”. Draw a thick freehand red circle around these three engagement metrics, as if marked with a digital pen. Visual style and constraints: Keep the screenshot clean and realistic, with crisp UI elements and a slightly imperfect hand-drawn comic. Use black, white, muted green, gray, pale pink, and red accents. Preserve all visible Chinese UI text exactly where specified. Do not add extra panels, extra icons, watermarks, or unrelated decorations.

请将上传的每一张旅行照片分别制作成一张独立的3:4竖版“复古像素工业旅行档案”编辑海报。每张照片单独输出，不进行多图拼贴。 这套风格不是把整张照片直接像素化，也不是给照片套上普通电脑边框。必须先从原照片中拆分出3至7个决定场景身份的真实对象，再将它们转化为像素精灵、局部预览、数据索引和结构切片，并放入一套完整运行的旧式电脑软件中重新拼装。 上方约占画面30%至38%，保留清楚可辨的原始照片。准确保持主体身份、建筑结构、数量、观察角度、空间关系、真实材质、自然光影和地点色彩。仅进行轻微艺术出版物调色与薄胶片颗粒处理，不得插画化、重构或改变原照片。 照片下方保留一条约7%至10%的炭黑或暗橄榄色呼吸带，在其中加入一个1至4个英文单词的可读像素标题。标题应来自照片的真实结构或主题，例如： HEAVEN AXIS ROOF SYSTEM LAKE PAVILION CANAL OBJECTS 下方约占画面50%至60%，构建一套完整的20世纪80年代末至90年代初图形工作站软件。主底色必须是肉眼可辨的深森林绿、墨青绿或褪色橄榄绿，不能铺成纯黑。 电脑区域必须同时具备： 1. 一个占据下方主要面积的完整主应用窗口； 2. 旧象牙色或烟草金色窄边框； 3. 横贯整个应用顶部的系统菜单栏； 4. 菜单文字逐字使用： FILE EDIT VIEW DATA TOOLS WINDOW HELP 5. 至少一个被选中的菜单； 6. 一个与选中菜单直接相连的展开式下拉菜单； 7. 下拉菜单可使用： IMPORT… EXPORT… LAYERS… FILTER… COLOR MAP STATISTICS… PRINT… 8. 至少三个互相遮挡的功能子窗口； 9. 每个子窗口都有标题栏、关闭按钮和明确功能； 10. 至少一个窗口拥有清楚的滚动条、箭头按钮和滑块； 11. 至少一个对象索引、色卡表、柱状图或状态窗口。 先分析照片并列出可以拆分的对象。例如： 传统建筑照片： 主屋顶、次屋顶、斗拱色带、山墙、台阶、栏杆、红墙。 园林照片： 亭阁、山石、湖岸、游船、树枝、屋顶碎片。 水乡照片： 石桥、临水木屋、阳台、瓦顶、水面、倒影、岸线。 人物或交通照片： 人物、车辆、道路、建筑立面、标志物和动作方向。 将这些真实元素分别转换为低分辨率像素精灵或结构模块。保留对象的轮廓、方向、数量和标志色，但删除微小纹理。中央工作区重新调整对象的大小和位置：主对象可以被放大成为视觉锚点，其他对象像可以移动的图层一样围绕它排列。 不要沿用原…

Create a wide cropped comparison graphic showing two side-by-side image-generation quality results on a plain white background. The canvas is an ultra-wide horizontal strip, about 5:1 aspect ratio, cropped tightly so only the top portion of each result is visible. At the top left above the first crop place the bold black uppercase label “MEDIUM”; near the upper center above the second crop place the bold black uppercase label “HIGH”. Each of the 2 comparison panels shows the same fitness-class style tiled layout: a photographic band at the top with gym equipment and dark exercise machines, underneath a deep teal rectangular title bar with large bold white Japanese text {argument name="main tile text" default="有酸素"}. Use thick white borders and sharp diagonal white dividers between tiles, giving the layout a dynamic slanted collage look. Count exactly 2 main visible teal title tiles with the full text “有酸素”, one in each panel, plus exactly 2 partially visible neighboring teal tiles at the right edge showing only the first character or fragment {argument name="partial tile text" default="ス"}. A thin lower white divider and the top edge of another teal row are barely visible along the bottom, cropped off. Emphasize the artifact being compared: around the white Japanese letters on the solid teal backgrounds, include subtle uneven brightness and saturation patches, faint haloing, and mottled texture, with the “HIGH” panel slightly cleaner but still showing visible nonuniformity. Keep the composition as a technical screenshot/comparison crop, not a polished poster. Use crisp sans-serif typography, high contrast, no extra captions, no watermark.

Create a centered, high-end 3D render of a single translucent glass orb on a pure white studio background. The orb is a perfectly smooth sphere made of thick glossy crystal, with a dark smoky interior and layered, flowing ribbon-like waves suspended inside it. The internal ribbons curve horizontally in elegant S-shaped bands: electric cyan and deep blue across the upper half, molten gold and orange through the middle, hot magenta and violet through the lower half. Add bright neon light streaks, tiny star-like glints, and caustic refractions following the curves, as if glowing energy is trapped inside the glass. Use {argument name="orb material" default="transparent glossy crystal glass"}, {argument name="main color palette" default="electric blue, cyan, gold, orange, magenta, violet"}, and {argument name="background color" default="pure white"}. Include large soft white specular highlights near the upper left and upper right of the sphere, a crisp luminous rim around the edge, subtle internal reflections, and a soft blurred colored shadow underneath in blue, purple, and pink. The composition should be minimal and product-shot-like, with the sphere occupying most of the square frame, no text, no logos, no extra objects, ultra-detailed, polished, futuristic, luminous, photorealistic, octane-render style.

Genera una imagen 16:9 de una captura de pantalla oficial de {argument name="operating system" default="Windows 12"}

Goal: Create a clean mobile-style social media screenshot demonstrating an AI background-removal result for a thumbnail image. Canvas: Vertical 488×641 px composition on a pure white background, with generous margins and a centered content column. Layout: At the top, place one rounded-corner rectangular photo, about 318 px wide, showing three Japanese schoolgirls in sailor-style uniforms in a warm sunset urban street scene. The central girl is closest to camera, smiling brightly and raising her right hand beside her face in a salute-like pose; the two other girls are behind her, one on the left partly cropped and looking serious, one on the right in the background. Use cinematic dusk lighting with orange highlights and shallow depth of field. Prompt bubble: Directly below the top photo, add one pale light-blue rounded speech bubble containing Japanese text in dark navy. Preserve the text exactly as: 「@画像を作成する 中央の女性だけをそのポーズのまま切り抜く。背景は透過させる(アルファチャンネル)。高画質化。クリアでシャープな画像。JPEGのノイズとアーティファクトを軽減。16:9で、肘の先まで納める。」 The bubble should be aligned under the image and look like a chat prompt. Result image: Below the bubble, show one large cut-out portrait of only the central girl from the top image, isolated against a white/transparent-looking background. She remains in the same salute pose, smiling, with twin ponytails, straight bangs, a pale pink sailor uniform, navy sailor collar with white stripes, and a red necktie. Re-light her cleanly and evenly so the face is sharper and brighter than in the top photo. Crop from upper torso to just past the raised elbow, with a soft rounded gray bottom container edge like an image preview. Add one small circular gray share/download button at the bottom-right corner of the result preview. Bottom UI: Along the lower left margin, draw exactly five small gray interaction icons in a row: copy/duplicate, thumbs up, thumbs down, share/upload arrow, and ellipsis menu. Customizable subject details: The main cutout subject is {argument name="character description" default="a smiling Japanese teenage schoolgirl with twin ponytails and straight bangs"}; the uniform is {argument name="outfit" default="pale pink sailor-style school uniform with navy collar, white stripes, and red necktie"}; the top background is {argument name="background scene" default="warm sunset urban street with cinematic bokeh"}; the chat bubble text is {argument name="prompt text" default="@画像を作成する 中央の女性だけをそのポーズのまま切り抜く。背景は透過させる(アルファチャンネル)。高画質化。クリアでシャープな画像。JPEGのノイズとアーティファクトを軽減。16:9で、肘の先まで納める。"}. Visual style: Photorealistic portraits embedded in a minimal app interface, crisp shadows, rounded corners, clean white UI, high-resolution thumbnail demonstration. Avoid extra people beyond the three in the top image and the single isolated cutout below; avoid watermarks or additional text.

Create a simple centered children's-book illustration of a cute full-body {argument name="animal" default="koala"} character on a plain white background. Use a hand-drawn crayon or thick marker style with rough, slightly wobbly black outlines and soft textured gray fill. The character faces forward, standing upright with a large rounded head, rounded body, short arms hanging at the sides, and short legs. Include exactly 2 large round ears, one on each side of the head, each with a black inner ear patch and no pink inner-ear color. The face has exactly 2 wide white circular eyes with tiny black pupils, 1 large vertical oval black nose centered between them, and 1 small curved smiling mouth below the nose. The body is {argument name="body color" default="warm medium gray"}, the outline is {argument name="outline color" default="black"}, and the overall mood is friendly, naive, and toddler-like. Keep the composition uncluttered, with no text, no shadows, no props, no scenery, and no extra characters.